DETAILED ACTION
This nonfinal action is in response to application 18/083,802 filed on 12/19/2022. Claims 1-14 remain pending in the application. Claims 1, 13, and 14 are independent claims.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Objections
Claims 1-2 and 13-14 are objected to because of the following informalities:
In claim 1, “A neural network-based method for detecting validity of a human movement” should read ““A neural network-based method for detecting validity of a human body movement” or be likewise amended to establish proper antecedent basis for further recitations of “the human body movement” in the claims.
In claims 2, 13, and 14 recitations of “the detection model” (e.g., “training the detection model” in claim 2, “wherein the detection model comprises” in claims 13-14) should read “the auto-encoder neural network-based detection model” or be likewise amended to avoid ambiguity and maintain consistency in claim terminology.
Appropriate corrections are required.
Claim Interpretation
As recited in MPEP § 2111, during patent examination, “the pending claims must
be given their broadest reasonable interpretation consistent with the specification”.
Under a broadest reasonable interpretation (BRI), claim terms must be given their plain
and ordinary meaning (i.e., the meaning that the term would have to a person of
ordinary skill in the art), unless applicant sets forth a special definition of a claim term
within the specification. The plain and ordinary meaning of a term “may be evidenced by
a variety of sources, including the words of the claims themselves, the specification,
drawings, and prior art”.
Independent claim 1 (and corresponding claims 13-14) recite the limitation “extracting a feature text based on the user movement dataset, wherein the feature text comprises a plurality of input sequences, and each of the input sequences
comprises at least a direction, a distance, and a step count”. Dependent claim 2 similarly recites “extracting a feature text based on the training dataset, wherein the feature text comprises a plurality of input sequences, and each of the plurality of the input sequences comprises at least a direction, a distance and a step count”.
The specification does not set forth description to particularly describe the contents of extracted “feature text”, and at most sets forth an example “pre-encoding process” [¶ 0039] wherein obtained input sequences comprise a series of movement steps, such as "2 meters and 3 steps towards the southwest; 1 meter and 2 steps towards the west; and 4 meters and 7 steps towards the southwest". However, the specification also describes the datasets at issue (“user movement dataset” and “training dataset”, from which feature text is extracted) as being constructed directly from raw sensor data (e.g., accelerometer/gyroscope of smart device [¶ 0010, 0016, 0023]), without setting forth an explicit procedure explaining or detailing how these raw measurement values are then converted into sentences describing a series of movements (as appears to be suggested by [¶ 0039]).
The limited disclosure of the specification thereby necessitates a broad interpretation of the term “feature text” in the claims, which is herein interpreted as encompassing a collection of features drawn from collected movement data (user movement dataset, training dataset) that is representable as a human-readable sequence of characters (i.e., numbers, vectors, etc).
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 2 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claim 2, it recites the limitation “using a minimization of a second loss function value of the second initial encoding and the predicted encoding as an objective function, until the loss function value is lower than a preset threshold”. There is insufficient antecedent basis for the claim terms “the predicted encoding” and “the loss function value” in the claims; the claim previously recites both “a first/second predicted encoding” and “a first/second loss function value” (in light of being dependent on parent claim 1), such that it is unclear which of the previously recited terms is being referred to. Consequently, one of ordinary skill in the art would not be reasonably apprised of the scope of the invention.
For purposes of examination and as best understood in light of the specification, the limitation is interpreted as “using a minimization of a second loss function value of the second initial encoding and the second predicted encoding as an objective function, until the second loss function value is lower than a preset threshold”.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-14 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The analysis of the claims will follow the 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (“2019 PEG”).
Independent Claims (Claim 1, Claim 13, Claim 14):
Step 1: Claim 1 is drawn to a method, claim 13 is drawn to a product, and claim 14 is drawn to an apparatus. Therefore, each of these claims falls under one of the four categories of statutory subject matter (process/method, machine/apparatus, manufacture/product, or composition of matter).
Step 2A Prong 1: Claims 1, 13, and 14 each recite a judicially recognized exception of an abstract idea.
Claim 1 recites, inter alia:
A method for detecting validity of a human movement, comprising: extracting a feature text based on the user movement dataset, wherein the feature text comprises a plurality of input sequences, and each of the plurality of input sequences comprises at least a direction, a distance, and a step count; transforming the feature text into a first initial encoding according to a preset coding rule; and determin[ing] whether the human body movement corresponding to the user movement dataset is valid – These limitations amount to a procedure of observing performance of a sequence of human movement, noting observed measurements of the movements performed in said sequence, and further analyzing said measurements via a series of data manipulations to determine “validity” (e.g., proper form/consistency) of the performed movement. They therefore recite a process of evaluation that a human could reasonably perform in the mind or using pen and paper.
map the first initial encoding into a first predicted encoding; wherein determining whether the human body movement corresponding to the user movement dataset is valid further comprises: obtaining a first loss function value based on the first initial encoding and the first predicted encoding, and determining whether the first loss function value is lower than a preset threshold; determining that the human body movement corresponding to the user movement dataset is valid in response to the first loss function value being lower than the preset threshold; and determining the human body movement corresponding to the user movement dataset is invalid in response to the first loss function value being not lower than the preset threshold – These limitations further recite a procedure of performing data manipulation, calculation, comparison and analysis to determine “validity” of a human movement, and thereby further recite a process of evaluation that a human could reasonably perform in the mind or using pen and paper.
Claims 13 and 14 recite substantially similar abstract idea limitations to those recited in claim 1, and therefore recite the same judicial exception.
Step 2A Prong 2: The following additional elements recited in claims 1, 13, and 14 do not integrate the recited judicial exceptions into a practical application.
Claim 1 additionally recites:
A neural network-based method, comprising; inputting the first initial encoding into an auto-encoder neural network-based detection model [to determine whether movement is valid]; wherein the auto-encoder neural network-based detection model comprises an encoder and a decoder, the encoder is configured to [map encoding] to an intermediate variable comprising a compressed eigenvector, and the decoder is configured to translate the intermediate variable [into predicted encoding] – These limitations do no more than generically invoke an autoencoder (i.e., encoder-decoder) model as a tool to perform an existing procedure of data manipulation and analysis, and thereby amount to mere instructions to apply an exception using a known computational architecture.
obtaining a user movement dataset, the user movement dataset comprising a plurality of n-dimensional movement data items corresponding to the human body movement; – These limitations amount no more than a process of gathering and organizing data to enable further analysis, and therefore recite insignificant extra-solution activity.
Claim 13 recites substantially similar abstract idea limitations to those recited in claim 1, and further recites:
A non-transitory computer-readable medium having stored content, wherein the stored content causes a computing system to perform automated operations, comprising: – This limitation amounts to mere instructions to implement an abstract idea on a computer or computer components.
Claim 14 recites substantially similar abstract idea limitations to those recited in claim 1, and further recites:
A system, comprising: one or more processors; a wireless communication module configured to obtain a user movement dataset; and at least one memory having stored instructions that, when executed by at least one of the one or more processors, cause the system to perform automated operations comprising: – These limitations amount to mere instructions to implement an abstract idea on a computer or computer components.
Step 2B: The additional elements recited in claims 1, 13, and 14, viewed individually or as an ordered combination, do not provide an inventive concept or otherwise amount to significantly more than the recited abstract ideas themselves.
Claim 1 additionally recites:
A neural network-based method, comprising; inputting the first initial encoding into an auto-encoder neural network-based detection model [to determine whether movement is valid]; wherein the auto-encoder neural network-based detection model comprises an encoder and a decoder, the encoder is configured to [map encoding] to an intermediate variable comprising a compressed eigenvector, and the decoder is configured to translate the intermediate variable [into predicted encoding] – Leveraging commonly utilized methods of factor analysis (e.g., eigen decomposition via principal component analysis (PCA)) when compressing data into latent space via autoencoder architectures is well-understood, routine, and conventional activity (see Ghojogh et al., “Factor Analysis, Probabilistic Principal Component Analysis, Variational Inference, and Variational Autoencoder: Tutorial and Survey” [Abstract and pages 1-2 Introduction]); as such, mere instructions to apply an exception using known computational architectures do not provide an inventive concept or significantly more to the recited abstract idea.
obtaining a user movement dataset, the user movement dataset comprising a plurality of n-dimensional movement data items corresponding to the human body movement; – Procedures of obtaining/retrieving data to enable further analysis are well-understood, routine, and conventional activity (see MPEP § 2106.05(d); “Storing and retrieving information in memory”, “Receiving or transmitting data over a network”), and therefore do not provide an inventive concept or significantly more to the recited abstract idea.
Claim 13 recites substantially similar abstract idea limitations to those recited in claim 1, and further recites:
A non-transitory computer-readable medium having stored content, wherein the stored content causes a computing system to perform automated operations, comprising: – Mere instructions to implement an abstract idea on a computer or computer components do not provide an inventive concept or significantly more to the recited abstract idea.
Claim 14 recites substantially similar abstract idea limitations to those recited in claim 1, and further recites:
A system, comprising: one or more processors; a wireless communication module configured to obtain a user movement dataset; and at least one memory having stored instructions that, when executed by at least one of the one or more processors, cause the system to perform automated operations comprising: – Mere instructions to implement an abstract idea on a computer or computer components do not provide an inventive concept or significantly more to the recited abstract idea.
Even when considered as an ordered combination, the additional elements recited in the claims ultimately do no more than place the claims in the context of invoking a generic autoencoder model as a tool to perform an abstract procedure of data manipulation and analysis. As such, claims 1, 13, and 14 are not patent eligible.
Dependent Claims (Claims 2-12):
Dependent claims 2-12 narrow the scope of independent claim 1, and likewise narrow the recited judicial exceptions. They recite abstract idea limitations that are similar to those recited within the independent claim (i.e., mental processes and/or mathematical concepts), and thereby merely expand on the already recited exception. The dependent claims also do not recite any further additional elements that successfully integrate the recited judicial exceptions into a practical application or provide significantly more than the recited abstract ideas themselves. Consequently, claims 2-12 are also rejected under 35 U.S.C. 101.
Step 1: Claims 2-12 are drawn to a method. Therefore, each of these claims falls under one of the four categories of statutory subject matter (process/method, machine/apparatus, manufacture/product, or composition of matter).
Step 2A Prong 1: Claims 2-12 each recite a judicially recognized exception of an abstract idea.
Claim 2 recites, inter alia:
extracting a feature text based on the training dataset, wherein the feature text comprises a plurality of input sequences, and each of the plurality of the input sequences comprises at least a direction, a distance and a step count; transforming the feature text into a second initial encoding according to a preset coding rule; obtain a second predicted encoding [from the second initial encoding] – These limitations further amount to a procedure of observing performance of a sequence of human movement, noting observed measurements of the movements performed in said sequence, and further analyzing said measurements via a series of data manipulations to determine “validity” (e.g., proper form/consistency) of the performed movement. They therefore recite a process of evaluation that a human could reasonably perform in the mind or using pen and paper.
[training] by using a minimization of a second loss function value of the second initial encoding and the predicted encoding as an objective function, until the loss function value is lower than a preset threshold – This limitation amounts to a procedure of using mathematical methods of optimization (minimize[ing] a loss function) to perform manipulation of data, and therefore recites mathematical calculation.
Claim 3 recites, inter alia:
wherein the first loss function value is a cross entropy of the first initial encoding and the first predicted encoding – This limitation further amounts to reciting a mathematical relationship between values.
Claim 4 recites, inter alia:
wherein the second loss function value is a cross entropy of the second initial encoding and the second predicted encoding – This limitation further amounts to reciting a mathematical relationship between values.
Claims 5-7 recite the same judicial exception as claim 1.
Claims 8-9 recite the same judicial exception as claim 2.
Claim 10 recites, inter alia:
dividing the user movement dataset into a plurality of subsets, such that a change rate in at least one of preset dimensions of the plurality of n-dimensional movement data items in each of the plurality of subsets does not exceed a preset value, and determining for each of the plurality of subsets whether a human body movement of the corresponding subset is valid – This limitation recites a pre-processing procedure of organizing data into groups based on implementation of preset conditions prior to making further determinations, and therefore recites a process of evaluation that a human could reasonably perform in the mind or using pen and paper.
Claim 11 recites, inter alia:
tagging movement data items collected during a pause period in response to a user's indication to pause a movement – This limitation amounts to a procedure of recognizing and tracking “pauses” in movement during an observation period, and therefore recites a process of evaluation that a human could reasonably perform in the mind or using pen and paper.
Claim 12 recites the same judicial exception as claim 1.
Step 2A Prong 2: Claims 3-4 and 10 do not recite any further additional elements besides those recited in the independent claims, and the following additional elements recited in claims 2, 5-9, and 11-12 also do not integrate the recited judicial exceptions into a practical application.
Claim 2 additionally recites:
training the detection model by a training dataset, the training dataset comprising a plurality of n-dimensional movement data items corresponding to the human body movement; and in at least one iterative loop, inputting the second initial encoding into the auto-encoder neural network-based detection model to [obtain predicted encoding]; and training the detection model [by using a minimization] – These limitations do no more than generically invoke an autoencoder (i.e., encoder-decoder) model as a tool to perform an existing procedure of data manipulation and analysis, and thereby amount to mere instructions to apply an exception using a known computational architecture.
Claim 5 additionally recites:
wherein each of the plurality of n-dimensional movement data items comprises at least longitude information, latitude information, and accumulated steps information – This limitation merely specifies a particular data source / type of data to be manipulated, and therefore recites insignificant extra-solution activity.
Claim 6 additionally recites:
wherein each of the plurality of n-dimensional movement data items further comprises at least one of time information, instantaneous movement speed, and altitude information – This limitation merely specifies a particular data source / type of data to be manipulated, and therefore recites insignificant extra-solution activity.
Claim 7 additionally recites:
wherein each of the plurality of n-dimensional movement data items further comprises at least one of a pulse rate, a body temperature, a blood oxygen value, and a blood pressure value – This limitation merely specifies a particular data source / type of data to be manipulated, and therefore recites insignificant extra-solution activity.
Claim 8 additionally recites:
collecting movement data items at a preset frequency by using at least one sensor in a mobile device to construct the user movement dataset or the training dataset – This limitation merely specifies a particular data source / type of data to be manipulated, and therefore recites insignificant extra-solution activity.
Claim 9 additionally recites:
collecting movement data items by invoking at least one of a GPS positioning service, an application program, a gyroscope of a smart device, or a combination thereof, to construct one of the training dataset and the user movement dataset – This limitation merely specifies a particular data source / type of data to be manipulated, and therefore recites insignificant extra-solution activity.
Claim 11 additionally recites:
removing the tagged movement data items when obtaining the user movement dataset – This limitation amounts to an intermediary processing step of managing the storage and retrieval of data, and therefore recites insignificant extra-solution activity.
Claim 12 additionally recites:
wherein the auto-encoder neural network based detection model is deployed on a device terminal with edge computing capability, and the method is executed by the device terminal – This limitation amounts to no more than mere instructions to implement an abstract procedure on a type of computer (e.g., smartphones / other “edge devices” of a network) or computing components.
Step 2B: The additional elements recited in claims 2, 5-9, and 10-12, viewed individually or as an ordered combination, do not provide an inventive concept or otherwise amount to significantly more than the recited abstract ideas themselves.
Claim 2 additionally recites:
training the detection model by a training dataset, the training dataset comprising a plurality of n-dimensional movement data items corresponding to the human body movement; and in at least one iterative loop, inputting the second initial encoding into the auto-encoder neural network-based detection model to [obtain predicted encoding]; and training the detection model [by using a minimization] – Generically invoking an autoencoder (i.e., encoder-decoder) model as a tool to perform an existing procedure of data manipulation and analysis does not provide an inventive concept or significantly more to the recited abstract idea.
Claim 5 additionally recites:
wherein each of the plurality of n-dimensional movement data items comprises at least longitude information, latitude information, and accumulated steps information – Leveraging global positioning system (GPS) data drawn from wearable sensors in the field of human activity recognition (HAR) is well-understood, routine, and conventional activity (see Chen et al., “Sensor-Based Activity Recognition” [Abstract and pages 790-792 Introduction]) and therefore does not provide an inventive concept or significantly more to the recited abstract idea.
Claim 6 additionally recites:
wherein each of the plurality of n-dimensional movement data items further comprises at least one of time information, instantaneous movement speed, and altitude information – Leveraging accelerometer data drawn from wearable sensors in the field of HAR is well-understood, routine, and conventional activity (see Chen et al., “Sensor-Based Activity Recognition” [Abstract and pages 790-792 Introduction]) and therefore does not provide an inventive concept or significantly more to the recited abstract idea.
Claim 7 additionally recites:
wherein each of the plurality of n-dimensional movement data items further comprises at least one of a pulse rate, a body temperature, a blood oxygen value, and a blood pressure value – Leveraging vital signs data drawn from wearable sensors in the field of HAR is well-understood, routine, and conventional activity (see Chen et al., “Sensor-Based Activity Recognition” [Abstract, pages 790-792 Introduction, pages 792-793 Wearable Sensor-Based Activity Monitoring]) and therefore does not provide an inventive concept or significantly more to the recited abstract idea
Claim 8 additionally recites:
collecting movement data items at a preset frequency by using at least one sensor in a mobile device to construct the user movement dataset or the training dataset – Leveraging sensor data drawn from smartphones/mobile devices in the field of HAR is well-understood, routine, and conventional activity (see Chen et al., “Sensor-Based Activity Recognition” [Abstract and pages 790-792 Introduction]) and therefore does not provide an inventive concept or significantly more to the recited abstract idea.
Claim 9 additionally recites:
collecting movement data items by invoking at least one of a GPS positioning service, an application program, a gyroscope of a smart device, or a combination thereof, to construct one of the training dataset and the user movement dataset – Leveraging sensor data drawn from smartphones/mobile devices in the field of HAR is well-understood, routine, and conventional activity (see Chen et al., “Sensor-Based Activity Recognition” [Abstract and pages 790-792 Introduction]) and therefore does not provide an inventive concept or significantly more to the recited abstract idea.
Claim 11 additionally recites:
removing the tagged movement data items when obtaining the user movement dataset – Managing stored/retrieved information is well-understood, routine, and conventional activity (see MPEP § 2106.05(d); “Storing and retrieving information in memory”) and therefore does not provide an inventive concept or significantly more to the recited abstract idea.
Claim 12 additionally recites:
wherein the auto-encoder neural network based detection model is deployed on a device terminal with edge computing capability, and the method is executed by the device terminal – Mere instructions to implement an abstract idea on a computer or computer components do not provide an inventive concept or significantly more to the recited abstract idea.
Even when considered as an ordered combination, the additional elements recited in the claims ultimately do no more than place the claims in the context of utilizing various commonly utilized types of sensor data when performing the recited abstract procedure and/or generically invoking an autoencoder model as a tool to perform the recited abstract procedure. As such, claims 2-12 also are not patent eligible.
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 (i.e., changing from AIA to pre-AIA ) 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, 8-10, and 12-14 are rejected under 35 U.S.C. 103 as being unpatentable over Bonetto et al., (“Seq2Seq RNN based Gait Anomaly Detection from Smartphone Acquired Multimodal Motion Data”, available arXiv 9 Nov 2019), hereinafter Bonetto, in view of Wang (“Recognition of Human Activities Using Continuous Autoencoders with Wearable Sensors”, published 4 February 2016).
Regarding claim 1, Bonetto teaches A neural network-based method for detecting validity of a human movement (“Here, we are concerned with gait analysis systems that extract meaningful information from a user’s movements to identify anomalies and changes in their walking style. The solution that is put forward is subject-specific, as the designed feature extraction and classification tools are trained on the subject under observation. A smartphone mounted on an ad-hoc made chest support is utilized to gather inertial data and video signals from its built-in sensors and rear-facing camera. The collected video and inertial data are preprocessed, combined and then classified by means of a Recurrent Neural Network (RNN) based Sequence-to-Sequence (Seq2Seq) model, which is used as a feature extractor, and a following Convolutional Neural Network (CNN) classifier. This architecture provides excellent results, being able to correctly assess anomalies in 100% of the cases, for the considered tests, surpassing the performance of support vector machine classifiers” [Bonetto Abstract]), comprising,
obtaining a user movement dataset, the user movement dataset comprising a plurality of n-dimensional movement data items corresponding to the human body movement (“As a relatively cheap and convenient way to acquire this data, we opted for a smartphone application, since camera and inertial sensors are already integrated into the device. Hence, we developed a chest support for smartphones (showed in Fig. 1(a)) and a motion sensing application that records accelerometer, gyroscope, and magnetometer measurements, while also recording a video sequence from the front camera” [Bonetto pages 2-3 Data Processing]; “Inertial signals are provided by built-in gyroscopic, magnetometric and accelerometric sensors. At each sampling epoch, each of these sensors returns a three-dimensional real sample (3-axes, for a total of 9-axes for the three sensors), related to the motion of the device along the three dimensions of the smartphone reference system” [Bonetto page 3 Inertial Data Acquisition and Synchronization];)
extracting a feature text based on the user movement dataset, wherein the feature text comprises a plurality of input sequences, and each of the plurality of input sequences comprises at least a direction, a distance, and a step count; (“Although the sampling frequency is high (typically around 100/200 samples/s, depending on the device), the time interval between consecutive samples is not constant. This is consistent with the findings of [35]. To cope with this, interpolation and resampling are performed prior to data analysis to convert the signals into the common sampling frequency of 200 Hz, as done in prior work [36]” [Bonetto page 3 Inertial Data Acquisition and Synchronization]; “The human gait follows a cyclic behavior featuring a periodic repetition of a pattern delimited by two consecutive steps. The stance phase starts with the instant when contact is made with the ground (usually occurring with the heel touching the ground first); this instant is called Initial Contact (IC). After that, the foot becomes flat on the ground and supports the full body weight (foot flat). Then, the heel begins to lift off the ground in preparation for the forward propulsion of the body and we finally have the take off phase, which ends the stance and is delimited by the instant of Final Contact (FC) of the foot with the ground. Afterwards, the weight of the body is moved to the other foot until the next IC occurs (swing time). A gait cycle (also referred to as stride or walking cycle) is defined as the time instants between two consecutive Initial Contacts (ICs) of the same foot. A pictorial representation of ICs and FCs is given in Fig. 3. IC and FC instants can be identified by analyzing the vertical component of the accelerometer data… It is thus possible to locate the walking cycles vectors in all the available signals. Each gait cycle is then normalized to a fixed length of 200 samples, stored on a descriptor and classified with the technique of Section II-B” [Bonetto page 4 Gait Cycle Extraction]; Multi-dimensional inertial signals provided by built-in gyroscopic, magnetometric and accelerometric sensors (i.e., tracked motion data indicating direction and distance) are sequenced based on sampling frequency, from which vectors (i.e., feature text) tracking walking cycles are further extracted from the signals and normalized to a length of 200 samples (i.e., sequences))
transforming the feature text into a first initial encoding according to a preset coding rule; (“Detrending: The signals extracted from the reference video have trend components on each of the three axes, which can heavily impact the data normalization, see Fig. 5 for examples. To remove such trends and obtain semi-stationary timeseries, we have fitted a linear model for each gait cycle. Then, the slope of the computed model has been used to remove the trend from the corresponding gait cycle… Normalization Data normalization is a standard step in data preparation for neural network based learning. It is performed by removing the mean and dividing the data by its standard deviation, computing these measures over the entire dataset. Here, this operation is performed for each of the 9 signals in a gait cycle” [Bonetto page 4 Detrending and Normalization]; Detrending and normalization are standard data pre-processing steps (i.e., coding rules) prior to inputting data to RNN for embedding into feature space) and
inputting the first initial encoding into an auto-encoder neural network-based detection model to determine whether the human body movement corresponding to the user movement dataset is valid; wherein the auto-encoder neural network-based detection model comprises an encoder and a decoder, the encoder is configured to map the first initial encoding to an intermediate variable, and the decoder is configured to translate the intermediate variable into a first predicted encoding; (“Next, we present an RNN based architecture to embed the gait signals into a feature space containing a fixed size, higher order representation of the gaits… In this work, we design an RNN-Seq2Seq encoder-decoder architecture to embed multi-dimensional fixed-length timeseries into a feature space that is suitable for a subsequent classification task. To achieve this, we utilize a deep RNN as the encoder and a shallow NN with linear activations as the decoder. In Fig. 7, we show the architecture of the designed Seq-2-Seq model. First, the multi-dimensional timeseries corresponding to a gait cycle is fed into the RNN encoder. Then, the output of the encoder is fed to the shallow linear decoder. The whole neural network chain is trained to reproduce the input sequence at its output. The final state of the encoder, is then utilized as a feature vector representing the embedding of the input sequence into a feature space that retains temporal information about the processed gait cycle” [Bonetto pages 4-5 Gait RNN Based Embedding]; The disclosed RNN-Seq2Seq encoder-decoder performs processing to embed the gait signals into feature space, wherein said feature vectors are further passed to a classifier to identify anomalies (i.e., determine validity))
wherein determining whether the human body movement corresponding to the user movement dataset is valid further comprises:
obtaining a first loss function value based on the first initial encoding and the first predicted encoding, and determining whether the first loss function value is lower than a preset threshold; (“To obtain encodings that are suitable for the subsequent classification task, the RNN-Seq2Seq module is only trained on “normal” walking cycles. This is because by learning to only reproduce normal gaits, when anomalous ones are given as input, the module should be unable to determine a correct embedding, and this is detected with high probability by a subsequent classifier” [Bonetto page 6 Training]; “Once the trained encoder has processed the whole sequence X, as explained in Section II-B, its final state SL−1 contains the higher order representation of the input time series. SL−1 is a multidimensional vector (i.e., a tensor)…. The multidimensional matrix S˜L−1 is fed into a multilayer CNN with REctified Linear Unit (RELU) activations….The CNN output is then flattened and fed to a 2-units layer with logistic activation functions. Hence, the final classification is obtained by means of a softmax layer…according to the notation used in Section II-B, at the output of the classification layer we obtain:
PNG
media_image1.png
82
397
media_image1.png
Greyscale
where indices 0 and 1 identify the “normal” and “anomalous” gaits, respectively, and σ(·) is the sigmoid function. According to Eq. (8), s represents the scores associated with each of the two classes for a given input gait. To obtain a probability distribution of the class assignment for a given input, a softmax operation is performed as shown in Eq. (9):
PNG
media_image2.png
137
653
media_image2.png
Greyscale
…The result of Eq. (9) is a probability distribution and, hence, X belongs to the class with the highest probability, i.e., either “0” meaning “normal gait”, or “1” meaning “anomalous gait” [Bonetto pages 6-7 CNN Classifier]; Using the result output by the trained encoder (wherein encoder-decoder is trained based on loss function utilizing distance between decoder output sequence (i.e., predicted encoding) and input sequence (i.e., initial encoding) (see equation 7
PNG
media_image3.png
153
631
media_image3.png
Greyscale
[page 6])), the model further processes the outputted data via CNN layers to obtain values of a final probability distribution (equation 9), wherein it is determined that the gait sequence at issue is valid in response to p (X ∈ anomalous gait) (i.e., first loss function value) being less than P (X ∈ normal gait) (i.e., in a binary distribution, p (X ∈ anomalous gait) being < 0.5)) determining that the human body movement corresponding to the user movement dataset is valid in response to the first loss function value being lower than the preset threshold; and determining the human body movement corresponding to the user movement dataset is invalid in response to the first loss function value being not lower than the preset threshold (“The result of Eq. (9) is a probability distribution and, hence, X belongs to the class with the highest probability, i.e., either “0” meaning “normal gait”, or “1” meaning “anomalous gait” [Bonetto page 7 Classifier]; also see Fig. 8 – “CNN-based binary classifier Architecture” including outputted Softmax probability distribution P[Normal], 1 – P[Normal] [Bonetto page 7]; As explained above, p (X ∈ anomalous gait) (or 1 – P[Normal]) being < 0.5 means that the gait sequence at issue is normal (i.e., valid), and p (X ∈ anomalous gait) being > 0.5 means that the gait sequence at issue is anomalous)
However, Bonetto does not expressly teach the encoder [being] configured to map the first initial encoding to an intermediate variable that compris[es] a compressed eigenvector.
In the same field of endeavor, Wang teaches a means of utilizing an autoencoder architecture to process sensor data for human activity recognition (HAR) tasks (“This paper provides an approach for recognizing human activities with wearable sensors. The continuous autoencoder (CAE) as a novel stochastic neural network model is proposed which improves the ability of model continuous data” [Wang Abstract]) wherein and an encoder is configured to map the first initial encoding to an intermediate variable that compris[es] a compressed eigenvector (“The original signals are acquired by accelerometer, gyroscope and magnetometer which have 125 ˆ 45 dimension of data in each 5 s window. Because the original signals do not have easily-detected features, TFFE method is proposed to extract features from the original sensor data…After feature extraction, the dimension of each data segment is 630 (= 14 ˆ 45). In this paper, PCA [27] method is adopted to reduce the dimension of features. The essence of PCA is to calculate the optimal linear combinations of features by linear transformation. The results of PCA represent the highest variance in the feature subspace. The eigenvalues and contribution rate of covariance matrix are shown in Figure 6. It can be seen that after being sorted in descending order, the contribution rate of the largest three eigenvalues accounts for more than 98% of total contribution rate. These eigenvalues can be used to form the transformation matrix. After PCA feature reduction, the dimension of each signal segment is reduced from 630 (=14 ˆ 45) to 42 (=14 ˆ 3)” [Wang pages 8-9 Feature Extraction and Feature Reduction]; Following a procedure of feature extraction, PCA may be utilized as a processing step for further feature compression).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated an encoder configured to map the first initial encoding to an intermediate variable that compris[es] a compressed eigenvector as taught by Wang into Bonetto because they are both directed towards autoencoder architectures to process sensor data for human activity recognition (HAR) tasks. Given the known applicability of dimensionality reduction techniques, such as principal component analysis (PCA), in the field of human activity recognition (HAR) (see, e.g., Haresamudram et al., “On the Role of Features in Human Activity Recognition” [page 79 Feature Extraction in HAR]), a person of ordinary skill in the art would recognize the value of incorporating the teachings of Wang into the RNN-Seq2Seq model of Bonetto to enable further means of dimensionality reduction [Wang page 9 Feature Reduction] especially for high-dimensionality data, thereby reducing computational requirements for the subsequent classification architecture.
Regarding claim 2, the combination of Bonetto and Wang teaches the limitations of parent claim 1, and Bonetto further teaches training the detection model by a training dataset, the training dataset comprising a plurality of n-dimensional movement data items corresponding to the human body movement (“To improve the quality of this information, and hence that of the reconstructed sequence Xˆ, a training phase is needed, as we discuss next” [Bonetto page 6 Shallow Decoder]);
wherein the training comprises: extracting a feature text based on the training dataset, wherein the feature text comprises a plurality of input sequences, and each of the plurality of the input sequences comprises at least a direction, a distance and a step count; ([Bonetto page 3 Inertial Data Acquisition and Synchronization] and [Bonetto page 4 Gait Cycle Extraction] as detailed in claim 1 above)
transforming the feature text into a second initial encoding according to a preset coding rule; and ([Bonetto page 4 Detrending and Normalization] as detailed in claim 1 above)
in at least one iterative loop, inputting the second initial encoding into the auto- encoder neural network-based detection model to obtain a second predicted encoding, and training the detection model by using a minimization of a second loss function value of the second initial encoding and the predicted encoding as an objective function, until the loss function value is lower than a preset threshold (“To obtain encodings that are suitable for the subsequent classification task, the RNN-Seq2Seq module is only trained on “normal” walking cycles. This is because by learning to only reproduce normal gaits, when anomalous ones are given as input, the module should be unable to determine a correct embedding, and this is detected with high probability by a subsequent classifier. The objective function to minimize during the training phase is the Mean Squared Distance (MSD). Indeed, by minimizing the MSD, one maximizes the match between the input (X) and the output (Xˆ ) sequences. This ensures that the embedding of the input gait cycles obtained by the RNN encoder is a good higher order representation of the original timeseries…To update the RNN-Seq2Seq weights and biases, we use a Stochastic Gradient Descent (SGD) algorithm with exponentially decaying learning rate…This strategy is intended to perform a fast minimization of the objective function at the beginning and to slow it down at later iterations making these last updates more accurate. Formally, the cost function to be minimized using a gradient descent strategy is
PNG
media_image4.png
160
546
media_image4.png
Greyscale
” [Bonetto page 6 Training]).
Regarding claim 3, the combination of Bonetto and Wang teaches the limitations of parent claim 1, and Bonetto further teaches wherein the first loss function value is a cross entropy of the first initial encoding and the first predicted encoding (“To train the classifier, we used both “normal” and “anomalous” gaits that have been labeled at acquisition time. The cross-entropy loss applied to the softmax output was selected as the function to minimize during training” [Bonetto page 7 CNN Classifier]).
Regarding claim 8, the combination of Bonetto and Wang teaches the limitations of parent claim 2, and Bonetto further teaches collecting movement data items at a preset frequency by using at least one sensor in a mobile device to construct the user movement dataset or the training dataset (“As with previous work on automatic gait analysis, the proposed system requires the acquisition of accurate motion data for the subject that is to be monitored… As a relatively cheap and convenient way to acquire this data, we opted for a smartphone application, since camera and inertial sensors are already integrated into the device” [Bonetto pages 2-3 Data Processing]).
Regarding claim 9, the combination of Bonetto and Wang teaches the limitations of parent claim 2, and Bonetto further teaches collecting movement data items by invoking at least one of a GPS positioning service, an application program, a gyroscope of a smart device, or a combination thereof, to construct one of the training dataset and the user movement dataset (“Motion is extracted using accelerometric and gyroscopic measurements from a smartphone device that moves integrally with the analyzed subject” [Bonetto page 2 Contributions of the paper]).
Regarding claim 10, the combination of Bonetto and Wang teaches the limitations of parent claim 1, and Bonetto further teaches dividing the user movement dataset into a plurality of subsets, such that a change rate in at least one of preset dimensions of the plurality of n-dimensional movement data items in each of the plurality of subsets does not exceed a preset value, and determining for each of the plurality of subsets whether a human body movement of the corresponding subset is valid (“Inertial samples are labelled by a timestamp that is relative to the smartphone’s system clock. Although the sampling frequency is high (typically around 100/200 samples/s, depending on the device), the time interval between consecutive samples is not constant. This is consistent with the findings of [35]. To cope with this, interpolation and resampling are performed prior to data analysis to convert the signals into the common sampling frequency of 200 Hz, as done in prior work [36]. Moreover, the power spectral density of the accelerometer signals show that sensor samples are affected by a significant amount of noise due to the irregularities of motion and to the sensitivity of the sensing platform. This noise is largely removed through a low pass filter with a cut-off frequency of 40 Hz” [Bonetto page 3 Inertial Data Acquisition and Synchronization]).
Regarding claim 12, the combination of Bonetto and Wang teaches the limitations of parent claim 1, and Bonetto further teaches wherein the auto-encoder neural network based detection model is deployed on a device terminal with edge computing capability, and the method is executed by the device terminal (“Inertial and video data are collected by means of a custom made Android application called “Activity Logger and Video” (ALV). ALV was tested on an Asus Zenfone 2 featuring a 2.3 GHz quad-core Intel CPU, 4 GB of RAM and an Android 5 “Lollipop” operating system…The application is used to set the acquisition parameters, collect information about the user (age, height, gender, etc.), collect and save data into the smartphone non-volatile memory and, optionally, send them to a File Transfer Protocol (FTP) server” [Bonetto page 7 Activity Logger Video Application]; “Moreover, it is worth noting that, once the initial training of the RNN-Seq2Seq model and the CNN-based classifier has been completed, the resulting model can be run on off-the-shelf smartphones as the Asus Zenfone 2 used in this work” [Bonetto page 9 Results]).
Regarding claim 13, it is a product claim that corresponds to the method of claim 1, which is already taught by the combination of Bonetto and Wang as detailed above. Bonetto further teaches A non-transitory computer-readable medium having stored content, wherein the stored content causes a computing system to perform automated operations, comprising: the claimed functions ([Bonetto page 7 Activity Logger Video Application] and [Bonetto page 9 Results]) Consequently, claim 13 is rejected for the same reasons as claim 1.
Regarding claim 14, it is an apparatus claim that corresponds to the method of claim 1, which is already taught by the combination of Bonetto and Wang as detailed above. Bonetto further teaches A system, comprising: one or more processors; a wireless communication module configured to obtain a user movement dataset; and at least one memory having stored instructions that, when executed by at least one of the one or more processors, cause the system to perform automated operations comprising: the claimed functions ([Bonetto page 7 Activity Logger Video Application] and [Bonetto page 9 Results]). Consequently, claim 14 is rejected for the same reasons as claim 1.
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Bonetto and Wang, as applied to claim 2 above, further in view of Saeed et al. (“Synthesizing and Reconstructing Missing Sensory Modalities in Behavioral Context Recognition”, published 6 Sep 2018), hereinafter Saeed.
Regarding claim 4, the combination of Bonetto and Wang teaches the limitations of parent claim 2.
However, the combination does not expressly teach wherein the second loss function value is a cross entropy of the second initial encoding and the second predicted encoding, as Bonetto utilizes mean square distance (MSD) for reconstruction loss rather than binary cross entropy [Bonetto page 6 Training]
In the same field of endeavor, Saeed teaches a means of utilizing an autoencoder architecture to process sensor data for human activity recognition (HAR) tasks (“Detection of human activities along with the associated context is of key importance for various application areas, including assisted living and well-being. To predict a user’s context in the daily-life situation a system needs to learn from multimodal data that are often imbalanced, and noisy with missing values. The model is likely to encounter missing sensors in real-life conditions as well (such as a user not wearing a smartwatch) and it fails to infer the context if any of the modalities used for training are missing. In this paper, we propose a method based on an adversarial autoencoder for handling missing sensory features and synthesizing realistic samples” [Saeed Abstract]) wherein the second loss function value (i.e., reconstruction loss) is a cross entropy of the second initial encoding and the second predicted encoding (“The classical autoencoder can be extended in several ways (see for a review [11]). For handling missing input data, a compelling strategy is to train an autoencoder with artificially corrupted input x˜, which acts as an implicit regularization… Formally, the DAE is trained with stochastic gradient descent to optimize the following objective function:
PNG
media_image5.png
30
257
media_image5.png
Greyscale
where L represents a loss function like squared error or binary cross entropy” [Saeed pages 5-6 Autoencoder]; “We employ binary cross-entropy (see Equation (7)) for reconstruction loss rather than MSE as it led to consistently better results in earlier exploration” [Saeed page 8 Model Architecture and Training]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated wherein the second loss function value (i.e., reconstruction loss) is a cross entropy of the second initial encoding and the second predicted encoding as taught by Saeed into the combination because both Bonetto and Saeed are directed towards utilizing an autoencoder architecture to process sensor data for human activity recognition (HAR) tasks. Given that a person of ordinary skill in the art would already recognize binary cross entropy loss to be compatible with a binary classification task as disclosed in Bonetto, incorporating the teachings of Saeed, which extend upon typical autoencoder architecture training, into the RNN-Seq2Seq encoder-decoder architecture of Bonetto would thereby enable said architecture to be more adaptable to sensor reading in real-time conditions (i.e., dealing with missing values) [Saeed Abstract].
Claims 5-6 are rejected under 35 U.S.C. 103 as being unpatentable over Bonetto and Wang, as applied to claim 1 above, further in view of Kaghyan et al. (“Accelerometer and GPS Sensor Combination Based System for Human Activity Recognition”, available IEEE Xplor 16 Jan 2014), hereinafter Kaghyan.
Regarding claim 5, the combination of Bonetto and Wang teaches the limitations of parent claim 1 and wherein each of the plurality of n-dimensional movement data items comprises accumulated steps information ([Bonetto page 3 Inertial Data Acquisition and Synchronization] and [Bonetto page 4 Gait Cycle Extraction] as detailed in claim 1 above).
However, the combination does not expressly teach wherein each of the plurality of n-dimensional movement data items comprises at least longitude information and latitude information.
In the same field of endeavor, Kaghyan teaches a means of utilizing machine learning models for processing sensor data for human activity recognition (HAR) tasks (“Current paper introduces an approach which allows recognizing activity, performed by human, using smartphone acceleration and positioning sensors. We introduce an approach that retrieves signal data and stores it SQLite portable mobile database. It uses asynchronous model of signal retrieving and storing procedures. After the signals were collected we applied noise reduction, time and frequency domain feature extraction processes for stored information and acquired high-dimensional feature patterns. These patterns were later transferred on remote server instead of raw signals. The classification stage was based on “learning with teacher” method. Incoming signal sequences were collected from sensors of mobile device and were analyzed using support vector machines (SVM) learning method” [Kaghyan Abstract]) wherein each of the plurality of n-dimensional movement data items comprises at least longitude information and latitude information (“So, it is obvious that the data to recognize human’s movement activity from the physical hardware sensors, and the combination of the accelerometer, the compass sensors and GPS are the most commonly used nowadays. And our approach uses accelerometer and GPS sensors for signal retrieving and further processing” [Kaghyan page 1 Introduction]; “For GPS we used following features: mean and minimal frame times, mean latitude and longitude, mean and minimal distances between two sequential coordinates, which calculates similar way (except geographical coordinates which calculated using formula of Euclidean distance” [Kaghyan page 5 Noise Reduction and Feature Extraction Concepts])
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated wherein each of the plurality of n-dimensional movement data items comprises at least longitude information and latitude information as taught by Kaghyan into Wang because they are both directed towards utilizing machine learning models for processing sensor data for human activity recognition (HAR) tasks. Given the known recognition of location sensors, such as GPS, being useful for activity recognition tasks and also commonly included in mobile devices (“A large class of activity recognition method exploits sensors embedded in mobile devices, which potentially overcomes deployment constraints due to broad availability. In terms of wide usability, smartphones that are equipped with various sensors (audio, video or motion detection) can be considered as perfect tool for short-term physical activity recognition. Broader list of useful mobile device sensors includes imaging camera, microphones, accelerometers, gyros and compasses, ambient light detectors, proximity sensors, location sensors (GPS&WLAN&network), WLAN and other wireless network signal readings” [Kaghyan page 2 Related Works]), a person of ordinary skill in the art would recognize the value of incorporating the teachings of Kaghyan to combine both location data and movement data collected from a mobile device to further boost predictive accuracy of the RNN-Seq2Seq model.
Regarding claim 6, the combination of Bonetto, Wang, and Kaghyan teaches the limitations of parent claim 5, and Bonetto further teaches wherein each of the plurality of n-dimensional movement data items further comprises at least one of time information, instantaneous movement speed, and altitude information (“Inertial samples are labelled by a timestamp that is relative to the smartphone’s system clock” [Bonetto page 3 Inertial Data Acquisition and Synchronization]).
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Bonetto, Wang, and Kaghyan, as applied to claim 1 above, further in view of Muse et al. (Pub. No. US 20210241869 A1, “Systems and Methods for Sharing Recorded Medical Data with Authorized Users”, published 08/05/2021), hereinafter Muse.
Regarding claim 7, the combination of Bonetto, Wang, and Kaghyan teaches the limitations of parent claim 5.
However, the combination does not expressly teach wherein each of the plurality of n-dimensional movement data items further comprises at least one of a pulse rate, a body temperature, a blood oxygen value, and a blood pressure value.
In the same field of endeavor, Muse teaches a means of utilizing machine learning models for processing sensor data for human activity recognition (HAR) tasks ([Muse ¶ 0016]) wherein each of the plurality of n-dimensional movement data items further comprises at least one of a pulse rate, a body temperature, a blood oxygen value, and a blood pressure value (“In some embodiments, the patient computing entity 102 authorizes the data computing entity 106 to obtain data from sensors or applications on the patient computing entity 102 (e.g., by transmitting data from the patient computing entity 102 to the data computing entity 106 for storage within a database). In some embodiments, the patient computing entity 102 provides consent to access data obtained from an accelerometer or a pedometer of the personal device, for example, based at least in part on received user input indicative of the provided consent. In some embodiments, the patient computing entity 102 provides consent to access data of body temperature of the first user from the personal device (and/or ambient temperature), for example, based at least in part on received user input indicative of the provided consent. Alternatively, the patient computing entity 102 provides consent to access data of heart rate or blood pressure of the first user from the personal device, for example, based at least in part on received user input indicative of the provided consent. The patient computing entity 102 provides consent to access data of one or more types of activity of the first user such as walking, running, biking, and/or the like, for example, based at least in part on received user input indicative of the provided consent. In some embodiments, the patient computing entity 102 provides consent to access data obtained from one or more additional devices associated with the first user, for example, based at least in part on received user input indicative of the provided consent. For example, the one or more personal devices can include and/or may be associated with a smart watch, a fitness tracker, smart clothing, or any other devices connected to the personal device and capable of collecting data of the first user” [Muse ¶ 0051]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated wherein each of the plurality of n-dimensional movement data items further comprises at least one of a pulse rate, a body temperature, a blood oxygen value, and a blood pressure value as taught by Muse into the combination because both Bonetto and Muse are directed towards utilizing machine learning models for processing sensor data for human activity recognition (HAR) tasks. Incorporating the teachings of Muse would enable expanding applicability of the RNN Seq2Seq model to medical applications (e.g., automating recognizing emergency medical events and providing relevant data to medical personnel) ([Muse ¶ 0018]).
Claim 11 is rejected under 35 U.S.C. 103 as being unpatentable over Bonetto and Wang, as applied to claim 1 above, further in view of Muse (Pub. No. US 20210241869 A1, “Systems and Methods for Sharing Recorded Medical Data with Authorized Users”, published 08/05/2021).
Regarding claim 11, the combination of Bonetto and Wang teaches the limitations of parent claim 1.
However, the combination does not expressly teach tagging movement data items collected during a pause period in response to a user's indication to pause a movement, and removing the tagged movement data items when obtaining the user movement dataset.
In the same field of endeavor, Muse teaches a means of utilizing machine learning models for processing sensor data for human activity recognition (HAR) tasks ([Muse ¶ 0016]) comprising tagging movement data items collected during a pause period in response to a user's indication to pause a movement, and removing the tagged movement data items when obtaining the user movement dataset (“In some embodiments, the patient computing device 102 determines one or more conditions for which satisfaction would override, or pause, the data computing entity 106 from recording data (or deleting data) of the first user. In some embodiments, the patient computing entity determines that under certain circumstances, such as during instances in which the first user is participating in particular activities, e.g., exercising, intermittent fasting for a medical test, the user's monitored medical data is expected to violate one or more of the ranges. Such embodiments would result in inaccurate recorded data, which would further mislead the medical practitioners in evaluating first user's medical conditions. For example, the patient computing entity 102 can determine when the first user enters a gym facility to workout, the data computing entity 106 pauses recording data. Additionally, in some embodiments, the patient computing entity 102 determines one or more actions, e.g., driving, or one or more locations, e.g., a gym, or proximity to specific people or places, e.g. to a medical laboratory, that would pause the recording data” [Muse ¶ 0054]; “In some embodiments, the patient computing entity 102 further determines one or more conditions occurrence of which triggers recording data. Once the one or more conditions are satisfied, the data computing entity 106 starts recording the data. As a non-limiting example, the data computing entity 106 starts recording data of the first user when the first user is traveling 200 miles from home and they have not used their personal device for the last 90 minutes, or when they were driving and an EMT has been within 20 feet for more than 5 minutes. In some embodiments, the patient computing entity 102 determines one or more thresholds and target values that would indicate anomalies specific to the first user. The one or more thresholds can be determined subjectively and for a specific first user. For example, when historically the first user's blood pressure is low, a blood pressure above 95/88 would be an anomaly for that user. In some embodiments, the first user can optionally consent to access and record additional types of data after initial set up. Alternatively, the first user can retract a consent to access and record one or more types of data after initial set up. The first user can further add or remove personal, medical or environmental data at any time after the initial set up” [Muse ¶ 0055]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated tagging movement data items collected during a pause period in response to a user's indication to pause a movement, and removing the tagged movement data items when obtaining the user movement dataset as taught by Muse into the combination because both Bonetto and Muse are directed towards utilizing machine learning models for processing sensor data for human activity recognition (HAR) tasks. Incorporating the teachings of Muse would enable avoidance of including data from anomaly events that would likely throw off predictive capabilities of the activity recognition model and, e.g., possibly mislead practitioners in a medical setting [Muse ¶ 0054].
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Ghojogh et al., (“Factor Analysis, Probabilistic Principal Component Analysis, Variational Inference, and Variational Autoencoder: Tutorial and Survey”) discloses a survey of factor analysis, probabilistic Principal Component Analysis (PCA), variational inference, and Variational Autoencoder (VAE) techniques, as well as the commonalities between them.
Chen et al., (“Sensor-Based Activity Recognition”) discloses a review of major approaches and methods associated with sensor-based activity monitoring, modeling, and recognition.
Haresamudram et al., (“On the Role of Features in Human Activity Recognition”) discloses a review of the role feature representations play in HAR using wearables.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to VIJAY M BALAKRISHNAN whose telephone number is (571) 272-0455. The examiner can normally be reached 10am-5pm EST Mon-Thurs.
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, JENNIFER WELCH can be reached on (571) 272-7212. 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.
/V.M.B./
Examiner, Art Unit 2143
/JENNIFER N WELCH/Supervisory Patent Examiner, Art Unit 2143