DETAILED ACTION
Notice of Pre-AIA or AIA Status
In the present application, filed on or after March 16, 2013, claims 1, 3-5, 7-13, 17-19, 23-24, 26-29, and 31 have been considered and examined under the first inventor to file provisions of the AIA .
Respond to Applicant’s Arguments/Remarks
Applicant’s arguments, see Remarks, filed 01/07/2026, with respect to the rejection(s) of claims 1, 3-5, 7-13, 15-21, 23-29, and 31-32, based solely on the claimed limitations as amended, have been fully considered but are moot because the arguments do not apply to the new combination of references including prior art being used in the current rejection (see below for detail) under new grounds of rejection, necessitated by amendment.
Claim Objections
The followings are reasons for claim objections:
Claims 5, 11, 19, and 23 are objected to because each claim of claims 5, 11, 19, and 23 indicates as “Currently Amended” without any underlines, strikethrough, or bracket. For the purpose of examination, Examiner interprets each claim dependency as indicated in the currently filed claims.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 3-5, 11, 13, 17, 24, 26-29, and 31 are rejected under 35 U.S.C. 103 as being unpatentable over Lustig et al. (Lustig – US 2019/0175076 A1) in view of Tokuchi et al. (Tokuchi – US 2019/0243958 A1), Karunaratne et al. (Karunaratne – US 9,795,322 B1), Sonenblum et al. (Sonenblum – US 10,357,186 B2), and Bourahmoune et al. (Bourahmoune - AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion). In this rejection, Examine relied on the publication date of Twenty-Eighth Internation Joint Conference on Artificial Intelligence (IJCAI-19) August 10-16, 2019).
As to claim 1, Lustig discloses a posture detection system for detecting a user’s posture comprising:
a pressure sensor unit including a plurality of sensors (Lustig: [0018], [0043], [0051], [0059], [0061], [0064], FIG. 1 the pressure sensor matrix 120, and FIG. 6 the pressure matrix sensor 610: The sensors may include sensors useful for determining pressure exerted on the chair at different locations. In particular, a plurality of pressure points may be monitored in a seat of the chair and a plurality of pressure point may be monitored in a backrest of the chair. In some exemplary embodiments, a matrix of 3×3 or more pressure points may be used for the seat, the backrest, or for both), each of the sensors being configured to detect a pressure applied from a user (Lustig: [0018], [0043], [0051], [0059], [0061], [0064], FIG. 1 the pressure sensor matrix 120, and FIG. 6 the pressure matrix sensor 610: Pressure Sensor Matrix 120 may comprise an N×M matrix of pressure sensors enabling sensing pressure excreted by a person sitting on seat and leaning back against backrest. The matrix may comprise 3×3 sensors, or more, such as 4×4, 6×6, 8×6, 10×8, or the like);
a controller (Lustig: FIG. 6 the processor 602) comprising a supervised machine learning (ML) classifier (Lustig: Abstract, [0043]-[0046], [0077], [0105], [0114], and FIG. 3-5: On Step 310, posture of the person may be estimated based on the sensor input. The estimation may be performed by the processor using a supervised classifier, such as a k-means, Support Vector Machines (SVM), or the like, which may be trained to deduce a posture based on input based on training data provided thereto. The training data may comprise sensor readings and correct labels thereof, indicating the posture. The posture estimation may be performed on the apparatus itself without requiring computation by an external computation platform, such as a cloud-based server. In some exemplary embodiments, the classifier may be trained offline and the trained model may be provided to the apparatus to be used locally. In some exemplary embodiments, the potential posture for estimation may include, for example, forward sloping, slump, side reliance, cross legged, no legs support, correct posture with back support, correct posture without back support, standing, or the like. Each posture may be associated with a different severity measurement, a different alleviating stretches or exercises, or the like) and configured to:
receive, in real-time, raw pressure data detected by the pressure sensor unit (Lustig: [0061], [0064], [0076]-[0077], [0108], [0110], and FIG. 6: Processor 602 may be configured to process sensor readings and select a feedback device to provide feedback to the person sitting on the chair);
classify the user’s posture and at least one user behavior (Lustig: Abstract, [0043]-[0046], [0077], [0105], [0114], and FIG. 3-5: On Step 310, posture of the person may be estimated based on the sensor input. The estimation may be performed by the processor using a supervised classifier, such as a k-means, Support Vector Machines (SVM), or the like, which may be trained to deduce a posture based on input based on training data provided thereto. The training data may comprise sensor readings and correct labels thereof, indicating the posture. The posture estimation may be performed on the apparatus itself without requiring computation by an external computation platform, such as a cloud-based server. In some exemplary embodiments, the classifier may be trained offline and the trained model may be provided to the apparatus to be used locally. In some exemplary embodiments, the potential posture for estimation may include, for example, forward sloping, slump, side reliance, cross legged, no legs support, correct posture with back support, correct posture without back support, standing, or the like. Each posture may be associated with a different severity measurement, a different alleviating stretches or exercises, or the like) other than posture based on the raw pressure data using the ML classifier to generate a classification (Lustig: [0112]-[0116] and FIG. 7: Current sitting time is displayed (702) together with a visual indication of aggregated posture information over time (706, 708)… Another user interface illustration is shown in FIG. 7B, which shows real time pressure map. Pressure maps corresponding to pressure sensor matrices are shown. Pressure Map 720 shows pressure sensed on backrest…aggregated information is shown relating sitting time and times in which the different postures were detected (730). A distribution of the different postures over time is visually show. In some cases, the postures may be abstracted into groups of Good, Bad, Break and Exercise…Animation 740 may be displayed showing how the person should exercise. A Visual Timer 742 may be displayed and used to time the exercise time of the person and indicate when the person may finish exercising); and
determine whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose and a posture pressure distribution (Lustig: Abstract, [0016], [0051], [0080], and FIG. 3),
a haptic feedback configured to provide feedback to the user by vibrating based on a result of the classification (Lustig: [0008], [0045], [0049], [0062], FIG. 1 the haptic feedback devices 130: the device may perform posture estimation locally on-device and may select a feedback from the local feedback devices, such as integrated vibration motors embedded with the device itself); and
a display unit (Lustig: FIG. 6 the user device 650) configured to perform a display according to the result of the classification (Lustig: [0051], [0112]-[0116], and FIG. 7: FIG. 7A shows an illustration of a user interface of an application program, in accordance with some exemplary embodiments of the disclosed subject matter. Current sitting time is displayed (702) together with a visual indication of aggregated posture information over time (706, 708). For example, indication of aggregative time in the last time window of an hour is shown for good posture (708) and rest time (706). Real-time message (710) is displayed if current estimated posture is incorrect. Real-time message (710) may indicate area of the body where the posture is incorrect (712)).
Lustig does not explicitly disclose
classify the user’s posture and at least one user behavior other than posture based on the raw pressure distribution data using the ML classifier to generate a classification; and
determine whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and a posture pressure distribution,
wherein the controller is further configured to select the recommended pose as a stretch pose from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose,
wherein the pose pressure distribution serves as an input to a classifier that is learned by a supervised machine learning tool and outputs, in real-time, a behavior label, which classifies a user behavior other than the posture of the user in real-time; and
wherein the controller is configured to generate the user’s posture label and the behavior label directly from the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling.
However, it has been known in the art of determining postures of a user to implement the controller further configured to determine whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and a posture pressure distribution, as suggested by Tokuchi, which discloses the controller further configured to determine whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and a posture pressure distribution (Tokuchi: Abstract, [0017], [0019]-[0024], [0029]-[0031], [0034], [0037], [0041]-[0042], and FIG. 6-8: In each case when the user places the sheet-shaped apparatus 100 on the same seat and sits in the seat, the pressure distribution sensed by the pressure distribution sensor 102 disposed on the body side falls within a certain range, although some variations may occur due to slight differences in sitting posture for individual occasions when the user sits in the seat, and likewise, the pressure distribution sensed by the pressure distribution sensor 102 disposed on the seat side falls within a certain range (e.g., assuming that the user sits back in the seat with correct posture in each compared case). Therefore, for the same user sitting in the same seat, the combination of the body-side and seat-side pressure distributions acting on the sheet-shaped apparatus falls within a certain range. Now, a case is considered in which the same user carries the sheet-shaped apparatus 100, and uses the sheet-shaped apparatus 100 by placing the sheet-shaped apparatus 100 on various seats in various places. In this case, if the material of the seating portion of a seat that the user sits on or the shape of the seating surface of the seat changes, both the pressure distribution sensed by the pressure distribution sensor 102 disposed on the body side and the pressure distribution sensed by the pressure distribution sensor 102 disposed on the seat side also change, and so does the combination of pressure distributions on the two sides).
Therefore, in view of teachings by Lustig and Tokuchi, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig to include the controller further configured to determine whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and a posture pressure distribution, as suggested by Tokuchi. The motivation for this is to determine a user posture based on pressure distribution information.
While the combination of Lustig and Tokuchi discloses the pressure data/distribution for training supervised machine learning to identify postures of a user (Lustig: [0064], [0111]-[0116], FIG. 3-5, and FIG. 7: It will be noted that the measured pressure may change over time not only based on changes in weight, but also on changes in weight distribution. If the person puts his leg on the ground, not all of his weight is distributed on the pressure sensors. Monitoring the readings from the pressure sensors over time may be useful in estimating actual weight based on the different pressure measurements and distributions thereof. In some exemplary embodiments, training datasets may be provided based on users' that explicitly provided their physical measurements, and the information may be used to train a supervised machine learning classifier to estimate a weight of another person based solely on the sensor readings), the combination of Lustig and Tokuchi does not explicitly disclose wherein the pose pressure distribution serves as an input to a classifier that is learned by a supervised machine learning tool and outputs, in real-time, a behavior label, which classifies a user behavior other than the posture of the user in real-time.
However, it has been known in the art of monitoring sitting postures of user to implement wherein the pose pressure distribution serves as an input to a classifier that is learned by a supervised machine learning tool and outputs, in real-time, a behavior label, which classifies a user behavior other than the posture of the user in real-time, as suggested by Karunaratne, which discloses wherein the pose pressure distribution (Karunaratne: column 9 lines 51-column 10 lines 26, column 13 lines 49-column 14 lines 21, column 15 lines 4-34, column 24 lines 63-column 25 lines 19, and FIG. 1-3: In this embodiment, there are seven pressure sensors and one angle sensor for capturing the pressure distribution on a smart seat cover with both a backrest portion 328 and a seat portion 348. In other embodiments, any number of pressure or angle sensors may be present) serves as an input to a classifier that is learned by a supervised machine learning tool and outputs, in real-time, a behavior label (Karunaratne: column 26 lines 29 - column 28 lines 5 and FIG. 19-21: Initial training data may be fed to the system manually by professionals via a web-based admin panel for one or more machine learning or other statistical algorithms 2120, which may be Naïve Bayes classification, K-Means, Support Vector Machine (SVM), a hybrid approach using multiple algorithms, or any other algorithm. Then, current and historical data 2110 from pressure sensor readings, angle sensor readings, health sensor readings, and behavioural sensor readings may be fed to the machine learning algorithms and medical predictions 2130 may be made based on the user's sitting behaviours and/or other health and behavioural data from the sensor readings), which classifies a user behavior (column 8 lines 5-8, column 9 lines 18-47, column 9 lines 64- column 10 lines 26, column 12 lines 5-44, column 23 lines 4-54, column 22 lines 15-53, FIG. 3-13 and FIG. 19-21: The backrest portion and the seat portion may be part of a contiguous arrangement, or may be two separate components held together by another attachment means. In some embodiments, the system may further comprise health sensors that monitor health indicators such as heart rate, blood pressure, and various hormone levels, and behavioural sensors, which may, independently or together with the health sensors, monitor stress levels as well as other salient behavioural features) other than the posture of the user in real-time (column 9 lines 64 - column 10 lines 26, column 19 lines 9-31, column 20 lines 66-column 21 lines 23, and FIG. 3-13: Again, such automatic notifications do not interfere with the user's activities, while providing constant, real-time feedback to ensure that posture improvements occur continuously but also progressively. Moreover, in some embodiments, the smart seat cover system may perform behaviour detection over time via health sensors and behavioral sensors to monitor health metrics that may collectively indicate the user's general state of health).
Therefore in view of teachings by Lustig, Tokuchi, and Karunaratne, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig and Tokuchi to include wherein the pose pressure distribution serves as an input to a classifier that is learned by a supervised machine learning tool and outputs, in real-time, a behavior label, which classifies a user behavior other than the posture of the user in real-time, as suggested by Karunaratne. The motivation for this is to identify sitting posture of a user and discourage sedentary behavior to minimize associated adverse health effects.
While the combination of Lustig, Tokuchi, and Karunaratne discloses the pressure data/distribution for training supervised machine learning to identify postures of a user (Lustig: [0064], [0111]-[0116], FIG. 3-5, and FIG. 7: It will be noted that the measured pressure may change over time not only based on changes in weight, but also on changes in weight distribution. If the person puts his leg on the ground, not all of his weight is distributed on the pressure sensors. Monitoring the readings from the pressure sensors over time may be useful in estimating actual weight based on the different pressure measurements and distributions thereof. In some exemplary embodiments, training datasets may be provided based on users' that explicitly provided their physical measurements, and the information may be used to train a supervised machine learning classifier to estimate a weight of another person based solely on the sensor readings), the combination of Lustig, Tokuchi, and Karunaratne does not explicitly disclose wherein the controller is configured to generate the user’s posture label and the behavior label directly from the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling.
However, it has been known in the art of monitoring sitting postures of user to implement wherein the controller is configured to generate the user’s posture label and the behavior label directly from the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling, as suggested by Sonenblum, which discloses classify the user’s posture and at least one user behavior other than posture based on the raw pressure distribution data using the ML classifier to generate a classification (Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: In other embodiments, the user can set goals 428 (e.g., change from left-leaning position to right-leaning position every 15 minutes) and the system can provide active feedback 432 to the user to assist the user in achieving such goals. Such feedback can include verbal indications and alarms transmitted to the user or to a clinician via a mobile app when the use data indicate that a characterized weight shifting movement event has not occurred within the amount of time set as a goal); and
wherein the controller is configured to generate the user’s posture label and the behavior label directly from the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling (Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: The seat sensors are placed based upon human anthropometry to accurately measure load and load distribution on the cushion by the wheelchair user during both static and dynamic seated postures. An algorithm capable of accommodating system creep over long durations of use can maintain the ability to accurately classify weight shift and in-seat movement activity. An algorithm is capable of classifying a continuum of weight-shifting activities ranging from partial load distribution resulting from leaning and reaching to complete unweighting of the cushion surface. The classification algorithm takes into account both magnitude and duration of loading to better distinguish different weight-shifting behaviors. Also, an algorithm can be used that is capable of monitoring the in-seat movement of a wheelchair users that results from transient re-distribution of cushion loading during dynamic changes in seated posture).
Therefore in view of teachings by Lustig, Tokuchi, Karunaratne, and Sonenblum, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, and Karunaratne to include wherein the controller is configured to generate the user’s posture label and the behavior label directly from the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling, as suggested by Sonenblum. The motivation for this is to identify sitting posture of a user and provide suggestions to the user based on characterization data.
The combination of Lustig, Tokuchi, Karunaratne, and Sonenblum does not explicitly disclose wherein the controller is further configured to select the recommended pose as a stretch pose from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose.
However, it has been known in the art of monitoring postures of users to implement wherein the controller is further configured to select the recommended pose as a stretch pose from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose, as suggested by Bourahmoune, which discloses wherein the controller is further configured to select the recommended pose as a stretch pose from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose (Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion: the heatmap for the stretch pose ‘Right Arm Cross’ (RAC) shows a high pressure reading on the upper left sensor. This captures accurately the pose performed by the experiment subjects where they extended the right arm across the chest and pressed the left arm on the right elbow. The pressure distribution heatmaps show that the six stretches are to some degree visually distinguishable from each other. Unlike sitting posture, a stretch is a static event that requires a conscious and dedicated effort from the user. Subjectivity in stretch interpretation and variable physical predisposition to perform the stretches result in greater inter-individual variability compared to the sitting posture recognition task).
Therefore in view of teachings by Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, Karunaratne, and Sonenblum to include wherein the controller is further configured to select the recommended pose as a stretch pose from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose, as suggested by Bourahmoune. The motivation for this is to identify sitting posture of a user and provide suggestions to the user based on characterization data.
As to claim 3, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune disclose the limitations of claim 1 further comprising the posture detection system according to Claim 1, wherein the pressure sensor unit is configured to detect a pressure applied from the user’s back, bottom or thighs (Lustig: [0018], [0043], [0051], [0059], [0061], [0064], FIG. 1 the pressure sensor matrix 120, and FIG. 6 the pressure matrix sensor 610: Pressure Sensor Matrix 120 may comprise an N×M matrix of pressure sensors enabling sensing pressure excreted by a person sitting on seat and leaning back against backrest. The matrix may comprise 3×3 sensors, or more, such as 4×4, 6×6, 8×6, 10×8, or the like and Tokuchi: Abstract, [0017], [0019]-[0024], [0029]-[0031], [0034], [0037], [0041]-[0042], and FIG. 6-8: In each case when the user places the sheet-shaped apparatus 100 on the same seat and sits in the seat, the pressure distribution sensed by the pressure distribution sensor 102 disposed on the body side falls within a certain range, although some variations may occur due to slight differences in sitting posture for individual occasions when the user sits in the seat, and likewise, the pressure distribution sensed by the pressure distribution sensor 102 disposed on the seat side falls within a certain range (e.g., assuming that the user sits back in the seat with correct posture in each compared case). Therefore, for the same user sitting in the same seat, the combination of the body-side and seat-side pressure distributions acting on the sheet-shaped apparatus falls within a certain range. Now, a case is considered in which the same user carries the sheet-shaped apparatus 100, and uses the sheet-shaped apparatus 100 by placing the sheet-shaped apparatus 100 on various seats in various places. In this case, if the material of the seating portion of a seat that the user sits on or the shape of the seating surface of the seat changes, both the pressure distribution sensed by the pressure distribution sensor 102 disposed on the body side and the pressure distribution sensed by the pressure distribution sensor 102 disposed on the seat side also change, and so does the combination of pressure distributions on the two sides, Karunaratne: column 9 lines 51-column 10 lines 26, column 13 lines 49-column 14 lines 21, column 15 lines 4-34, column 24 lines 63-column 25 lines 19, and FIG. 1-13: In this embodiment, there are seven pressure sensors and one angle sensor for capturing the pressure distribution on a smart seat cover with both a backrest portion 328 and a seat portion 348. In other embodiments, any number of pressure or angle sensors may be present, Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: The seat sensors are placed based upon human anthropometry to accurately measure load and load distribution on the cushion by the wheelchair user during both static and dynamic seated postures. An algorithm capable of accommodating system creep over long durations of use can maintain the ability to accurately classify weight shift and in-seat movement activity. An algorithm is capable of classifying a continuum of weight-shifting activities ranging from partial load distribution resulting from leaning and reaching to complete unweighting of the cushion surface. The classification algorithm takes into account both magnitude and duration of loading to better distinguish different weight-shifting behaviors. Also, an algorithm can be used that is capable of monitoring the in-seat movement of a wheelchair users that results from transient re-distribution of cushion loading during dynamic changes in seated posture, and Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion).
As to claim 4, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune disclose the limitations of claim 1 further comprising the posture detection system according to Claim 1, wherein the pressure sensor unit is provided in a backrest and is configured to detect a pressure applied from the user’s back (Lustig: [0018], [0043], [0051], [0059], [0061], [0064], FIG. 1 the pressure sensor matrix 120, and FIG. 6 the pressure matrix sensor 610: Pressure Sensor Matrix 120 may comprise an N×M matrix of pressure sensors enabling sensing pressure excreted by a person sitting on seat and leaning back against backrest. The matrix may comprise 3×3 sensors, or more, such as 4×4, 6×6, 8×6, 10×8, or the like and Tokuchi: Abstract, [0017], [0029]-[0031], [0034], [0037], [0041]-[0042], [0058], FIG. 6-8, and FIG. 10: When in use, the sheet-shaped apparatus 100 is placed over the area of a chair 20 from a seating portion 22 to a backrest 24 as illustrated in FIG. 10. In FIG. 10, the sheet-shaped apparatus 100 is depicted thicker than in reality for easy recognition, Karunaratne: column 9 lines 51-column 10 lines 26, column 13 lines 49-column 14 lines 21, column 15 lines 4-34, column 24 lines 63-column 25 lines 19, and FIG. 1-13: In this embodiment, there are seven pressure sensors and one angle sensor for capturing the pressure distribution on a smart seat cover with both a backrest portion 328 and a seat portion 348. In other embodiments, any number of pressure or angle sensors may be present, and Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion).
As to claim 5, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune disclose the limitations of claim 4 further comprising the posture detection system according to Claim 4, further comprising a seating face sensor unit provided in the user’s seating face and configured to detect the pressure applied from the user’s bottom (Lustig: [0018], [0043], [0051], [0059], [0061], [0064], FIG. 1 the pressure sensor matrix 120, and FIG. 6 the pressure matrix sensor 610: Pressure Sensor Matrix 120 may comprise an N×M matrix of pressure sensors enabling sensing pressure excreted by a person sitting on seat and leaning back against backrest. The matrix may comprise 3×3 sensors, or more, such as 4×4, 6×6, 8×6, 10×8, or the like and Tokuchi: Abstract, [0017], [0029]-[0031], [0034], [0037], [0041]-[0042], [0058], FIG. 6-8, and FIG. 10: When in use, the sheet-shaped apparatus 100 is placed over the area of a chair 20 from a seating portion 22 to a backrest 24 as illustrated in FIG. 10. In FIG. 10, the sheet-shaped apparatus 100 is depicted thicker than in reality for easy recognition and Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: The seat sensors are placed based upon human anthropometry to accurately measure load and load distribution on the cushion by the wheelchair user during both static and dynamic seated postures.), wherein the controller is configured to classify the user’s posture based on detection data of the seating face sensor unit (Lustig: Abstract, [0043]-[0046], [0077], [0105], [0114], and FIG. 3-5: On Step 310, posture of the person may be estimated based on the sensor input. The estimation may be performed by the processor using a supervised classifier, such as a k-means, Support Vector Machines (SVM), or the like, which may be trained to deduce a posture based on input based on training data provided thereto. The training data may comprise sensor readings and correct labels thereof, indicating the posture. The posture estimation may be performed on the apparatus itself without requiring computation by an external computation platform, such as a cloud-based server. In some exemplary embodiments, the classifier may be trained offline and the trained model may be provided to the apparatus to be used locally. In some exemplary embodiments, the potential posture for estimation may include, for example, forward sloping, slump, side reliance, cross legged, no legs support, correct posture with back support, correct posture without back support, standing, or the like. Each posture may be associated with a different severity measurement, a different alleviating stretches or exercises, or the like, Tokuchi: Abstract, [0017], [0019]-[0024], [0029]-[0031], [0034], [0037], [0041]-[0042], and FIG. 6-8: In each case when the user places the sheet-shaped apparatus 100 on the same seat and sits in the seat, the pressure distribution sensed by the pressure distribution sensor 102 disposed on the body side falls within a certain range, although some variations may occur due to slight differences in sitting posture for individual occasions when the user sits in the seat, and likewise, the pressure distribution sensed by the pressure distribution sensor 102 disposed on the seat side falls within a certain range (e.g., assuming that the user sits back in the seat with correct posture in each compared case), Karunaratne: column 9 lines 51-column 10 lines 26, column 13 lines 49-column 14 lines 21, column 15 lines 4-34, column 24 lines 63-column 25 lines 19, and FIG. 1-13: In this embodiment, there are seven pressure sensors and one angle sensor for capturing the pressure distribution on a smart seat cover with both a backrest portion 328 and a seat portion 348. In other embodiments, any number of pressure or angle sensors may be present, Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5, and Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion).
As to claim 11, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune disclose the limitations of claim 5 further comprising the posture detection system according to Claim 5, wherein the haptic feedback includes a plurality of actuators for vibrating the backrest or the seating face, and the controller is configured to operate the actuators in a pattern according to the result of the classification (Lustig: [0008], [0045], [0049], [0062], FIG. 1 the haptic feedback devices 130: the device may perform posture estimation locally on-device and may select a feedback from the local feedback devices, such as integrated vibration motors embedded with the device itself and Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion), and the controller is configured to operate the actuators in a pattern according to the result of the classification (Lustig: [0008], [0045], [0049], [0062], FIG. 1 the haptic feedback devices 130: In some exemplary embodiments, feedback is provided to the person using a haptic feedback device, such as a vibrating motor, a kinesthetic feedback device, a tactile feedback device, or the like. In some exemplary embodiments, several different forms of haptic feedbacks may be used to relay a different message. For example, a same vibration motor may vibrate intermittently for 3 seconds to relay one message, and vibrate continuously for 3 seconds to relay another message. In addition, there may be several vibration motors used to provide different feedbacks. For example, there may be two motors located on the seat of the chair, one located in the center of the left half of the chair and the other in the center of the right half of the chair. A vibration by the left motor may indicate the person needs to mind his posture relating to his left hand-side (e.g., tilting to the left excessively); a vibration by the right motor may similarly relate to the person's right side; and a simulations vibration by both motors may indicate the person has correctly corrected his posture. As yet another example, a relatively long vibration by both motors may prompt the user to stand up after a determination of a length sit session is determined. As yet another example, a vibration by both motors in a non-continuous manner may indicate bad posture which needs to be corrected, Karunaratne: column 9 lines 18-47, column 10 lines 13-40, column 14 lines 22-45, column 22 lines 45-48, and FIG. 13-16: Depending on the user's preferences, such an alert may be sent via audio or silent vibrations, pinches, or even electric shocks. Again, such automatic notifications do not interfere with the user's activities, while providing constant, real-time feedback to ensure that posture improvements occur continuously but also progressively, and Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion).
As to claim 13, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune disclose the limitations of claim 1 further comprising the posture detection system according to Claim 1, wherein the display unit is configured to display the recommended pose for the user (Lustig: [0051], [0112]-[0116], and FIG. 7: FIG. 7A shows an illustration of a user interface of an application program, in accordance with some exemplary embodiments of the disclosed subject matter. Current sitting time is displayed (702) together with a visual indication of aggregated posture information over time (706, 708). For example, indication of aggregative time in the last time window of an hour is shown for good posture (708) and rest time (706). Real-time message (710) is displayed if current estimated posture is incorrect. Real-time message (710) may indicate area of the body where the posture is incorrect (712)), and, in response to determining whether the user’s pose matches the recommended pose according to a result of the detection by the pressure sensor unit (Lustig: [0051], [0112]-[0116], and FIG. 7: FIG. 7A shows an illustration of a user interface of an application program, in accordance with some exemplary embodiments of the disclosed subject matter. Current sitting time is displayed (702) together with a visual indication of aggregated posture information over time (706, 708). For example, indication of aggregative time in the last time window of an hour is shown for good posture (708) and rest time (706). Real-time message (710) is displayed if current estimated posture is incorrect. Real-time message (710) may indicate area of the body where the posture is incorrect (712)), feedback is provided according to a result of the determination (Lustig: [0016], [0051],[0077], [0080]-[0081], [0111]-[0116], and FIG. 7: or example, the same message indicating “leaning to the left”, may be provided using haptic feedback, by a written message, using audial feedback, and using visual feedback. In some cases, different content messages may be provided such as textual message indicating of the incorrect posture, the person may be informed of an informative study attesting to the dangerous and adverse affects of such incorrect posture, the person may receive an exercise suggestion to begin performing to strengthen muscles that are relevant to the incorrect posture, the person may receiving an immediate stretch suggestion that may be suitable to improve posture, potentially specifically with relation to the incorrect posture (e.g., stretching the right side of the body to cause the back to be positioned in a more symmetric manner), the person may be shown a visual indication of his current posture and an indication of how to improve the posture, such as by visually emphasizing a body part to be moved, the person may be provided with a visual pressure map showing the pressure sensed by the pressure sensors, or the like and Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion).
As to claim 17, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune disclose the limitations of claim 1 further comprising the posture detection system according to Claim 1, wherein user information about the user’s physical features is input to the controller (Lustig: [0052], [0062]-[0065], [0081], [0116], and FIG. 7: In some exemplary embodiments, the automated adjustment may be performed based on physical measurements of the person, such as provided explicitly by the user, or gathered from the sensors and analysis thereof. In some exemplary embodiments, height of the person may be inferred from the pressure sensors from the backseat. For example, the first pressure sensors in the backrest that sense pressure may indicate a length of the lumber of the person. Additionally or alternatively, the weight of the person may be estimated based on the amount of pressure overtime. In some exemplary embodiments, the weight of the person may be based on calculated of average pressure overtime, combined with the angles of the seat and backrest. Additionally or alternatively, estimation as to the amount of pressure exerted by the legs on the surface that is not sensed by the sensors may be computed and utilized. It will be noted that the measured pressure may change over time not only based on changes in weight, but also on changes in weight distribution. If the person puts his leg on the ground, not all of his weight is distributed on the pressure sensors. Monitoring the readings from the pressure sensors over time may be useful in estimating actual weight based on the different pressure measurements and distributions thereof. In some exemplary embodiments, training datasets may be provided based on users' that explicitly provided their physical measurements, and the information may be used to train a supervised machine learning classifier to estimate a weight of another person based solely on the sensor readings), and the controller is configured to define the user’s ideal posture based on the user information (Lustig: [0052], [0062]-[0065], [0081], [0116], and FIG. 7: Another user interface illustration is shown in FIG. 7F. Information about the specific chair model may be shown (750). The chair model may be a-priori known if the apparatus is embedded within the chair and is permanently attached thereto. Additionally or alternatively, the user may manual indicate the model. Suggested chair setting may be displayed (755). The suggested settings may be determined based on the physical measurements of the person, based on expert knowledge relating to the chair, combination thereof, or the like. In some exemplary embodiments, the suggested setting may be displayed to the user. Additionally or alternatively, the suggested setting may be applied automatically. In some exemplary embodiments, statistical information about the usage of the chair, such as average sitting time (760) may be displayed and Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion).
As to claim 24, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune discloses the limitations of claim 1 further comprising the posture detection system according to Claim 1, wherein the display unit is configured to display a report including at least one of:
summary of a sedentary performance or activeness (Lustig: [0112]-[0116], and FIG. 7: Another user interface illustration is shown in FIGS. 7C and 7D, which show user analytics. In FIG. 7C, aggregated information is shown relating sitting time and times in which the different postures were detected (730), Karunaratne: column 4 lines 9-20, column 29 lines 18-38, lines 40-52, and FIG. 19-23: Starting at step 2210, the BACKGURU smart seat cover system may acquire a Correct Posture Score for the user at step 2220, and may determine whether the score is less than or equal to a Silver threshold score S at step 2230. In determining that the score is less than or equal to S, the system may add the user to a “Bronze” category at step 2240, and ends the process at step 2280. In determining that the score is not less than or equal to S at step 2230, the system may further check whether the score is greater than S and less than a Gold threshold score G at step 2250. In determining that the score is greater than S and less than G, the system may add the user to a “Silver” category at step 2260, and ends the process at step 2280. In determining that the score is greater than or equal to G at step 2250, the system may add the user to a “Gold” category at step 2270, and ends the process at step 2280 and Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: In other embodiments, the user can set goals 428 (e.g., change from left-leaning position to right-leaning position every 15 minutes) and the system can provide active feedback 432 to the user to assist the user in achieving such goals. Such feedback can include verbal indications and alarms transmitted to the user or to a clinician via a mobile app when the use data indicate that a characterized weight shifting movement event has not occurred within the amount of time set as a goal),
a score of the posture or activeness (Lustig: [0051], [0083], and [0086]: The desired outcome may be an immediate outcome (e.g., changing the posture) or long-term outcome (e.g., improving muscle strength, and thereby improving posture in the long-run). Based on the outcome, the disclosed subject matter may improve its future feedbacks and provide more efficient feedbacks for the same person. In some exemplary embodiments, the feedback history may be fed into a classifier for training the classifier to predict an outcome score to a feedback. The classifier may be utilized when a new feedback is selected between alternative potential feedbacks to select the feedback with the highest predicted score), wherein the score of the posture or activeness is determined for a time period based on at least one of a sitting time duration, a percentage of occurrence of the posture, a frequency of breaks, a duration of breaks, pressure distribution symmetry value and a detection of performing stretches (Karunaratne: column 4 lines 9-20, column 28 lines 52-59, column 29 lines 18-38, lines 40-52, lines 59-column 30 lines 12, and FIG. 19-23: A challenge type selection menu 2415 may allow the user to select a type for the new challenge. For example, the user may elect to maintain a correct posture, or correct an incorrect posture over a configurable period of time. A percentage selection menu 2420 may allow the user to select a minimum percentage of correct sitting time which users involved in the challenge should attain in order to meet the challenge over a duration as set through a menu 2425 and Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: In other embodiments, the user can set goals 428 (e.g., change from left-leaning position to right-leaning position every 15 minutes) and the system can provide active feedback 432 to the user to assist the user in achieving such goals. Such feedback can include verbal indications and alarms transmitted to the user or to a clinician via a mobile app when the use data indicate that a characterized weight shifting movement event has not occurred within the amount of time set as a goal) and
recommended action including a stretch pose, exercise routine or a sedentary guidance (Lustig: [0112]-[0116], and FIG. 7: Another user interface illustration is shown in FIG. 7E, exemplifies exercise information screen. Animation 740 may be displayed showing how the person should exercise. A Visual Timer 742 may be displayed and used to time the exercise time of the person and indicate when the person may finish exercising), wherein the stretch pose is associated with the classified posture and wherein the sedentary guidance is classified based on a history of the user’s postures and a score of the posture or activeness of the user.
As to claim 26, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune discloses the limitations of claim 1 further the posture detection system according to Claim 1, wherein
a pressure distribution is measured by detection data of the pressure sensor unit (Lustig: [0064], [0111]-[0116], FIG. 3-5, and FIG. 7, Karunaratne: column 9 lines 51-column 10 lines 26, column 13 lines 49-column 14 lines 21, column 15 lines 4-34, column 24 lines 63-column 25 lines 19, and FIG. 1-3: In this embodiment, there are seven pressure sensors and one angle sensor for capturing the pressure distribution on a smart seat cover with both a backrest portion 328 and a seat portion 348. In other embodiments, any number of pressure or angle sensors may be present, and Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: The seat sensors are placed based upon human anthropometry to accurately measure load and load distribution on the cushion by the wheelchair user during both static and dynamic seated postures), and
input data of the supervised machine learning tool (Lustig: [0064], [0111]-[0116], FIG. 3-5, and FIG. 7, Karunaratne: column 26 lines 29 - column 28 lines 5 and FIG. 19-21: Initial training data may be fed to the system manually by professionals via a web-based admin panel for one or more machine learning or other statistical algorithms 2120, which may be Naïve Bayes classification, K-Means, Support Vector Machine (SVM), a hybrid approach using multiple algorithms, or any other algorithm. Then, current and historical data 2110 from pressure sensor readings, angle sensor readings, health sensor readings, and behavioural sensor readings may be fed to the machine learning algorithms and medical predictions 2130 may be made based on the user's sitting behaviours and/or other health and behavioural data from the sensor readings, and Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: The seat sensors are placed based upon human anthropometry to accurately measure load and load distribution on the cushion by the wheelchair user during both static and dynamic seated postures. An algorithm capable of accommodating system creep over long durations of use can maintain the ability to accurately classify weight shift and in-seat movement activity. An algorithm is capable of classifying a continuum of weight-shifting activities ranging from partial load distribution resulting from leaning and reaching to complete unweighting of the cushion surface. The classification algorithm takes into account both magnitude and duration of loading to better distinguish different weight-shifting behaviors. Also, an algorithm can be used that is capable of monitoring the in-seat movement of a wheelchair users that results from transient re-distribution of cushion loading during dynamic changes in seated posture) includes information of a physical feature of the user, the detection data of the pressure sensor unit (Lustig: [0064], [0111]-[0116], FIG. 3-5, and FIG. 7: It will be noted that the measured pressure may change over time not only based on changes in weight, but also on changes in weight distribution. If the person puts his leg on the ground, not all of his weight is distributed on the pressure sensors. Monitoring the readings from the pressure sensors over time may be useful in estimating actual weight based on the different pressure measurements and distributions thereof. In some exemplary embodiments, training datasets may be provided based on users' that explicitly provided their physical measurements, and the information may be used to train a supervised machine learning classifier to estimate a weight of another person based solely on the sensor readings and Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: In other embodiments, the user can set goals 428 (e.g., change from left-leaning position to right-leaning position every 15 minutes) and the system can provide active feedback 432 to the user to assist the user in achieving such goals. Such feedback can include verbal indications and alarms transmitted to the user or to a clinician via a mobile app when the use data indicate that a characterized weight shifting movement event has not occurred within the amount of time set as a goal), a score of the posture or activeness (Lustig: [0051], [0083], and [0086]: The desired outcome may be an immediate outcome (e.g., changing the posture) or long-term outcome (e.g., improving muscle strength, and thereby improving posture in the long-run). Based on the outcome, the disclosed subject matter may improve its future feedbacks and provide more efficient feedbacks for the same person. In some exemplary embodiments, the feedback history may be fed into a classifier for training the classifier to predict an outcome score to a feedback. The classifier may be utilized when a new feedback is selected between alternative potential feedbacks to select the feedback with the highest predicted score and Karunaratne: column 4 lines 9-20, column 28 lines 52-59, column 29 lines 18-38, lines 40-52, lines 59-column 30 lines 12, and FIG. 19-23: A challenge type selection menu 2415 may allow the user to select a type for the new challenge. For example, the user may elect to maintain a correct posture, or correct an incorrect posture over a configurable period of time. A percentage selection menu 2420 may allow the user to select a minimum percentage of correct sitting time which users involved in the challenge should attain in order to meet the challenge over a duration as set through a menu 2425), and the time of day (Lustig: [0051], [0112]-[0116], and FIG. 7: FIG. 7A shows an illustration of a user interface of an application program, in accordance with some exemplary embodiments of the disclosed subject matter. Current sitting time is displayed (702) together with a visual indication of aggregated posture information over time (706, 708). For example, indication of aggregative time in the last time window of an hour is shown for good posture (708) and rest time (706). Real-time message (710) is displayed if current estimated posture is incorrect. Real-time message (710) may indicate area of the body where the posture is incorrect (712), and Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion).
As to claim 27, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune discloses the limitations of claim 1 further comprising the posture detection system according to Claim 1, wherein the user’s state is predicted according to a result of the prediction of the user’s behavior (Lustig: Abstract, [0043]-[0046], [0064]-[0065], [0077], [0105], [0111]-[0116], FIG. 3-5 and FIG. 7: In some exemplary embodiments, the potential posture for estimation may include, for example, forward sloping, slump, side reliance, cross legged, no legs support, correct posture with back support, correct posture without back support, standing, or the like, Karunaratne: column 26 lines 29 - column 28 lines 5 and FIG. 19-21: Initial training data may be fed to the system manually by professionals via a web-based admin panel for one or more machine learning or other statistical algorithms 2120, which may be Naïve Bayes classification, K-Means, Support Vector Machine (SVM), a hybrid approach using multiple algorithms, or any other algorithm. Then, current and historical data 2110 from pressure sensor readings, angle sensor readings, health sensor readings, and behavioural sensor readings may be fed to the machine learning algorithms and medical predictions 2130 may be made based on the user's sitting behaviours and/or other health and behavioural data from the sensor readings and Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion).
As to claim 28, Lustig discloses a posture detection method for detecting a user’s posture, the posture detection method comprising:
receiving, in real-time, raw pressure distribution data detected by a pressure sensor unit comprising a plurality of sensors, each of the sensors configured to detect a pressure applied from a user (Lustig: [0061], [0064], [0076]-[0077], [0108], [0110], and FIG. 6: Processor 602 may be configured to process sensor readings and select a feedback device to provide feedback to the person sitting on the chair);
detecting a pressure applied from a user (Lustig: [0018], [0043], [0051], [0059], [0061], [0064], FIG. 1 the pressure sensor matrix 120, and FIG. 6 the pressure matrix sensor 610: Pressure Sensor Matrix 120 may comprise an N×M matrix of pressure sensors enabling sensing pressure excreted by a person sitting on seat and leaning back against backrest. The matrix may comprise 3×3 sensors, or more, such as 4×4, 6×6, 8×6, 10×8, or the like) using a pressure sensor unit (Lustig: [0018], [0043], [0051], [0059], [0061], [0064], FIG. 1 the pressure sensor matrix 120, and FIG. 6 the pressure matrix sensor 610: The sensors may include sensors useful for determining pressure exerted on the chair at different locations. In particular, a plurality of pressure points may be monitored in a seat of the chair and a plurality of pressure point may be monitored in a backrest of the chair. In some exemplary embodiments, a matrix of 3×3 or more pressure points may be used for the seat, the backrest, or for both), the pressure sensor unit including a sheet shape or a padded shape and including plurality of sensors, and each of the sensors being configured to detect the pressure applied from the user (Lustig: [0018], [0043], [0051], [0059], [0061], [0064], FIG. 1 the pressure sensor matrix 120, and FIG. 6 the pressure matrix sensor 610);
classifying the user’s posture and a user behavior of the user (Lustig: Abstract, [0043]-[0046], [0077], [0105], [0114], and FIG. 3-5: On Step 310, posture of the person may be estimated based on the sensor input. The estimation may be performed by the processor using a supervised classifier, such as a k-means, Support Vector Machines (SVM), or the like, which may be trained to deduce a posture based on input based on training data provided thereto. The training data may comprise sensor readings and correct labels thereof, indicating the posture. The posture estimation may be performed on the apparatus itself without requiring computation by an external computation platform, such as a cloud-based server. In some exemplary embodiments, the classifier may be trained offline and the trained model may be provided to the apparatus to be used locally. In some exemplary embodiments, the potential posture for estimation may include, for example, forward sloping, slump, side reliance, cross legged, no legs support, correct posture with back support, correct posture without back support, standing, or the like. Each posture may be associated with a different severity measurement, a different alleviating stretches or exercises, or the like) other than the posture of the user based on the raw pressure distribution data (Lustig: [0112]-[0116] and FIG. 7: Current sitting time is displayed (702) together with a visual indication of aggregated posture information over time (706, 708)… Another user interface illustration is shown in FIG. 7B, which shows real time pressure map. Pressure maps corresponding to pressure sensor matrices are shown. Pressure Map 720 shows pressure sensed on backrest…aggregated information is shown relating sitting time and times in which the different postures were detected (730). A distribution of the different postures over time is visually show. In some cases, the postures may be abstracted into groups of Good, Bad, Break and Exercise…Animation 740 may be displayed showing how the person should exercise. A Visual Timer 742 may be displayed and used to time the exercise time of the person and indicate when the person may finish exercising)) using a supervised machine learning (ML) classifier to generate a classification (Lustig: Abstract, [0043]-[0046], [0077], [0105], [0114], and FIG. 3-5: On Step 310, posture of the person may be estimated based on the sensor input. The estimation may be performed by the processor using a supervised classifier, such as a k-means, Support Vector Machines (SVM), or the like, which may be trained to deduce a posture based on input based on training data provided thereto. The training data may comprise sensor readings and correct labels thereof, indicating the posture. The posture estimation may be performed on the apparatus itself without requiring computation by an external computation platform, such as a cloud-based server. In some exemplary embodiments, the classifier may be trained offline and the trained model may be provided to the apparatus to be used locally. In some exemplary embodiments, the potential posture for estimation may include, for example, forward sloping, slump, side reliance, cross legged, no legs support, correct posture with back support, correct posture without back support, standing, or the like. Each posture may be associated with a different severity measurement, a different alleviating stretches or exercises, or the like);
providing feedback to the user by vibrating based on a result of the classification (Lustig: Abstract, [0045], [0047], [0062], FIG. 1, and FIG. 6: In some exemplary embodiments, feedback is provided to the person using a haptic feedback device, such as a vibrating motor, a kinesthetic feedback device, a tactile feedback device, or the like. In some exemplary embodiments, several different forms of haptic feedbacks may be used to relay a different message. For example, a same vibration motor may vibrate intermittently for 3 seconds to relay one message, and vibrate continuously for 3 seconds to relay another message. In addition, there may be several vibration motors used to provide different feedbacks); and
performing a display according to the result of the classification (Lustig: [0051], [0112]-[0116], and FIG. 7: FIG. 7A shows an illustration of a user interface of an application program, in accordance with some exemplary embodiments of the disclosed subject matter. Current sitting time is displayed (702) together with a visual indication of aggregated posture information over time (706, 708). For example, indication of aggregative time in the last time window of an hour is shown for good posture (708) and rest time (706). Real-time message (710) is displayed if current estimated posture is incorrect. Real-time message (710) may indicate area of the body where the posture is incorrect (712)).
Lustig does not explicitly disclose classifying the user’s posture and a user behavior of the user other than the posture of the user based on the raw pressure distribution data;
determining whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and a posture pressure distribution,
wherein the recommended pose as a stretch pose selected from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose,
wherein the pose pressure distribution serves as an input to the supervised machine learning classifier that outputs, in real-time, a behavior label classifying the user behavior other than the posture of the user in real-time;
wherein classifying the user’s posture and the behavior label is performed directly by the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling.
However, it has been known in the art of determining postures of a user to implement determining whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and a posture pressure distribution, as suggested by Tokuchi, which discloses determining whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and a posture pressure distribution (Tokuchi: Abstract, [0017], [0019]-[0024], [0029]-[0031], [0034], [0037], [0041]-[0042], and FIG. 6-8: In each case when the user places the sheet-shaped apparatus 100 on the same seat and sits in the seat, the pressure distribution sensed by the pressure distribution sensor 102 disposed on the body side falls within a certain range, although some variations may occur due to slight differences in sitting posture for individual occasions when the user sits in the seat, and likewise, the pressure distribution sensed by the pressure distribution sensor 102 disposed on the seat side falls within a certain range (e.g., assuming that the user sits back in the seat with correct posture in each compared case). Therefore, for the same user sitting in the same seat, the combination of the body-side and seat-side pressure distributions acting on the sheet-shaped apparatus falls within a certain range. Now, a case is considered in which the same user carries the sheet-shaped apparatus 100, and uses the sheet-shaped apparatus 100 by placing the sheet-shaped apparatus 100 on various seats in various places. In this case, if the material of the seating portion of a seat that the user sits on or the shape of the seating surface of the seat changes, both the pressure distribution sensed by the pressure distribution sensor 102 disposed on the body side and the pressure distribution sensed by the pressure distribution sensor 102 disposed on the seat side also change, and so does the combination of pressure distributions on the two sides).
Therefore, in view of teachings by Lustig and Tokuchi, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig to include determining whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and a posture pressure distribution, as suggested by Tokuchi. The motivation for this is to determine a user posture based on pressure distribution information.
While the combination of Lustig and Tokuchi discloses the pressure data/distribution for training supervised machine learning to identify postures of a user (Lustig: [0064], [0111]-[0116], FIG. 3-5, and FIG. 7: It will be noted that the measured pressure may change over time not only based on changes in weight, but also on changes in weight distribution. If the person puts his leg on the ground, not all of his weight is distributed on the pressure sensors. Monitoring the readings from the pressure sensors over time may be useful in estimating actual weight based on the different pressure measurements and distributions thereof. In some exemplary embodiments, training datasets may be provided based on users' that explicitly provided their physical measurements, and the information may be used to train a supervised machine learning classifier to estimate a weight of another person based solely on the sensor readings), the combination of Lustig and Tokuchi does not explicitly disclose classifying the user’s posture and a user behavior of the user other than the posture of the user based on the raw pressure distribution data.
However, it has been known in the art of monitoring sitting postures of user to implement classifying the user’s posture and a user behavior of the user other than the posture of the user based on the raw pressure distribution data, as suggested by Karunaratne, which discloses classifying the user’s posture (Karunaratne: column 26 lines 29 - column 28 lines 5 and FIG. 19-21: Initial training data may be fed to the system manually by professionals via a web-based admin panel for one or more machine learning or other statistical algorithms 2120, which may be Naïve Bayes classification, K-Means, Support Vector Machine (SVM), a hybrid approach using multiple algorithms, or any other algorithm. Then, current and historical data 2110 from pressure sensor readings, angle sensor readings, health sensor readings, and behavioural sensor readings may be fed to the machine learning algorithms and medical predictions 2130 may be made based on the user's sitting behaviours and/or other health and behavioural data from the sensor readings) and a user behavior of the user (column 8 lines 5-8, column 9 lines 18-47, column 9 lines 64- column 10 lines 26, column 12 lines 5-44, column 23 lines 4-54, column 22 lines 15-53, FIG. 3-13 and FIG. 19-21: The backrest portion and the seat portion may be part of a contiguous arrangement, or may be two separate components held together by another attachment means. In some embodiments, the system may further comprise health sensors that monitor health indicators such as heart rate, blood pressure, and various hormone levels, and behavioural sensors, which may, independently or together with the health sensors, monitor stress levels as well as other salient behavioural features.) other than the posture of the user (column 9 lines 64 - column 10 lines 26, column 19 lines 9-31, column 20 lines 66-column 21 lines 23, and FIG. 3-13: Again, such automatic notifications do not interfere with the user's activities, while providing constant, real-time feedback to ensure that posture improvements occur continuously but also progressively. Moreover, in some embodiments, the smart seat cover system may perform behaviour detection over time via health sensors and behavioral sensors to monitor health metrics that may collectively indicate the user's general state of health) based on the raw pressure distribution data (Karunaratne: column 9 lines 51-column 10 lines 26, column 13 lines 49-column 14 lines 21, column 15 lines 4-34, column 24 lines 63-column 25 lines 19, and FIG. 1-3: In this embodiment, there are seven pressure sensors and one angle sensor for capturing the pressure distribution on a smart seat cover with both a backrest portion 328 and a seat portion 348. In other embodiments, any number of pressure or angle sensors may be present).
Therefore in view of teachings by Lustig, Tokuchi, and Karunaratne, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig and Tokuchi to include classifying the user’s posture and a user behavior of the user other than the posture of the user based on the raw pressure distribution data, as suggested by Karunaratne. The motivation for this is to identify sitting posture of a user and discourage sedentary behavior to minimize associated adverse health effects.
While the combination of Lustig, Tokuchi, and Karunaratne discloses the pressure data/distribution for training supervised machine learning to identify postures of a user (Lustig: [0064], [0111]-[0116], FIG. 3-5, and FIG. 7: It will be noted that the measured pressure may change over time not only based on changes in weight, but also on changes in weight distribution. If the person puts his leg on the ground, not all of his weight is distributed on the pressure sensors. Monitoring the readings from the pressure sensors over time may be useful in estimating actual weight based on the different pressure measurements and distributions thereof. In some exemplary embodiments, training datasets may be provided based on users' that explicitly provided their physical measurements, and the information may be used to train a supervised machine learning classifier to estimate a weight of another person based solely on the sensor readings), the combination of Lustig, Tokuchi, and Karunaratne does not explicitly disclose wherein the pose pressure distribution serves as an input to the supervised machine learning classifier that outputs, in real-time, a behavior label classifying the user behavior other than the posture of the user in real-time;
wherein classifying the user’s posture and the behavior label is performed directly by the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling.
However, it has been known in the art of monitoring sitting postures of user to implement wherein the pose pressure distribution serves as an input to the supervised machine learning classifier that outputs, in real-time, a behavior label classifying the user behavior other than the posture of the user in real-time;
wherein classifying the user’s posture and the behavior label is performed directly by the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling, as suggested by Sonenblum, which discloses classify the user’s posture and at least one user behavior other than posture based on the raw pressure distribution data using the ML classifier to generate a classification (Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: In other embodiments, the user can set goals 428 (e.g., change from left-leaning position to right-leaning position every 15 minutes) and the system can provide active feedback 432 to the user to assist the user in achieving such goals. Such feedback can include verbal indications and alarms transmitted to the user or to a clinician via a mobile app when the use data indicate that a characterized weight shifting movement event has not occurred within the amount of time set as a goal); and
wherein the pose pressure distribution serves as an input to the supervised machine learning classifier that outputs, in real-time, a behavior label classifying the user behavior other than the posture of the user in real-time; wherein classifying the user’s posture and the behavior label is performed directly by the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling (Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: The seat sensors are placed based upon human anthropometry to accurately measure load and load distribution on the cushion by the wheelchair user during both static and dynamic seated postures. An algorithm capable of accommodating system creep over long durations of use can maintain the ability to accurately classify weight shift and in-seat movement activity. An algorithm is capable of classifying a continuum of weight-shifting activities ranging from partial load distribution resulting from leaning and reaching to complete unweighting of the cushion surface. The classification algorithm takes into account both magnitude and duration of loading to better distinguish different weight-shifting behaviors. Also, an algorithm can be used that is capable of monitoring the in-seat movement of a wheelchair users that results from transient re-distribution of cushion loading during dynamic changes in seated posture).
Therefore in view of teachings by Lustig, Tokuchi, Karunaratne, and Sonenblum, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, and Karunaratne to include wherein the pose pressure distribution serves as an input to the supervised machine learning classifier that outputs, in real-time, a behavior label classifying the user behavior other than the posture of the user in real-time;
wherein classifying the user’s posture and the behavior label is performed directly by the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling, as suggested by Sonenblum. The motivation for this is to identify sitting posture of a user and provide suggestions to the user based on characterization data.
The combination of Lustig, Tokuchi, Karunaratne, and Sonenblum does not explicitly disclose wherein the recommended pose as a stretch pose selected from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose.
However, it has been known in the art of monitoring postures of users to implement wherein the recommended pose as a stretch pose selected from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose, as suggested by Bourahmoune, which discloses wherein the recommended pose as a stretch pose selected from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose (Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion: the heatmap for the stretch pose ‘Right Arm Cross’ (RAC) shows a high pressure reading on the upper left sensor. This captures accurately the pose performed by the experiment subjects where they extended the right arm across the chest and pressed the left arm on the right elbow. The pressure distribution heatmaps show that the six stretches are to some degree visually distinguishable from each other. Unlike sitting posture, a stretch is a static event that requires a conscious and dedicated effort from the user. Subjectivity in stretch interpretation and variable physical predisposition to perform the stretches result in greater inter-individual variability compared to the sitting posture recognition task).
Therefore in view of teachings by Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, Karunaratne, and Sonenblum to include wherein the recommended pose as a stretch pose selected from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose, as suggested by Bourahmoune. The motivation for this is to identify sitting posture of a user and provide suggestions to the user based on characterization data.
As to claim 29, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune discloses the limitations of claim 28 further comprising the posture detection method according to Claim 28, further comprising comparing a pressure distribution of a template with a current pressure distribution (Tokuchi: Abstract, [0017], [0019]-[0024], [0029]-[0031], [0034], [0037], [0041]-[0042], and FIG. 6-8 and Karunaratne: column 9 lines 51-column 10 lines 26, column 13 lines 49-column 14 lines 21, column 15 lines 4-34, column 24 lines 63-column 25 lines 19, and FIG. 1-3: In this embodiment, there are seven pressure sensors and one angle sensor for capturing the pressure distribution on a smart seat cover with both a backrest portion 328 and a seat portion 348. In other embodiments, any number of pressure or angle sensors may be present), wherein the pressure distribution of the template has been calibrated by using the user’s physical information (Lustig: [0052], [0062]-[0065], [0081], [0116], and FIG. 7: In some exemplary embodiments, the automated adjustment may be performed based on physical measurements of the person, such as provided explicitly by the user, or gathered from the sensors and analysis thereof. In some exemplary embodiments, height of the person may be inferred from the pressure sensors from the backseat. For example, the first pressure sensors in the backrest that sense pressure may indicate a length of the lumber of the person. Additionally or alternatively, the weight of the person may be estimated based on the amount of pressure overtime. In some exemplary embodiments, the weight of the person may be based on calculated of average pressure overtime, combined with the angles of the seat and backrest. Additionally or alternatively, estimation as to the amount of pressure exerted by the legs on the surface that is not sensed by the sensors may be computed and utilized. It will be noted that the measured pressure may change over time not only based on changes in weight, but also on changes in weight distribution. If the person puts his leg on the ground, not all of his weight is distributed on the pressure sensors. Monitoring the readings from the pressure sensors over time may be useful in estimating actual weight based on the different pressure measurements and distributions thereof. In some exemplary embodiments, training datasets may be provided based on users' that explicitly provided their physical measurements, and the information may be used to train a supervised machine learning classifier to estimate a weight of another person based solely on the sensor readings, Karunaratne: column 10 lines 56-column 11 lines 12, column 16 lines 37-column 17 lines 35, column 24 lines 54- column 26 lines 16, FIG. 17-18 and FIG. 21: In some embodiments, the user may be required to maintain a correct posture for a certain calibration period, such as 30 seconds, 60 seconds, or 90 seconds, for the auto-profiling or calibration process to complete. Measurement values from engaged pressure sensors and angle sensor, may be averaged across the calibration period to obtain a calibration reference for scaling ranges and thresholds for posture identification. In some embodiments. users of different body shapes or having different body weight distributions may require further adjustments to the scaled ranges and thresholds. For example, females generally have wider hips and may exert relatively more pressure on the BR and BL pressure sensors, and Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion).
As to claim 31, Lustig discloses a posture detection method for detecting a user’s posture, the posture detection method comprising:
using a pressure sensor unit (Lustig: [0018], [0043], [0051], [0059], [0061], [0064], FIG. 1 the pressure sensor matrix 120, and FIG. 6 the pressure matrix sensor 610: The sensors may include sensors useful for determining pressure exerted on the chair at different locations. In particular, a plurality of pressure points may be monitored in a seat of the chair and a plurality of pressure point may be monitored in a backrest of the chair. In some exemplary embodiments, a matrix of 3×3 or more pressure points may be used for the seat, the backrest, or for both) to detect, in real-time, a posture pressure distribution generated by a user, wherein each sensor is configured to detect the pressure applied from the user (Lustig: [0018], [0043], [0051], [0059], [0061], [0064], FIG. 1 the pressure sensor matrix 120, and FIG. 6 the pressure matrix sensor 610: Pressure Sensor Matrix 120 may comprise an N×M matrix of pressure sensors enabling sensing pressure excreted by a person sitting on seat and leaning back against backrest. The matrix may comprise 3×3 sensors, or more, such as 4×4, 6×6, 8×6, 10×8, or the like)
comparing the posture pressure distribution to a pressure distribution of a template that has been calibrated by using the user’s physical information (Lustig: [0052], [0062]-[0065], [0081], [0116], and FIG. 7: In some exemplary embodiments, the automated adjustment may be performed based on physical measurements of the person, such as provided explicitly by the user, or gathered from the sensors and analysis thereof. In some exemplary embodiments, height of the person may be inferred from the pressure sensors from the backseat. For example, the first pressure sensors in the backrest that sense pressure may indicate a length of the lumber of the person. Additionally or alternatively, the weight of the person may be estimated based on the amount of pressure overtime. In some exemplary embodiments, the weight of the person may be based on calculated of average pressure overtime, combined with the angles of the seat and backrest. Additionally or alternatively, estimation as to the amount of pressure exerted by the legs on the surface that is not sensed by the sensors may be computed and utilized. It will be noted that the measured pressure may change over time not only based on changes in weight, but also on changes in weight distribution. If the person puts his leg on the ground, not all of his weight is distributed on the pressure sensors. Monitoring the readings from the pressure sensors over time may be useful in estimating actual weight based on the different pressure measurements and distributions thereof. In some exemplary embodiments, training datasets may be provided based on users' that explicitly provided their physical measurements, and the information may be used to train a supervised machine learning classifier to estimate a weight of another person based solely on the sensor readings) and serves as an input to supervised ML classifier (Lustig: Abstract, [0043]-[0046], [0077], [0105], [0114], and FIG. 3-5: On Step 310, posture of the person may be estimated based on the sensor input. The estimation may be performed by the processor using a supervised classifier, such as a k-means, Support Vector Machines (SVM), or the like, which may be trained to deduce a posture based on input based on training data provided thereto. The training data may comprise sensor readings and correct labels thereof, indicating the posture. The posture estimation may be performed on the apparatus itself without requiring computation by an external computation platform, such as a cloud-based server. In some exemplary embodiments, the classifier may be trained offline and the trained model may be provided to the apparatus to be used locally. In some exemplary embodiments, the potential posture for estimation may include, for example, forward sloping, slump, side reliance, cross legged, no legs support, correct posture with back support, correct posture without back support, standing, or the like. Each posture may be associated with a different severity measurement, a different alleviating stretches or exercises, or the like), wherein the classifier is learned by a supervised machine learning tool (Lustig: Abstract, [0043]-[0046], [0077], [0105], [0114], and FIG. 3-5: On Step 310, posture of the person may be estimated based on the sensor input. The estimation may be performed by the processor using a supervised classifier, such as a k-means, Support Vector Machines (SVM), or the like, which may be trained to deduce a posture based on input based on training data provided thereto. The training data may comprise sensor readings and correct labels thereof, indicating the posture. The posture estimation may be performed on the apparatus itself without requiring computation by an external computation platform, such as a cloud-based server. In some exemplary embodiments, the classifier may be trained offline and the trained model may be provided to the apparatus to be used locally. In some exemplary embodiments, the potential posture for estimation may include, for example, forward sloping, slump, side reliance, cross legged, no legs support, correct posture with back support, correct posture without back support, standing, or the like. Each posture may be associated with a different severity measurement, a different alleviating stretches or exercises, or the like) and outputs, in real-time, a behavior label that classifies (Lustig: Abstract, [0043]-[0046], [0064]-[0065], [0077], [0105], [0111]-[0116], FIG. 3-5 and FIG. 7: In some exemplary embodiments, the potential posture for estimation may include, for example, forward sloping, slump, side reliance, cross legged, no legs support, correct posture with back support, correct posture without back support, standing, or the like), in real-time, at least one of a user posture and a user behavior of the user other than the posture of the user to generate a classification result (Lustig: [0051], [0112]-[0116], and FIG. 7: FIG. 7A shows an illustration of a user interface of an application program, in accordance with some exemplary embodiments of the disclosed subject matter. Current sitting time is displayed (702) together with a visual indication of aggregated posture information over time (706, 708). For example, indication of aggregative time in the last time window of an hour is shown for good posture (708) and rest time (706). Real-time message (710) is displayed if current estimated posture is incorrect. Real-time message (710) may indicate area of the body where the posture is incorrect (712)); and
providing haptic feedback to the user based on the classification result (Lustig: Abstract, [0045], [0047], [0062], FIG. 1, and FIG. 6: In some exemplary embodiments, feedback is provided to the person using a haptic feedback device, such as a vibrating motor, a kinesthetic feedback device, a tactile feedback device, or the like. In some exemplary embodiments, several different forms of haptic feedbacks may be used to relay a different message. For example, a same vibration motor may vibrate intermittently for 3 seconds to relay one message, and vibrate continuously for 3 seconds to relay another message. In addition, there may be several vibration motors used to provide different feedbacks).
Lustig does not explicitly disclose
applying pressure to a plurality of pressure sensors that include a sheet shape or a padded shape;
classifying the user’s posture and a user behavior of the user other than the posture of the user based on the posture pressure distribution using a supervised machine learning (ML) classifier to generate a classification result, wherein classifying the user’s posture and the user behavior is performed directly by the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling;
determining whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and the posture pressure distribution;
wherein the recommended pose as a stretch pose selected from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose, and
comparing the posture pressure distribution to a pressure distribution of a template that has been calibrated by using the user’s physical information and serves as an input to supervised ML classifier, wherein the classifier is learned by a supervised machine learning tool and outputs, in real-time, a behavior label that classifies, in real-time, at least one of a user posture and a user behavior of the user other than the posture of the user to generate a classification result.
However, it has been known in the art of determining postures of a user to implement applying pressure to a plurality of pressure sensors that include a sheet shape or a padded shape; determining whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and the posture pressure distribution; and
comparing the posture pressure distribution to a pressure distribution of a template that has been calibrated by using the user’s physical information, as suggested by Tokuchi, which discloses applying pressure to a plurality of pressure sensors that include a sheet shape or a padded shape; determining whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and the posture pressure distribution (Tokuchi: Abstract, [0017], [0019]-[0024], [0029]-[0031], [0034], [0037], [0041]-[0042], and FIG. 6-8: In each case when the user places the sheet-shaped apparatus 100 on the same seat and sits in the seat, the pressure distribution sensed by the pressure distribution sensor 102 disposed on the body side falls within a certain range, although some variations may occur due to slight differences in sitting posture for individual occasions when the user sits in the seat, and likewise, the pressure distribution sensed by the pressure distribution sensor 102 disposed on the seat side falls within a certain range (e.g., assuming that the user sits back in the seat with correct posture in each compared case). Therefore, for the same user sitting in the same seat, the combination of the body-side and seat-side pressure distributions acting on the sheet-shaped apparatus falls within a certain range. Now, a case is considered in which the same user carries the sheet-shaped apparatus 100, and uses the sheet-shaped apparatus 100 by placing the sheet-shaped apparatus 100 on various seats in various places. In this case, if the material of the seating portion of a seat that the user sits on or the shape of the seating surface of the seat changes, both the pressure distribution sensed by the pressure distribution sensor 102 disposed on the body side and the pressure distribution sensed by the pressure distribution sensor 102 disposed on the seat side also change, and so does the combination of pressure distributions on the two sides); and comparing the posture pressure distribution to a pressure distribution of a template that has been calibrated by using the user’s physical information (Tokuchi: Abstract, [0017], [0019]-[0024], [0029]-[0031], [0034], [0037], [0041]-[0042], and FIG. 6-8: If it is determined at S38 that someone is seated, the sheet-linked program 204 obtains distribution characteristics data from the sensor data in the same manner as at S14 (S40), and determines whether the obtained distribution characteristics data matches any registered distribution characteristics data included in the mode settings information (S42). In the determination at S42, rather than checking all pieces of registered distribution characteristics data included in the mode settings information, the sheet-linked program 204 may only determine whether the distribution characteristics data obtained at S40 matches the registered distribution characteristics data previously determined at S16 prior to the user leaving the seat).
Therefore, in view of teachings by Lustig and Tokuchi, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig to include applying pressure to a plurality of pressure sensors that include a sheet shape or a padded shape; determining whether a current pose of the user matches a recommended pose based on a ranking of a similarity metric between a recommended pose pressure distribution and the posture pressure distribution; and comparing the posture pressure distribution to a pressure distribution of a template that has been calibrated by using the user’s physical information, as suggested by Tokuchi. The motivation for this is to determine a user posture based on pressure distribution information.
While the combination of Lustig and Tokuchi discloses the pressure data/distribution for training supervised machine learning to identify postures of a user (Lustig: [0064], [0111]-[0116], FIG. 3-5, and FIG. 7: It will be noted that the measured pressure may change over time not only based on changes in weight, but also on changes in weight distribution. If the person puts his leg on the ground, not all of his weight is distributed on the pressure sensors. Monitoring the readings from the pressure sensors over time may be useful in estimating actual weight based on the different pressure measurements and distributions thereof. In some exemplary embodiments, training datasets may be provided based on users' that explicitly provided their physical measurements, and the information may be used to train a supervised machine learning classifier to estimate a weight of another person based solely on the sensor readings), the combination of Lustig and Tokuchi does not explicitly disclose wherein the classifier is learned by a supervised machine learning tool and outputs, in real-time, a behavior label that classifies, in real-time, at least one of a user posture and a user behavior of the user other than the posture of the user to generate a classification result.
However, it has been known in the art of monitoring sitting postures of user to implement wherein the classifier is learned by a supervised machine learning tool and outputs, in real-time, a behavior label that classifies, in real-time, at least one of a user posture and a user behavior of the user other than the posture of the user to generate a classification result, as suggested by Karunaratne, which discloses
wherein the classifier is learned by a supervised machine learning tool (Karunaratne: column 26 lines 29 - column 28 lines 5 and FIG. 19-21: Initial training data may be fed to the system manually by professionals via a web-based admin panel for one or more machine learning or other statistical algorithms 2120, which may be Naïve Bayes classification, K-Means, Support Vector Machine (SVM), a hybrid approach using multiple algorithms, or any other algorithm. Then, current and historical data 2110 from pressure sensor readings, angle sensor readings, health sensor readings, and behavioural sensor readings may be fed to the machine learning algorithms and medical predictions 2130 may be made based on the user's sitting behaviours and/or other health and behavioural data from the sensor readings) and outputs, in real-time, a behavior label that classifies, in real-time, at least one of a user posture and a user behavior of the user (column 8 lines 5-8, column 9 lines 18-47, column 9 lines 64- column 10 lines 26, column 12 lines 5-44, column 23 lines 4-54, column 22 lines 15-53, FIG. 3-13 and FIG. 19-21: The backrest portion and the seat portion may be part of a contiguous arrangement, or may be two separate components held together by another attachment means. In some embodiments, the system may further comprise health sensors that monitor health indicators such as heart rate, blood pressure, and various hormone levels, and behavioural sensors, which may, independently or together with the health sensors, monitor stress levels as well as other salient behavioural features.) other than the posture of the user to generate a classification result (column 9 lines 64 - column 10 lines 26, column 19 lines 9-31, column 20 lines 66-column 21 lines 23, and FIG. 3-13: Again, such automatic notifications do not interfere with the user's activities, while providing constant, real-time feedback to ensure that posture improvements occur continuously but also progressively. Moreover, in some embodiments, the smart seat cover system may perform behaviour detection over time via health sensors and behavioral sensors to monitor health metrics that may collectively indicate the user's general state of health).
Therefore in view of teachings by Lustig, Tokuchi, and Karunaratne, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig and Tokuchi to include wherein the classifier is learned by a supervised machine learning tool and outputs, in real-time, a behavior label that classifies, in real-time, at least one of a user posture and a user behavior of the user other than the posture of the user to generate a classification result, as suggested by Karunaratne. The motivation for this is to identify sitting posture of a user and discourage sedentary behavior to minimize associated adverse health effects.
While the combination of Lustig, Tokuchi, and Karunaratne discloses the pressure data/distribution for training supervised machine learning to identify postures of a user (Lustig: [0064], [0111]-[0116], FIG. 3-5, and FIG. 7: It will be noted that the measured pressure may change over time not only based on changes in weight, but also on changes in weight distribution. If the person puts his leg on the ground, not all of his weight is distributed on the pressure sensors. Monitoring the readings from the pressure sensors over time may be useful in estimating actual weight based on the different pressure measurements and distributions thereof. In some exemplary embodiments, training datasets may be provided based on users' that explicitly provided their physical measurements, and the information may be used to train a supervised machine learning classifier to estimate a weight of another person based solely on the sensor readings), the combination of Lustig, Tokuchi, and Karunaratne does not explicitly disclose classifying the user’s posture and a user behavior of the user other than the posture of the user based on the posture pressure distribution using a supervised machine learning (ML) classifier to generate a classification result, wherein classifying the user’s posture and the user behavior is performed directly by the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling.
However, it has been known in the art of monitoring sitting postures of user to implement classifying the user’s posture and a user behavior of the user other than the posture of the user based on the posture pressure distribution using a supervised machine learning (ML) classifier to generate a classification result, wherein classifying the user’s posture and the user behavior is performed directly by the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling, as suggested by Sonenblum, which discloses classifying the user’s posture and a user behavior of the user other than the posture of the user based on the posture pressure distribution using a supervised machine learning (ML) classifier to generate a classification result (Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: In other embodiments, the user can set goals 428 (e.g., change from left-leaning position to right-leaning position every 15 minutes) and the system can provide active feedback 432 to the user to assist the user in achieving such goals. Such feedback can include verbal indications and alarms transmitted to the user or to a clinician via a mobile app when the use data indicate that a characterized weight shifting movement event has not occurred within the amount of time set as a goal); and
wherein classifying the user’s posture and the user behavior is performed directly by the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling (Sonenblum: Abstract, column 3 lines 25 – 42, column 5 lines 21-43, column 5 lines 51 – column 6 lines 17, and FIG. 4-5: The seat sensors are placed based upon human anthropometry to accurately measure load and load distribution on the cushion by the wheelchair user during both static and dynamic seated postures. An algorithm capable of accommodating system creep over long durations of use can maintain the ability to accurately classify weight shift and in-seat movement activity. An algorithm is capable of classifying a continuum of weight-shifting activities ranging from partial load distribution resulting from leaning and reaching to complete unweighting of the cushion surface. The classification algorithm takes into account both magnitude and duration of loading to better distinguish different weight-shifting behaviors. Also, an algorithm can be used that is capable of monitoring the in-seat movement of a wheelchair users that results from transient re-distribution of cushion loading during dynamic changes in seated posture).
Therefore in view of teachings by Lustig, Tokuchi, Karunaratne, and Sonenblum, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, and Karunaratne to include classifying the user’s posture and a user behavior of the user other than the posture of the user based on the posture pressure distribution using a supervised machine learning (ML) classifier to generate a classification result, wherein classifying the user’s posture and the user behavior is performed directly by the supervised ML classifier without requiring additional physiological sensor data, environmental sensor data, or manual professional labeling, as suggested by Sonenblum. The motivation for this is to identify sitting posture of a user and provide suggestions to the user based on characterization data.
The combination of Lustig, Tokuchi, Karunaratne, and Sonenblum does not explicitly disclose wherein the recommended pose as a stretch pose selected from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose.
However, it has been known in the art of monitoring postures of users to implement wherein the recommended pose as a stretch pose selected from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose, as suggested by Bourahmoune, which discloses wherein the recommended pose as a stretch pose selected from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose (Bourahmoune: AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion: the heatmap for the stretch pose ‘Right Arm Cross’ (RAC) shows a high pressure reading on the upper left sensor. This captures accurately the pose performed by the experiment subjects where they extended the right arm across the chest and pressed the left arm on the right elbow. The pressure distribution heatmaps show that the six stretches are to some degree visually distinguishable from each other. Unlike sitting posture, a stretch is a static event that requires a conscious and dedicated effort from the user. Subjectivity in stretch interpretation and variable physical predisposition to perform the stretches result in greater inter-individual variability compared to the sitting posture recognition task).
Therefore in view of teachings by Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, Karunaratne, and Sonenblum to include wherein the recommended pose as a stretch pose selected from a plurality of candidate stretch poses by ranking the similarity metric between (i) a historic posture pressure distribution associated with the classified posture and (ii) each candidate stretch pose pressure distribution, and selecting a least similar stretch pose, as suggested by Bourahmoune. The motivation for this is to identify sitting posture of a user and provide suggestions to the user based on characterization data.
Claims 7-10 are rejected under 35 U.S.C. 103 as being unpatentable over Lustig et al. (Lustig – US 2019/0175076 A1) in view of Tokuchi et al. (Tokuchi – US 2019/0243958 A1), Karunaratne et al. (Karunaratne – US 9,795,322 B1) ), Sonenblum et al. (Sonenblum – US 10,357,186 B2) and Bourahmoune et al. (Bourahmoune - AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion), and further in view of James et al. (James – GB 2547495 A).
As to claim 7, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune discloses the limitations of claim 1 except for the claimed limitations of the posture detection system according to Claim 1, wherein the pressure sensor unit comprises:
a first layer including a plurality of sensing electrodes formed of conductive fabric or conductive tape;
a second layer including a conductive sheet with a variable resistance changing according to the pressure applied from the user; and
a third layer including at least one counter electrode placed to face the plurality of sensing electrodes, the counter electrode being formed of conductive fabric or conductive tape,
wherein the second layer is placed between the first and third layer.
However, it has been known in the art of designing pressure sensing devices to implement wherein the pressure sensor unit comprises: a first layer including a plurality of sensing electrodes formed of conductive fabric or conductive tape; a second layer including a conductive sheet with a variable resistance changing according to the pressure applied from the user; and a third layer including at least one counter electrode placed to face the plurality of sensing electrodes, the counter electrode being formed of conductive fabric or conductive tape, wherein the second layer is placed between the first and third layer, as suggested by James, which discloses wherein the pressure sensor unit comprises: a first layer including a plurality of sensing electrodes formed of conductive fabric or conductive tape (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, and FIG. 1-4 the conductive sheets 18, 20, and 24); a second layer including a conductive sheet with a variable resistance changing according to the pressure applied from the user (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, and FIG. 1-4 the conductive sheets 18, 20, and 24); and a third layer including at least one counter electrode placed to face the plurality of sensing electrodes, the counter electrode being formed of conductive fabric or conductive tape (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, and FIG. 1-4 the conductive sheets 18, 20, and 24), wherein the second layer is placed between the first and third layer (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, and FIG. 1-4 the conductive sheets 18, 20, and 24: First and second pressure sensors are provided in the cushion 10. The first and second pressure sensors are disposed in different horizontal planes in the cushion 10. In this embodiment, the pressure sensors are between the cushions 12, 14. The pressure monitors are also stacked one above the other within the cushion 10. In this embodiment, the first pressure sensor lies above the second pressure sensor in the cushion 10. The pressure sensors act as pressure switches. Fabric sheets 17 are provided between the top cushion 12 and the cushioned layer 16, about the first and second pressure sensors. The first pressure sensor includes two electrical conductors (or conductive sheets) 18, 20, and a resilient electrical insulator (or insulating layer) 22 disposed between the sheets 18, 20. The second pressure sensor also includes two conductive sheets 20, 24, and an insulating layer 26 disposed between the sheets 20, 24. In this embodiment, the pressure sensors share a common conductive sheet 20. However, other embodiments may have distinct conductive sheets for each pressure sensor).
Therefore in view of teachings by Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, and James, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune to include wherein the pressure sensor unit comprises: a first layer including a plurality of sensing electrodes formed of conductive fabric or conductive tape; a second layer including a conductive sheet with a variable resistance changing according to the pressure applied from the user; and a third layer including at least one counter electrode placed to face the plurality of sensing electrodes, the counter electrode being formed of conductive fabric or conductive tape, wherein the second layer is placed between the first and third layer, as suggested by James. The motivation for this is to implement a known specific design of pressure sensors for monitoring pressure applied by a user.
As to claim 8, Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, and James disclose the limitations of claim 7 further comprising the posture detection system according to Claim 7, wherein the plurality of sensing electrodes are formed of conductive tape, the plurality of sensing comprising electrodes being in contact with the second layer (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, page 12 lines 1-15, and FIG. 1-4 the conductive sheets 18, 20, and 24: In Figure 4, a person sat on the cushion 10 has compressed both the first and second pressure sensors, connecting all of the conductive sheets 18, 20, 24. This corresponds to excessively high pressure being exerted on seated parts of that person’s body. Remedial action may be needed soon, or immediately, to prevent increased risk of pressure ulcer formation).
As to claim 9, Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, and James disclose the limitations of claim 7 further comprising the posture detection system according to Claim 7, wherein
the pressure sensor unit further comprises a fourth layer (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, page 12 lines 1-15, and FIG. 1-4 the conductive sheets 18, 20, and 24 and the insulating layers 22 and 26: The first pressure sensor includes two electrical conductors (or conductive sheets) 18, 20, and a resilient electrical insulator (or insulating layer) 22 disposed between the sheets 18, 20. The second pressure sensor also includes two conductive sheets 20, 24, and an insulating layer 26 disposed between the sheets 20, 24. In this embodiment, the pressure sensors share a common conductive sheet 20. However, other embodiments may have distinct conductive sheets for each pressure sensor) placed between the first layer and the second layer and formed by a foam material (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, page 12 lines 1-15, and FIG. 1-4 the conductive sheets 18, 20, and 24 and the insulating layers 22 and 26),
the fourth layer includes a plurality of openings corresponding to the plurality of sensing electrodes, respectively (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, page 12 lines 1-15, and FIG. 1-4 the conductive sheets 18, 20, and 24 and the insulating layers 22 and 26: Each pressure sensor includes first and second electrical conductors 18, 20, 24 and a resilient electrical insulator 22, 26 disposed between the electrical conductors. Each electrical insulator includes one or more apertures 28, 30 through which the adjacent electrical conductors come into contact when a threshold pressure is applied to the cushion. The electrical conductors of the first pressure sensor come into contact at a first threshold pressure, and the electrical conductors of the second pressure sensor come into contact at a higher second threshold pressure), and
when the pressure applied from the user exceeds a predetermined value, the plurality of sensing electrodes is brought into contact with the conductive sheet through the opening (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, page 12 lines 1-15, and FIG. 1-4 the conductive sheets 18, 20, and 24 and the insulating layers 22 and 26: Applying pressure to cushion 10 compresses the insulating layers 22, 26 and allows the conductive sheets 18, 20, 24 to approach each other more closely. It can be seen from Figure 1 that the lower insulating layer 26 is around twice as thick as the upper insulating layer 22 in this embodiment. In this embodiment, the layers are 10 mm and 5 mm thick respectively. Therefore, the lower insulating layer is more difficult to compress. The actual thicknesses can be calibrated accordingly to the weight of the intended user. Overall, this gives the second pressure sensor a higher effective pressure threshold before its conductive sheets 20, 24 can come into contact, relative to the pressure threshold of the first pressure sensor).
As to claim 10, Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, and James disclose the limitations of claim 7 further comprising the posture detection system according to Claim 7, wherein the pressure sensor unit further comprises a fourth layer placed between the second layer and the third layer (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, page 12 lines 1-15, and FIG. 1-4 the conductive sheets 18, 20, and 24 and the insulating layers 22 and 26: The first pressure sensor includes two electrical conductors (or conductive sheets) 18, 20, and a resilient electrical insulator (or insulating layer) 22 disposed between the sheets 18, 20. The second pressure sensor also includes two conductive sheets 20, 24, and an insulating layer 26 disposed between the sheets 20, 24. In this embodiment, the pressure sensors share a common conductive sheet 20. However, other embodiments may have distinct conductive sheets for each pressure sensor),
the fourth layer includes a plurality of openings corresponding to the plurality of sensing electrodes, respectively (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, page 12 lines 1-15, and FIG. 1-4 the conductive sheets 18, 20, and 24 and the insulating layers 22 and 26: Each pressure sensor includes first and second electrical conductors 18, 20, 24 and a resilient electrical insulator 22, 26 disposed between the electrical conductors. Each electrical insulator includes one or more apertures 28, 30 through which the adjacent electrical conductors come into contact when a threshold pressure is applied to the cushion. The electrical conductors of the first pressure sensor come into contact at a first threshold pressure, and the electrical conductors of the second pressure sensor come into contact at a higher second threshold pressure), and
when the pressure applied from the user exceeds a predetermined value, the counter electrode is brought into contact with the conductive sheet through the opening (James: Abstract, page 5 lines 26-page 6 lines 11, page 8 lines 1 – page 9 lines 17, page 12 lines 1-15, and FIG. 1-4 the conductive sheets 18, 20, and 24 and the insulating layers 22 and 26: Applying pressure to cushion 10 compresses the insulating layers 22, 26 and allows the conductive sheets 18, 20, 24 to approach each other more closely. It can be seen from Figure 1 that the lower insulating layer 26 is around twice as thick as the upper insulating layer 22 in this embodiment. In this embodiment, the layers are 10 mm and 5 mm thick respectively. Therefore, the lower insulating layer is more difficult to compress. The actual thicknesses can be calibrated accordingly to the weight of the intended user. Overall, this gives the second pressure sensor a higher effective pressure threshold before its conductive sheets 20, 24 can come into contact, relative to the pressure threshold of the first pressure sensor).
Claims 12 and 23 are rejected under 35 U.S.C. 103 as being unpatentable over Lustig et al. (Lustig – US 2019/0175076 A1) in view of Tokuchi et al. (Tokuchi – US 2019/0243958 A1), Karunaratne et al. (Karunaratne – US 9,795,322 B1), Sonenblum et al. (Sonenblum – US 10,357,186 B2), and Bourahmoune et al. (Bourahmoune - AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion), and further in view of Yang et al. (Yang – US 2017/0092094 A1).
As to claim 12, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune disclose the limitations of claim 4 except for the claimed limitations of the posture detection system according to Claim 4, wherein
each of the sensors is configured to detect, as a reference pressure, a pressure when the user is sitting with his/her back leaning against the backrest with a reference posture,
the controller is configured to calculate a difference value between the reference pressure of each of the sensors and a current pressure, and
the controller is configured to calculate a balance in a left and right direction and a balance in a vertical direction based on the difference value of each of the sensors.
However, it has been known in the art of user posture detection to implement wherein
each of the sensors is configured to detect, as a reference pressure, a pressure when the user is sitting with his/her back leaning against the backrest with a reference posture,
the controller is configured to calculate a difference value between the reference pressure of each of the sensors and a current pressure, and
the controller is configured to calculate a balance in a left and right direction and a balance in a vertical direction based on the difference value of each of the sensors, as suggested by Yang, which discloses wherein
each of the sensors is configured to detect, as a reference pressure, a pressure when the user is sitting with his/her back leaning against the backrest with a reference posture (Yang: [0045]-[0046], [0051], [0054]-[0055], [0057], [0064], [0088], and FIG. 11: The control module 62 controls other modules 50 and may be controlled by the user, e.g., via the display module 58. For example, the control module 62 may turn on and/or turn off data acquisition of the acquisition module 52, may control the measurement parameters of the measurement module 54, may retrieve and/or update calibration data (e.g., activity thresholds) from/to the measurement module 54, and/or may determine which postures are calculated with the posture module 56 (as described further herein)),
the controller is configured to calculate a difference value between the reference pressure of each of the sensors and a current pressure (Yang: [0050]-[0066], [0070]-[0071], FIG. 5-10 and FIG. 12: As indicated in FIG. 13, calculating 104 may include determining 120 that the chair is in use, e.g., by determining that at least one sensor is active (i.e., indicates a body force, or related physical interaction, that is greater than a sensor-specific activity threshold)), and
the controller is configured to calculate a balance in a left and right direction (Yang: [0050]-[0066], [0070]-[0071], [0075], FIG. 5-10 and FIG. 12: Calculating 104 may include determining 130 that the posture type is leaning when the right armrest sensor value indicates a significantly different arm force than the left armrest sensor value. A significantly different arm force may be the magnitude of the difference of right armrest sensor value and the left armrest sensor value being greater than a predetermined threshold (e.g., an arm force imbalance threshold). The direction of the lean may be determined by the sign of the difference of the right armrest sensor value and the left armrest sensor value) and a balance in a vertical direction based on the difference value of each of the sensors (Yang: [0050]-[0066], [0070]-[0071], [0075]-[0076], FIG. 5-10 and FIG. 12: Calculating 104 may include determining 132 that the posture type is sitting on edge when the right shoulder sensor value indicates inactivity, the left shoulder sensor value indicates inactivity, the lumbar sensor value indicates inactivity, the right buttocks sensor value indicates inactivity, the left buttocks sensor value indicates inactivity, the right thigh sensor value indicates activity, and the left thigh sensor value indicates activity).
Therefore, in view of teachings by Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, and Yang, it would have been obvious to one of the ordinary skill in the art before ethe effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune, to include wherein
each of the sensors is configured to detect, as a reference pressure, a pressure when the user is sitting with his/her back leaning against the backrest with a reference posture,
the controller is configured to calculate a difference value between the reference pressure of each of the sensors and a current pressure, and
the controller is configured to calculate a balance in a left and right direction and a balance in a vertical direction based on the difference value of each of the sensors, as suggested by Yang. The motivation for this is to determine a user posture based on sensing information located at predetermined locations on a chair.
As to claim 23, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune disclose the limitations of claim 1 except for the claimed limitations of the posture detection system according to Claim 1, wherein when the posture classified by the controller continues for a predetermined period or longer, the haptic feedback is configured to provide the feedback by the vibration.
However, it has been known in the art of user posture detection to implement when the posture classified by the controller continues for a predetermined period or longer, the haptic feedback is configured to provide the feedback by the vibration, as suggested by Yang, which discloses when the posture classified by the controller continues for a predetermined period or longer, the haptic feedback is configured to provide the feedback by the vibration (Yang: Abstract, [0004], [0083]-[0086], and FIG. 12: Methods include reading sensor values from sensors in an ergonomics awareness chair, calculating a posture type based on the sensor values, determining a time the user has been continuously sitting properly, and, if the time is greater than a predetermined threshold time, alerting the user to take a break… Alerting 114 may include displaying a message to the user on a user-interface computer. The message may be visual, audio, and/or tactile. If the alerting 114 includes indicating a posture type, alerting 114 may include displaying corrective and/or preventative action instructions. If the alerting 114 includes indicating a break is due, alerting 114 may include indicating the duration of the break and/or the end of the break).
Therefore, in view of teachings by Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, and Yang, it would have been obvious to one of the ordinary skill in the art before ethe effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune, to include when the posture classified by the controller continues for a predetermined period or longer, the haptic feedback is configured to provide the feedback by the vibration, as suggested by Yang. The motivation for this is to inform a user to take a break in response to detecting the user has been sitting for a long period of time.
Claim 18 is rejected under 35 U.S.C. 103 as being unpatentable over Lustig et al. (Lustig – US 2019/0175076 A1) in view of Tokuchi et al. (Tokuchi – US 2019/0243958 A1) and Karunaratne et al. (Karunaratne – US 9,795,322 B1), Sonenblum et al. (Sonenblum – US 10,357,186 B2), and Bourahmoune et al. (Bourahmoune - AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion), and further in view of Hu (Hu – US 20160089059 A1).
As to claim 18, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune disclose the limitations of claim 1 further comprising the posture detection system according to Claim 1, wherein the controller comprises a data storage unit configured to store the detection data of the pressure sensor unit (Lustig: [0105], [0110], and FIG. 6 the memory 607: In some exemplary embodiments, Memory 607 may retain program code operative to cause Processor 602 to perform acts associated with any of the subcomponents of Apparatus 600. Memory 607 may retain readings obtained from sensors, classifiers used for posture estimation and feedback selection, feedback history, or the like. Memory 607 may retain rules, selection rules, or the like, used by Apparatus 600), except for the claimed limitations of wherein the controller comprises a data storage unit configured to store the detection data of the pressure sensor unit for a plurality of the users, and the controller is configured to refer to the data stored in the data storage unit and identify the user according to the result of the detection by the pressure sensor unit.
However, it has been known in the art of determining user posture to implement the controller comprises a data storage unit configured to store the detection data of the pressure sensor unit for a plurality of the users, and
the controller is configured to refer to the data stored in the data storage unit and identify the user according to the result of the detection by the pressure sensor unit, as suggested by Hu, which discloses the controller comprises a data storage unit configured to store the detection data of the pressure sensor unit for a plurality of the users (Hu: [0039], [0083], [0085], [0091], FIG. 5 and FIG. 10: In some embodiments, the database of acceptable and/or unacceptable posture data includes data collected from the user's past use history. In other embodiments, the database of acceptable and/or unacceptable posture data includes data collected from a plurality of other users. In other embodiments, the database includes medically recommended values or ranges of values), and
the controller is configured to refer to the data stored in the data storage unit and identify the user according to the result of the detection by the pressure sensor unit (Hu: [0014], [0039], [0049], [0051], [0083], [0085], [0087]-[0089], [0091], and FIG. 10: In other embodiments, the database is stored directly on the portable computing device. In some embodiments, the database of acceptable and/or unacceptable posture data includes data collected from the user's past use history. In other embodiments, the database of acceptable and/or unacceptable posture data includes data collected from a plurality of other users. In other embodiments, the database includes medically recommended values or ranges of values).
Therefore, in view of teachings by Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, and Hu, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune to include the controller comprises a data storage unit configured to store the detection data of the pressure sensor unit for a plurality of the users, and
the controller is configured to refer to the data stored in the data storage unit and identify the user according to the result of the detection by the pressure sensor unit, as suggested by Hu. The motivation for this is to remotely monitor conditions of a user.
Claim 19 is rejected under 35 U.S.C. 103 as being unpatentable over Lustig et al. (Lustig – US 2019/0175076 A1) in view of Tokuchi et al. (Tokuchi – US 2019/0243958 A1), Karunaratne et al. (Karunaratne – US 9,795,322 B1), Sonenblum et al. (Sonenblum – US 10,357,186 B2), and Bourahmoune et al. (Bourahmoune - AI-powered Posture Training: Application of Machine Learning in Sitting Posture Recognition Using the LifeChair Smart Cushion), and further in view of Benson et al. (Benson – US 10,786,162 B2) and Kogure et al. (Kogure – US 2021/0106256 A1).
As to claim 19, Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune disclose the limitations of claim 4 except for the claimed limitations of the posture detection system according to Claim 4, further comprising a vibration sensor provided in the backrest configured to detect a vibration applied from the user, wherein the vibration sensor is configured to detect the user’s heart beats per minute or respiration rate according to a result of the detection of the vibration sensor.
However it has been known in the art of seating device to implement a sensor provided in the backrest configured to detect a vibration applied from the user, wherein the sensor is configured to detect the user’s heart beats per minute or respiration rate according to a result of the detection of the sensor, as suggested by Benson, which discloses a sensor provided in the backrest (Benson: FIG. 1 the ECG sensor system 18 disposed in the seat back 14) configured to detect a vibration applied from the user (Benson: Abstract, column 4 lines 41-45, column 7 lines 26 – column 8 lines 5, FIG. 1, and FIG. 8: First and second ECG receivers 24, 26 and ECG mat 28 cooperate to provide an ECG sensor 34. ECG sensor 34 is coupled to a seat cushion 36 and surrounded by trim 38 as shown in FIG. 2. ECG sensor 34 is configured to provide means for detecting electrical signals in occupant 50 through first, second, and N.sup.th clothing layers 41, 42, and 43N as shown in FIG. 2), wherein the sensor is configured to detect the user’s heart beats per minute or respiration rate according to a result of the detection of the sensor (Benson: Abstract, column 7 lines 26 – column 8 lines 5, column 11 lines 15 – column 12 lines 16 FIG. 1, FIG. 8 and FIG. 11: Computer 54 uses the ECG signal to determine heart rate 61, heart-rate variability 65, stress level 64, a pulse-transit time 66, and blood pressure 62 as shown in FIG. 9. ECG signal 58 is obtained when first and second ECG receivers 24, 26 sense electrical signals in occupant 50. Based on the output of the processing, computer 54 may perform a predetermined action. The predetermined action may be storing the calculated values in memory 542 of computer 54. The predetermined action may be activating output 56 to communicate the output to the occupant).
Therefore in view of teachings by Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, and Benson, it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, Karunaratne, Sonenblum, and Bourahmoune, to include a sensor provided in the backrest configured to detect a vibration applied from the user, wherein the sensor is configured to detect the user’s heart beats per minute or respiration rate according to a result of the detection of the sensor, as suggested by Benson. The motivation for this is to monitor conditions of a user sitting in a chair.
The combination of Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, and Benson does not explicitly disclose the sensor as the vibration sensor.
However it has been known in the art of seating device to implement the sensor as the vibration sensor, as suggested by Kogure, which discloses the sensor as the vibration sensor (Kogure: Abstract, [0033]-[0034], [0044]-[0046], and FIG. 1-5: The first calculation unit 120 is configured to acquire a biological signal of the user from the vibration data, and calculate biological information values (such as a breathing rate, a heart rate, and the amount of activity). In this embodiment, the first calculation unit extracts a breathing component and a heart beat component from the vibration (body vibration) data acquired by the detection unit 110. The first calculation unit 110 may obtain a breathing rate and a heart rate from the extracted breathing component and heart beat component based on a breathing interval and a heart beat interval. Alternatively, the first calculation unit 120 may analyze (e.g. Fourier transform) the periodicity of the vibration data and calculate a breathing rate and a heart rate from the peak frequency).
Therefore in view of teachings by Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, Benson, and Kogure it would have been obvious to one of the ordinary skill in the art before the effective filing date of the claimed invention to implement in the posture detection system of Lustig, Tokuchi, Karunaratne, Sonenblum, Bourahmoune, and Benson to include the sensor as the vibration sensor, as suggested by Kogure. The motivation for this is to implement a known types of sensor for monitoring conditions of a user.
Citation of Pertinent Art
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure:
Imoto et al., US 2024/0242842 A1, discloses information processing apparatus, information processing method, and program.
Kaku et al.US 2024/0131397 A1, discloses service providing apparatus.
Walsh et al., US 2011/0275939 A1, discloses ergonomic sensor pad with feedback to user and method of use.
Bourahmoune et al., 2022, Intelligent Posture Training: Machine-Learning-Powered Human Sitting Posture Recognition Based on a Pressure-Sensing IoT Cushion.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to QUANG PHAM whose telephone number is (571)-270-3668. The examiner can normally be reached 09:00 AM - 05:00 PM.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, QUAN-ZHEN WANG can be reached at (571)-272-3114. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/QUANG PHAM/Primary Examiner, Art Unit 2685