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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 14 January 2026 has been entered.
Claim Rejections - 35 USC § 103
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim(s) 1, 7-8, 10, 16-17, and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Cohen et al. (US Pub. No. 2017/0162197, published June 2017, hereinafter “Cohen”) in view of Eleftheriou et al (US Patent No. 11410682, filed July 2019, hereinafter “Eleftheriou”) and further in view of Milton et al. (US Pub. No. 2020/0017117, published Jan. 2020, hereinafter “Milton”).
Regarding claim 1, Cohen teaches a computer-implemented method comprising:
receiving a global intermediate intervention score threshold and a global baseline sensory feature data object for the cognitive condition based on data received from [[the]] a plurality of vehicles associated with a population of users with the cognitive condition, wherein the vehicle is one of the plurality of vehicles and comprises a plurality of sensors configured to transmit the data to the one or more processors (Cohen, [0035] – “In other words, user sensors 160 may measure physiological and other characteristics of the user. A given user sensor 160 may include, without limitation, a camera directed toward the user (e.g., to measure the user's pupil, monitor the user's head movements, etc.), microphone (e.g., to measure the user's speech), galvanic skin response receptors (e.g., mounted on a steering wheel or other input), and/or other sensor that can measure a physiological characteristic of the user.”, [0060] – “FIG. 3 depicts a conversation management system 108 configured to assess a human user's cognitive state and/or a situational state and adjust natural language conversations from a computer based on the cognitive and/or situational state, according to an implementation of the invention. For example, conversation management system 108 may adjust the extent to which a user may interact with NLP system 106 based on a user's cognitive state and/or a situational state associated with the user.”, [0007] – “The user's state may also be inferred from vehicular sensor information (e.g., vehicle is veering in and out of road travel lanes) and from historic models of user behavior or physiological states”, [0075] – “In some instances, the sensor information from user sensors 160 may be stored in a database for baseline comparison. For instance, user assessment engine 320 may model physiological characteristics (as determined from the sensor information) of a user for baseline purposes… User assessment engine 320 may obtain the baseline physiological information for the user and compare current physiological characteristics with the baseline physiological characteristics. For example, user assessment engine 320 may model a user's gaze (e.g., position of pupils) over a typical driving session to determine what a “normal” or average gaze is like for a user while driving. Such baseline gaze may be used to compare to a user's current gaze to determine whether the current pupil position deviates beyond an average range, causing user assessment engine 320 to determine that the user is distracted. In another example, a user's speech over time may be stored in the database to infer normal speech of the user, which may be compared to current speech of the user, which may be impaired.”, and in [0082] – “In an implementation, conversation adjuster 330 may use a combination of one or more user cognitive states and one or more situational states to determine whether and what action to take. For instance, the adjustment rules may specify that a given action should be taken if the user is sleepy and a road hazard exists and take another action if the user is sleepy and bad weather conditions are present.” - teaches receiving a global intermediate intervention score threshold (adjustment rules are intervention score thresholds, conversation adjuster 330 may use combination of cognitive state and situation state to determine what action to take based on adjustment rules which specify thresholds for an action to be taken according to assessments as in [0073]) and a global baseline sensory feature data object (baseline physiological information for baseline purpose) for the cognitive condition (user assessment engine, which assesses user cognitive condition, may obtain baseline physiological information to assess sensor features gathered from user sensors 160) based on data received from a plurality of vehicles associated with a population of users with the cognitive condition (user cognitive state may be inferred from historic models of user behavior or physiological states), wherein the vehicle is one of the plurality of vehicles (current user operating vehicle) and comprises a plurality of sensors configured to transmit data to the one or more processors (vehicle comprises plurality of sensors configured to transmit data as in Cohen at [0033]))
identifying a real-time sensory timeseries data object comprising aggregated sensor measurements captured by the plurality of sensors during a current prediction window (Cohen, [0063] – “Situation assessment engine 310 may receive third party information from data providers 120 and/or situation sensor information from environment sensors 150. In some instances, situation assessment engine 310 may determine a situational state based on individual (single) pieces of information (e.g., traffic data from data providers 120 or throttle data from a throttle sensor). In other instances, situation assessment engine 310 may fuse various third party information and/or various sensor information to determine an assessment of a situational state.” and in [0064] – “The individual or fused information may be continuously monitored to update the state of the situation (e.g., continuously update the situation score). The term “continuously” is used herein to mean periodically perform an action until a terminating signal or event (e.g., a vehicle being placed in park or turned off) has occurred.” – teaches identifying a real-time (continuous monitoring of information) sensory timeseries data object comprising aggregated sensor measurements (situation assessment engine fuses various sensor information) captured by the plurality of sensors during a current prediction window (fused information is continuously monitored by periodically performing an action until terminating signal, thus capturing information during a current prediction window))
generating, using the real-time risk scoring machine learning model, and based at least in part on the real-time sensory timeseries data object and a global baseline sensory feature data object for the cognitive condition, a real-time risk score associated with the current prediction window that indicates a severity level of the cognitive condition with respect to the monitored user (Cohen, [0073] – “In an implementation, user assessment engine 320 may use user sensor information from user sensors 160 to assess a user's cognitive state (e.g., the user's emotional state, whether the user is sleepy, intoxicated, or otherwise distracted). For example, user assessment engine 320 may use sensor information to generate a cognitive state score, in a manner similar to that described above with respect to situational state scores. Adjustment rules may include rules for scoring cognitive state for given types of sensor information, which may be informed by statistical analysis of risk factors of user cognitive states. For instance, sleepiness may present one level of risk while being distracted may present another level of risk. As such, different cognitive state scores may be assessed based on different levels of risk a given cognitive state presents. Furthermore, and also similar to that described above with respect to situational state scores, different cognitive states, when combined, may result in greater risk, which may be reflected in an overall cognitive state score that represents the greater risk. For example, a user who is sleepy and distracted may have a greater risk of being involved in a car accident than a user who is only sleepy or only distracted. Each sensor may generate a probabilistic assessment of a feature of the user's cognitive state (sleepiness, distraction, emotion (e.g. angry)), looking at dashboard, etc.” and in [0066] – “These parameters may, in turn, be defined based on risk tables associated with different types and combinations of activities and risks, which may be informed by statistical data (e.g., accident rates, fatality rates, etc.). In some instances, machine learning may be used to determine optimal combinations of weights, types of situation assessments, confidence scores or probability assignments in the various situation assessments.” – teaches generating, using a real-time risk scoring machine learning model (machine learning used to determine confidence scores or probability assignments in various situation assessments, sensors generate probabilistic assessment of user cognitive state for risk score) and based at least in part on the real-time sensory timeseries data object and a global baseline sensory feature data object for the cognitive condition (cognitive state score is based on user assessment engine 320 using user sensors 160, user sensor information may be fused and provided continuously as in Cohen at [0063-0064] and has baseline physiological information for baseline purposes as in [0075]), a real-time risk score (cognitive state score determines score based on cognitive state associated with risk) associated with the current prediction window (continuous assessment) that indicates a severity level of the cognitive condition with respect to the monitored user (cognitive state score indicates risk associated with cognitive condition of the monitored user, where cognitive state such as sleepy and distracted is at greater risk than cognitive state of sleepy and not distracted)),
wherein (i) the global baseline sensory feature data object comprises a plurality of real-time sensory features captured by [[a]] the plurality of vehicles associated with the population of users with the cognitive condition (Cohen, [0007] – “For instance, the sensor information may be used to assess a driver's voice through voice analysis, perspiration through galvanic skin response electrodes, assess alertness or attention via pupil analysis based on images from interior cameras, and/or other user sensor information used to assess a driver's state. The user's state may also be inferred from vehicular sensor information (e.g., vehicle is veering in and out of road travel lanes) and from historic models of user behavior or physiological states.”, [0025] – “Data providers 120 may provide data relevant to situational states. Such data may include, without limitation, traffic data (e.g., congestion information, road construction, etc.), weather data (e.g., forecasts for a given location, current weather conditions, etc.), vehicle-to-vehicle data sources (e.g., vehicle-to-vehicle communication information), vehicle-to-server data sources (e.g., vehicle sensor information communicated from vehicles to a server), and/or other information that may be used to assess a situation.”, [0033] – “In an implementation, environment sensors 150 may measure or otherwise gather (e.g., sense) information relating to a situation associated with the user. In other words, environment sensors 150 may sense the (intra-vehicle and/or extra-vehicle) surroundings of a user but not direct measurements of the user himself (which the user sensors 160 sense). A given situation sensor 150 may include, without limitation, a vehicle state sensor (e.g., steering wheel sensor, throttle sensor, brake sensor, seatbelt sensor, etc.), a Global Positioning System (“GPS”) device that provides location information, a camera directed to outside the vehicle (e.g., a lane keeping/warning system), a radar directed to outside the vehicle (e.g., a collision mitigation system, automatic radar, etc.), a thermometer (e.g., to sense freezing conditions), a rain sensor, and/or other sensor that can measure the environment associated with a user (e.g., surroundings of a vehicle operated by the user).”, [0063] – “Situation assessment engine 310 may receive third party information from data providers 120 and/or situation sensor information from environment sensors 150. In some instances, situation assessment engine 310 may determine a situational state based on individual (single) pieces of information (e.g., traffic data from data providers 120 or throttle data from a throttle sensor). In other instances, situation assessment engine 310 may fuse various third party information and/or various sensor information to determine an assessment of a situational state.” – teaches wherein the global baseline sensory feature data object (sensor information stored in database for baseline comparison or baseline purposes as in [0075]) comprises a plurality of real-time sensory features captured by the plurality of vehicles (environment sensors 150 and user sensors 160) associated with the cognitive condition (user state inferred from historical models of user behavior and psychological states))
generating, based at least in part on the real-time risk score, an intermediate intervention score (Cohen, [0083] – “In an implementation, conversation adjuster 330 may generate an overall score based on a user cognitive assessment score and a situational assessment score. For instance, the adjustment rules may specify a first weight for the user cognitive assessment score and a second weight for the situational assessment score. The weights in this context may assign a level of importance to the user cognitive assessment score and the situational assessment score.” – teaches generating, based at least in part on the real-time risk score (user cognitive assessment score), an intermediate intervention score (overall score based on user cognitive assessment scores and situational assessment scores))
performing, responsive to determining that the intermediate intervention score satisfies the global intermediate intervention score threshold, one or more intermediate intervention operations (Cohen, [0082] – “In an implementation, conversation adjuster 330 may use a combination of one or more user cognitive states and one or more situational states to determine whether and what action to take. For instance, the adjustment rules may specify that a given action should be taken if the user is sleepy and a road hazard exists and take another action if the user is sleepy and bad weather conditions are present.” – teaches performing, response to determining that the intermediate intervention score satisfies the global intermediate intervention score threshold (responsive to adjustment rule being satisfied, adjustment rule determines action based on current states and assessment scores as in Cohen at [0081]), one or more intermediate intervention operations (specifies a given action that should be taken))
Cohen fails to explicitly teach receiving one or more intermediate intervention response features associated with the one or more intermediate intervention operations and generating, using the intermediate risk scoring machine learning model, and based at least in part on the one or more intermediate intervention response features, an intermediate risk score.
However, analogous to the field of the claimed invention, Eleftheriou teaches:
receiving one or more intermediate intervention response features associated with the one or more intermediate intervention operations (Eleftheriou, Page 17, Col. 18 Lines 42-48 – “the system can: access the set of sensors on the wearable device; detect an instance of a distress emotion exhibited by the user; send a signal to the user by vibrating the wearable device; access a stress coaching protocol generated for the user; and prompt the user via the mobile device to begin a breathing coaching activity to help the user regulate the stress emotional state” – teaches receiving one or more intermediate intervention response features (set of sensor detecting regulation of stressed emotional state during coaching protocol) subsequent to performing one or more intermediate intervention operations (coaching protocols)),
generating, using the intermediate risk scoring machine learning model, and based at least in part on the one or more intermediate intervention response features, an intermediate risk score (Eleftheriou, Page 17, Col. 18 Lines 52-58 – “ The system can record an effectiveness value for each of the coaching protocols for the target emotion based on the time elapsed from the start of the coaching protocol to the end of the instance of the target emotion and either promote or remove particular coaching protocols based on the efficacy of each coaching protocol” – teaches generating an intermediate risk score (effectiveness value) based at least in part on the one or more intermediate intervention response features (based on time elapsed from start of coaching protocol to end of instance of target emotion)),
Therefore, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate the intermediate intervention response features and intermediate risk score based on the intermediate intervention response features of Eleftheriou to the user assessment, sensors, and vehicles of Cohen. Doing so would provide metrics for measuring the efficacy and success rate of intervention protocols (Eleftheriou, Pg. 17, Col. 18) for emotions that are considered in evaluations of user cognitive conditions and risks (Cohen, [0073]).
The combination of Cohen and Eleftheriou fails to explicitly teach: generating, by one or more processors and using a hierarchical intervention recommendation machine learning framework that comprises (i) a real-time risk scoring machine learning model, (ii) an intermediate risk scoring machine learning model, and (iii) a risk aggregation machine learning model, an adjusted risk score indicating a likelihood of a cognitive impairment of a monitored user associated with a vehicle and a cognitive condition, generating, using the risk aggregation machine learning model and based at least in part on the real-time risk score and the intermediate risk score, [[an]] the adjusted risk score, generating, by the one or more processors and responsive to the adjusted risk score confirming the real-time risk score, an alternative navigation route of the vehicle which diverts the vehicle to a geographic location associated with a risk score below a threshold, and updating, by the one or more processors, a navigation route of a GPS system of the vehicle based on the alternative navigation route.
However, analogous to the field of the claimed invention, Milton teaches:
generating, by one or more processors and using a hierarchical intervention recommendation machine learning framework that comprises (i) a real-time risk scoring machine learning model, (ii) an intermediate risk scoring machine learning model, and (iii) a risk aggregation machine learning model (Milton, [0043] – “In some embodiments, an agent may perform a training activity and send results of the training activity to a subsequent infrastructure layer as part of a federated learning (FL) architecture. For example, each of a set of computing devices attached to one of a corresponding set of vehicles may be part of a FL population of devices assigned to perform a training operation or other machine-learning operation. Training weights, training results, or other results from one or more agents may be sent to a subsequent layer during a FL round without exposing the data used for training to any other agent on the same layer.” – teaches a federated machine learning framework that comprises multiple models at multiple levels of a hierarchy, as in [0096] – “each selected vehicle may transmit its respective state values (e.g. gradients or perceptron weights) from a training operation to a data center of the local computing layer after encrypting the state values into cryptographic hash values. One or more applications on the local computing layer or on the top-view computing layer may then modify a global model based on the results from each of the selected vehicles and transmit the modified global model to each of the selected vehicles for further training.” – thus teachings a hierarchical federated machine learning framework comprising models at the agent or vehicle level, data centers or intermediate level, and the global or aggregated level), an adjusted risk score indicating a likelihood of a cognitive impairment of a monitored user associated with a vehicle and a cognitive condition (Milton, [0058] – “Some embodiments may associate a risk value with a behavior, operator profile, or vehicle profile.” and in [0101] – “Some embodiments may include training and using one or more of the learning methods described above to predict a future vehicle behavior based on known sensor data from a plurality of vehicles. For example, in some embodiments, a data center in a local computing layer may train and then use an ensemble system comprising a federated machine-learning system having an attention model and a CNN that uses vehicle geolocation data, vehicle velocity data, and LIDAR-detected object data from a plurality of vehicles to predict whether a vehicle having a particular vehicle profile will encounter an accident in a 1 month period.” – teaches generating a risk score indicating a likelihood of cognitive impairment of a monitored user associated with a vehicle and a cognitive condition (risk score associated with vehicle operator behavior and/or profile, thus generating a risk score of a monitored user associated with a vehicle and their cognitive condition)), by:
generating, using the risk aggregation machine learning model and based at least in part on the real-time risk score and the intermediate risk score, [[an]] the adjusted risk score (Milton, [0021] – “These attributes can be combined to form vehicle profiles, operator profiles, or location profiles, each which are further described below and can be used to determine vehicle adjustment values. In some cases, the profiles may be implicit in trained machine-learning model parameters. In addition, these attributes can be combined to provide risk indicators for locations, vehicle types, or operator types, wherein vehicle adjustment values can be pushed to vehicles in response to one or more risk values meeting or exceeding a risk threshold.” and in [0058] – “Alternatively, or in addition, some embodiments may determine or update a risk value based on specific two-part relationships between a vehicle operator and a vehicle. Alternatively, or in addition, some embodiments may determine or update a risk value based on three-part relationships between vehicle operators using a specific vehicle to visit a specific place as determined based on a road network graph portion.” – teaches generating, using the risk aggregation machine learning model and based at least in part on the real-time risk score and the intermediate risk score, the adjusted risk score (risk values may be updated based on two-part or three-part relationships between vehicle operators, vehicles, and/or specific locations, thus generating an adjusted risk based at least in part on various risk scores))
generating, by the one or more processors and responsive to the adjusted risk score confirming the real-time risk score, an alternative navigation route of the vehicle which diverts the vehicle to a geographic location associated with a risk score below a threshold (Milton, [0018] – “In some cases, fleet operators (e.g., in trucking, ride-share platforms, or delivery services) may adjust routing of configuration of fleet over geographic areas responsive to outputs of the trained models.” – teaches generating an alternative navigation route of the primary vehicle (adjust routing configuration) which diverts vehicle to a geographic location associated with a risk score below a threshold, as in [0021] – “In addition, these attributes can be combined to provide risk indicators for locations, vehicle types, or operator types, wherein vehicle adjustment values can be pushed to vehicles in response to one or more risk values meeting or exceeding a risk threshold.” – teaches risk indicators for locations and adjustment values confirming risk score thresholds); and
updating, by the one or more processors, a navigation route of a GPS system of the vehicle based on the alternative navigation route (Milton, [0028] – “Some embodiments may include, as on-board vehicle sensors, satellite navigation sensors such as global positioning system (GPS) sensors, Global Navigation Satellite System (GLONASS) sensors, Galileo sensors, etc. to provide one or more geolocations.” and in [0018] – “In some cases, fleet operators (e.g., in trucking, ride-share platforms, or delivery services) may adjust routing of configuration of fleet over geographic areas responsive to outputs of the trained models” – teaches updating a navigation route of a GPS system of the primary vehicle based on the alternative navigation route (GPS provides geolocation, system updates route configurations responsive to trained models)).
Therefore, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate the hierarchical federated learning framework, aggregation, plurality of vehicles, adjusted risk scores, and vehicle sensor data of Milton to the cognitive condition assessments, vehicles, interventions, and real-time and intermediate risk scores of Cohen and Eleftheriou. Doing so would provide a federated machine-learning architecture that supports active learning to infer things about vehicles, drivers, and places based on relatively high-bandwidth on-board (Milton, [0017] and improve accuracy, computation speed, and data privacy for autonomous driving systems (Milton, [0043]).
Claims 10 and 19 incorporate substantively all the limitations of claim 1 in a system and non-transitory computer-readable storage media, and are rejected on similar grounds as above. Cohen teaches the processors and storage media of these claims at [0021] – “Computer system 110 may include one or more processors 112 (also interchangeably referred to herein as processors 112, processor(s) 112, or processor 112 for convenience), one or more storage devices 114, and/or other components.”
Regarding claim 7, the combination of Cohen, Eleftheriou, and Milton teaches the computer-implemented method of claim 1, wherein:
a server computing entity is configured to provide global intermediate intervention score thresholds to the vehicle according to a predefined time interval (Eleftheriou, Page 15, Col. 13 Lines 41-49 – “The system can assign confidence scores for the remaining instances of the first target emotion to build the emotion model only with instances above the minimum confidence score, such that the emotion model only contains a minimum amount of data to accurately predict instances of the target emotion, which can then be stored locally on the wearable device (and/or stored remotely on the mobile device or the remote computer system).” – teaches a server computing entity (remote computer system) providing the global intermediate intervention score threshold (minimum confidence score) to the particular computing entity (remote computer system). In addition to the previously cited passages, Eleftheriou further teaches providing the global intermediate intervention score thresholds according to a predefined time interval, as in Page 15, Col. 13 Lines 58- 63 – “For example, the wearable device can send packets of data to the mobile device once per minute following the detection of an instance of a happy emotion and—upon detection of an instance of an angry emotion—begin to send packets of data to the mobile device once every five seconds to better monitor and manage the negative emotion in the user.” – teaches providing global intermediate intervention score thresholds (minimum confidence score) according to a predefined time interval (sends data once per minute upon following detection)).
The combination of Cohen and Eleftheriou fails to teach providing a global intermediate intervention score threshold to a primary vehicle.
However, analogous to the field of the claimed invention, Milton teaches:
a server computing entity configured to provide a global intermediate intervention score to a vehicle (Milton, [0130] – “In some embodiments, an application executing on the top-view computing layer may apply one or more control-system adjustment values based on sensor data from a first and second vehicle to modify the operations of a third vehicle, wherein applying the control-system adjustment value may be done through an actual adjustment of vehicle operations or through a simulation. In response, the top-view computing layer may determine whether applying the one or more control-system adjustment values to the third vehicle would reduce a risk value below a risk threshold value, or whether the proposed control-system adjustment value violates any other adjustment value limitations programmed into one or more media of the top-view computing layer.” – teaches a server computing entity (top-view layer) providing a global intermediate intervention threshold (adjustment values applied to risk values that would result in risk below a threshold risk) to a primary vehicle)
Therefore, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate the risk scores and vehicles of Milton to the intervention scores thresholds of Cohen and Eleftheriou in order to provide the global intermediate intervention score threshold to a primary vehicle. Doing so would improve accuracy, computation speed, and data privacy for autonomous driving systems (Milton, [0043]).
Claim 16 is similar to claim 7, hence similarly rejected.
Regarding claim 8, the combination of Cohen, Eleftheriou, and Milton teaches the computer-implemented method of claim 7, wherein the server computing entity is configured to generate the global intermediate intervention score threshold based at least in part on:
(i) a statistical distribution measure of sensory data associated with a plurality of client computing entities associated with the server computing entity (Cohen, [0023] – “Computer system 110 may be configured as a device remote from and communicating with (via network 102) user device 140. For example, computer system 110 may include a networked server (or other type) of computing device that is programmed to assist user device 140 adjust conversations with users by analyzing, in whole or in part, a user's cognitive state and/or situational state. In this sense, computer system 110 may receive some or all available sensor information via network 102 and assess the user's cognitive and/or situational state.”, [0066] – “As such, the adjustment rules may specify weights for different combinations of third party or sensor information. Such weights and corresponding situation scores may be predefined in the adjustment rules, which may be informed by various predefined parameters. These parameters may, in turn, be defined based on risk tables associated with different types and combinations of activities and risks, which may be informed by statistical data (e.g., accident rates, fatality rates, etc.). In some instances, machine learning may be used to determine optimal combinations of weights, types of situation assessments, confidence scores or probability assignments in the various situation assessments.” and in [0073] – “Each sensor may generate a probabilistic assessment of a feature of the user's cognitive state (sleepiness, distraction, emotion (e.g. angry)), looking at dashboard, etc.” – teaches a statistical distribution (probabilistic assessment) measure of sensory data (each sensor) associated with a plurality of client computing entities (sensors associated with user) associated with the server computing entity (system configured as device remote from and communicating with a networked server that analyzes some or all available sensor information))
and (ii) historical sensory data associated with the particular client computing entity (Cohen, [0007] – “The user's state may also be inferred from vehicular sensor information (e.g., vehicle is veering in and out of road travel lanes) and from historic models of user behavior or physiological states.” and in [0075] – “In some instances, the sensor information from user sensors 160 may be stored in a database for baseline comparison... User assessment engine 320 may obtain the baseline physiological information for the user and compare current physiological characteristics with the baseline physiological characteristics. For example, user assessment engine 320 may model a user's gaze (e.g., position of pupils) over a typical driving session to determine what a “normal” or average gaze is like for a user while driving. Such baseline gaze may be used to compare to a user's current gaze to determine whether the current pupil position deviates beyond an average range, causing user assessment engine 320 to determine that the user is distracted.” – teaches historical sensory data associated with the particular client computing entity (information from sensors may be stored in database, for example, user assessment may model user gaze over a driving session to determine average gaze to be used to compare current gaze to stored historical gaze)).
The combination of Cohen and Eleftheriou fails to explicitly teach sensory data associated with a plurality of vehicles.
However, analogous to the field of the claimed invention, Milton teaches:
sensory data associated with the plurality of vehicles (Milton, [0019] – “Some embodiments may use multi-layer hardware and software infrastructure to implement a multilayer vehicle learning infrastructure to receive vehicle sensor data and push control-system adjustments or other adjustment values based on machine-learning operations comprising sensor data from multiple vehicles, operator metrics stored in vehicle operator profiles, or features associated with geographic locations.”)
Therefore, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate the sensory data associated with a plurality of vehicles of Milton to the sensory features and cognitive assessments of Cohen and Eleftheriou in order to generate a global intermediate intervention score threshold based at least in part on sensory data associated with a plurality of vehicles. Doing so would improve accuracy, computation speed, and data privacy for autonomous driving systems (Milton, [0043]).
Claim 17 is similar to claim 8, hence similarly rejected.
Claim(s) 2-3, 11-12, and 20-22 is/are rejected under 35 U.S.C. 103 as being unpatentable over Cohen, Eleftheriou, and Milton as applied to claims 1, 10, and 19 above, and further in view of O’Toole et al. (US Pub. No. 2021/0216928, published July 2021, hereinafter “O’Toole”).
Regarding claim 2, the combination of Cohen, Eleftheriou, and Milton teach the computer-implemented method of claim 1, wherein:
The combination of Cohen, Eleftheriou, and Milton fails to explicitly teach the current prediction window is associated with N historical prediction windows, a historical prediction window of the N historical prediction windows is associated with a historical risk score, and generating the intermediate intervention score comprises: determining a risk score timeseries data object based at least in part on the real- time risk score and the historical risk score, determining an upward linearity score associated with the risk score timeseries data object; and generating the intermediate intervention score based at least in part on the real- time risk score and the upward linearity score.
However, analogous to the field of risk prediction and automated intervention, O’Toole teaches:
the current prediction window is associated with N historical prediction windows, a historical prediction window of the N historical prediction windows is associated with a historical risk score (O’Toole, [0169] – “Risk summary section 102 may present historical risk data aggregated by campus, city, country, region, or asset geofence, over a given date range, broken down by risk profile (e.g., asset protection). A user may select a risk profile category (e.g., asset protection risk) from the drop down menu 107. A list of risk profiles 108 is displayed. A risk profile may be for a region 109, country, city, campus, or asset 110. For each listed profile, summary information is displayed, such as the date the profile was applied 109, profile status 110 (e.g., ‘active’), a risk score range applicable to the profiles 111, and the elapsed times since the lowest 112, medium 113, and highest 114 risk scores were recorded.” – teaches the current prediction window associated with historical prediction window, where each historical prediction window is associated with a historical risk score (elapsed times since lowest, medium, and highest risk scores recorded, presents historical risk data by given date range)), and generating the intermediate intervention score comprises:
determining a risk score timeseries data object based at least in part on the real- time risk score and the historical risk score, determining an upward linearity score associated with the risk score timeseries data object (O’Toole, [0357] – “The systems and methods could determine whether a weather condition is abnormal based on analyzing historical data (e.g., historic temperature ranges, snowfall amounts, etc.) for a predefined amount of time in the past (e.g., the past five years). If the weather condition is abnormal, a risk score can be generated based on the abnormal weather condition such that the value of the risk score is increased due to the abnormality of the weather condition.” – teaches determining an upward linearity score associated with the risk score timeseries data, as supported in [0131] – “A graphical indication of a risk trend 503 is also displayed. In the example shown, an upwards arrow is used to indicate that the risk score of the relevant asset is increasing.” – here, O’Toole teaches the upward linearity score (risk trend 503)); and
generating the intermediate intervention score based at least in part on the real- time risk score and the upward linearity score (O’Toole, [0357] – “The systems and methods could determine whether a weather condition is abnormal based on analyzing historical data (e.g., historic temperature ranges, snowfall amounts, etc.) for a predefined amount of time in the past (e.g., the past five years). If the weather condition is abnormal, a risk score can be generated based on the abnormal weather condition such that the value of the risk score is increased due to the abnormality of the weather condition.” – teaches generating an intermediate intervention score based at least in part on the real-time risk score and the upward linearity score, by generating a risk score based on the abnormal conditions detected based on the historical data and trends). Therefore, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate the upward linearity score of O’Toole to the risk and intermediate intervention scores of Cohen, Eleftheriou, and Milton. Doing so would provide situational awareness while the baseline risk 334 score can be used for analyzing long term trends on an asset or neighborhood (O’Toole, [0352]).
Claims 11 and 20 are similar to claim 2, hence similarly rejected.
Regarding claim 3, the combination of Cohen, Eleftheriou, Milton, and O’Toole teaches the computer-implemented method of claim 2, wherein determining the upward linearity score comprises:
performing M iterations of a linearity score inference routine, wherein M=<N (O’Toole, [0386] – “Over time, the analytics service 628 can collect and store the data in the historical weather database, i.e., perform the steps 2402 and 2604 iteratively for days, months, years, decades, etc.” – teaches performing iterations of a linearity score inference routine), and wherein an iteration of the M iterations comprises:
determining an enlarged subset of the risk score timeseries data object that is associated with the real-time risk score and most recent m historical risk scores, wherein m is initially set to one and incremented at an end of the iteration (O’Toole, [0426] – “The element 2604 provides an indication of a dynamic risk score for the building an a tabulation of each of the threats affecting the building, for example, if another threat is affecting the building outside of the “Foil Break Alarm,” an active shooter threat, the active shooter threat and/or the foil break alarm can be shown in the element 2604 along with an indication of the risk score value for the particular threat.” – teaches determining an enlarged subset of the risk score timeseries data (dynamic risk score, tabulation of each of the threats) objected that is associated with the real-time risk score and the most recent historical risk scores, as supported in [0427] – “The monitoring client 128 can be configured to determine whether the risk score has risen and/or fallen over a predefined time period” – teaches m historical risk scores (risk scores over predefined time period)), and
determining, based at least in part on the enlarged subset and a linear trend data object for the enlarged subset, an upward linear score (O’Toole, [0427] – “Element 2706 provides an indication of whether the risk score has been rising and/or falling for a predefined time period. The monitoring client 128 can be configured to determine whether the risk score has risen and/or fallen over a predefined time period and can provide the risk card 2502 with an indication of the amount that the risk score has risen or fallen.” – teaches an upward linear score (client 128 provides card 2502 with amount risk score has risen or fallen) based at least in part on the enlarged subset (element 2604, which can be repeated iteratively as in [0386]) and a linear trend data object (element 2706)); and
determining the upward linearity score based at least in part on the plurality of upward linear score (O’Toole, [0428] – “The risk card 2502 includes the dynamic risk score which corresponds to the current risk score from real time active threats. Then it also includes baseline risk score which shows the risk score over an extended period of time. Combination of these two together makes it a meaningful insight” – teaches determining the upward linearity score (dynamic risk, baseline risk) based on the plurality of linearity score (monitoring client 128 providing amount risk score has risen or fallen over predefined time period to risk card 2502, as in [0427])).
Claim 12 is similar to claim 3, hence similarly rejected.
Regarding claim 21, the combination of Cohen, Eleftheriou, Milton, and O’Toole teaches the computer-implemented method of claim 2, wherein the real-time sensory timeseries data object comprises at least one of: one or more physiological condition measurements, one or more behavioral condition measurements, one or more environmental condition measurements, or one or more operation condition measurements (Cohen, [0033] – “In an implementation, environment sensors 150 may measure or otherwise gather (e.g., sense) information relating to a situation associated with the user. In other words, environment sensors 150 may sense the (intra-vehicle and/or extra-vehicle) surroundings of a user but not direct measurements of the user himself (which the user sensors 160 sense). A given situation sensor 150 may include, without limitation, a vehicle state sensor (e.g., steering wheel sensor, throttle sensor, brake sensor, seatbelt sensor, etc.), a Global Positioning System (“GPS”) device that provides location information, a camera directed to outside the vehicle (e.g., a lane keeping/warning system), a radar directed to outside the vehicle (e.g., a collision mitigation system, automatic radar, etc.), a thermometer (e.g., to sense freezing conditions), a rain sensor, and/or other sensor that can measure the environment associated with a user (e.g., surroundings of a vehicle operated by the user).”, [0034] – “In an implementation, user sensors 160 may measure or otherwise gather information of a user. In other words, user sensors 160 may measure physiological and other characteristics of the user. A given user sensor 160 may include, without limitation, a camera directed toward the user (e.g., to measure the user's pupil, monitor the user's head movements, etc.), microphone (e.g., to measure the user's speech), galvanic skin response receptors (e.g., mounted on a steering wheel or other input), and/or other sensor that can measure a physiological characteristic of the user.”, and in [0063] – “In other instances, situation assessment engine 310 may fuse various third party information and/or various sensor information to determine an assessment of a situational state.” – teaches wherein the real-time sensory timeseries data object (fused sensor information) comprises one or more of physiological condition measurements, behavioral condition measurements, environmental condition measurements (extra-vehicle information, information of vehicle surroundings), or operation condition measurements (intra-vehicle information)).
Claim 22 is similar to claim 21, hence similarly rejected.
Claim(s) 4 and 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Cohen, Eleftheriou, and Milton as applied to claims 1 and 10 above, and further in view of Morales et al. (NPL: Automatic Prediction of Maintenance Intervention Types in Roads using Machine Learning and Historical Records, published Aug. 2018, hereinafter “Morales”).
Regarding claim 4, the combination of Cohen, Eleftheriou, and Milton teaches the computer-implemented method of claim 1, wherein generating the real-time risk score comprises:
The combination of Cohen, Eleftheriou, and Milton fails to explicitly teach generating a real-time sensory feature data object based at least in part on the real-time sensory timeseries data object, generating, based at least in part on the real-time sensory feature data object and the global baseline sensory feature data object, a real-time sensory deficit data object, and generating the real-time risk score based at least in part on the real-time sensory deficit data object.
However, analogous to the field of risk prediction and automated intervention, Morales teaches:
generating a real-time sensory feature data object based at least in part on the real-time sensory timeseries data object (Morales, A Framework for Estimating Pre-Alerts Based on Feature Limits Paragraph 4 – “Figure 2 represents the estimated value of a feature with a known probability distribution.” – teaches generating a real-time sensory feature data objected based on the real-time sensory timeseries data object),
generating, based at least in part on the real-time sensory feature data object and the global baseline sensory feature data object, a real-time sensory deficit data object (Morales, A Framework for Estimating Pre-Alerts Based on Feature Limits Paragraph 4 – “A pre-alert is generated when the condition of the asset, as identified by its explanatory features, surpasses a pre-set threshold under a specific probability in a forecast scenario.” – teaches generating a real-time sensory deficit data objected (pre-alert) based at least in part on the real-time sensory feature data object (condition) and the global baseline sensory feature data objected (pre-set threshold)), and
generating the real-time risk score based at least in part on the real-time sensory deficit data object (Morales, A Framework for Estimating Pre-Alerts Based on Feature Limits Paragraph 4 – “Any triggered pre-alert is then quantified; a TSL is associated with each explanatory feature, defined according to the distance from the predicted value to the pre-set limits (e.g. RTi or RTi+1). Both a TSL and the associated degree of uncertainty (γ) are defined; the straightforward evaluation of that distance can follow a multiplicity of criteria, as predefined by the maintenance managerial body (MMB).” – teaches generating a real-time risk score (quantified pre-alert) based on the real-time sensory deficit data object (pre-alert generated based on the feature and baseline threshold)).
Therefore, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate the feature generation and risk score based on the real-time sensory deficit data object of Morales to the method of Cohen, Eleftheriou, and Milton in order to generate a risk score based on the real-time sensory deficit data object. Doing so would demand a higher reliability for features considered more relevant to the maintenance, or intervention, activities (Morales, A Framework for Estimating Pre-Alerts Based on Feature Limits Paragraph 5)
Claim 13 is similar to claim 4, hence similarly rejected.
Claim(s) 5-6, and 14-15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Cohen, Eleftheriou, Milton, and Morales as applied to claim 4 and 13 above, and further in view of Pulicharam et al. (US Pub. No. 2022/0246296, filed Feb. 2021, hereinafter “Pulicharam”).
Regarding claim 5, the combination of Cohen, Eleftheriou, Milton, and Morales teaches the computer-implemented method of claim 4, wherein:
The combination of Cohen, Eleftheriou, Milton, and Morales fails to explicitly teach the real-time sensory feature data object comprises a plurality of per-sensor real-time sensory features corresponding to the plurality of sensors, and the global baseline sensory feature data object comprises a plurality of per-sensor baseline sensory features corresponding to the plurality of sensors.
However, analogous to the field of risk prediction and automated intervention, Pulicharam teaches:
the real-time sensory feature data object comprises a plurality of per-sensor real-time sensory features corresponding to the plurality of sensors, and the global baseline sensory feature [[s]] data object comprises a plurality of per-sensor baseline sensory features corresponding to the plurality of sensors (Pulicharam, [0044] – “When the patient application 115 determines that it has received valid data for the health measurement, the patient application 115 can provide instructions for taking a second health measurement using a second health sensor. The process described above can be repeated for each of the health sensors in the patient's testing routine” – teaches real-time sensory feature data object (health measurement) comprising a per-sensor real-time sensory feature for each sensor of one or more sensors (process repeated for each sensor), and a global baseline sensory feature data object comprising per-sensor baseline sensory features, as in [0043] - “The patient application 115 can determine if the data is valid. Valid data may be data that falls within a predefined range. For example, a valid heart rate may be between 20 beats/minute and 220 beats/minute. If the data falls outside the predefined range, the patient application 115 can instruct the user to take the health measurement again. Valid data may also be data that has a particular characteristic. For example, if the health sensor is a scale, the patient application 115 may expect to receive a single data point indicating the patient's measured weight. If the data is instead time-series data with multiple data points, the patient application 115 may determine that the data is invalid” – teaches a global baseline sensory feature per-sensor (predefined range, particular characteristics dependent on the sensor, determining valid data per sensor per characteristic based on a baseline).
Therefore, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate the per-sensor features for each of the real-time sensory feature data object and global baseline sensory feature data object of Pulicharam to the method of Cohen, Eleftheriou, Milton, and Morales in order to establish the data objects comprising per-sensor features. Doing so would ensure acquisition of valid data for the purpose of accurately estimating conditions and risks (Pulicharam, [0043] and [0048]).
Claim 14 is similar to claim 5, hence similarly rejected.
Regarding claim 6, the combination of Cohen, Eleftheriou, Milton, Morales, and Pulicharam teaches the computer-implemented method of claim 5, wherein:
the real-time sensory deficit data object comprises a plurality of per-sensory real-time sensory deficit values corresponding to the plurality of sensors, and the computer implemented method further comprises (Pulicharam, [0055] – “the system can obtain a dataset including a plurality of values for each of a plurality of health measurements and metadata about the dataset (210). The system can obtain the dataset from a patient device (e.g., the patient device 110 of FIG. 1), which can temporarily store the dataset after the patient has taken health measurements using a plurality of health sensors (e.g., the sensors 105 of FIG. 1) that are wirelessly coupled to the patient device. Additionally or alternatively, the system can obtain the dataset from a wearable device of the patient. Additionally or alternatively, the system can obtain the dataset from an electronic medical record (EMR) of the patient.” – teaches a real-time sensory deficit data object (system obtains a dataset) comprising a per-sensor real-time sensory deficit value for each sensor (plurality of values for a plurality of measurements) of the one or more sensors):
generating the real-time sensory deficit data object comprises, for a sensor of the plurality of sensors, determining a per-sensory real-time sensory deficit value of the plurality of per-sensory real-time sensory deficit values based at least in part on a per-sensor real-time sensory feature, for the sensor, of the plurality of per-sensor real-time sensory features and a per-sensor baseline sensory feature, for the sensor, of the plurality of per-sensor baseline sensory features (Pulicharam, [0055] – “the system can obtain a dataset including a plurality of values for each of a plurality of health measurements and metadata about the dataset (210).” – teaches generating the real-time sensory deficit data object (dataset) comprising per-sensory real-time sensory deficit values (plurality of values for each a plurality of health measurements) based at least in part on the per-sensor real-time sensory features (health measurements) and the per-sensor baseline sensory feature for the sensor, as in [0043] – “The patient application 115 can determine if the data is valid. Valid data may be data that falls within a predefined range.” – teaches per-sensor baseline sensory features), and
generating the real-time risk score is based at least in part on the plurality of per-sensor baseline sensory feature (Pulicharam, [0063] – “The risk score may indicate the likelihood that the patient will experience the adverse health event within a specified time period. The time period may be the next day, the next week, the next month, the next two months, the next three months, the next four months, the next year, or more. As mentioned above, in one example, the intermediate scores are based on health measurement values from the preceding days in the current month. Therefore, the risk score in one month may be independent from the risk score in another month. The system can update the risk score for a particular month as it obtains new data during that month (e.g., as it obtains daily health measurement values from the patient).” – teaches generating a real-time risk score based at least in part on the plurality of per-sensor baseline sensory feature (health measurements, validated by baseline sensor features).
Claim 15 is similar to claim 6, hence similarly rejected.
Response to Arguments
Applicant’s arguments, see pp. 1-3 of Remarks, filed 14 January 2026, with respect to the rejection(s) of claim(s) 1, 7-8, 10, 16-17, and 19 under 35 U.S.C. 103 have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made over Cohen et al. (US Pub. No. ) in view of Eleftheriou and further in view of Milton. Cohen teaches the amended limitations of claim 1 regarding “receiving a global intermediate intervention score threshold and a global baseline sensory feature data object for the cognitive condition…”, “identifying a real-time sensory timeseries data object…”, “generating, using the real-time risk scoring machine learning model and based at least in part on the real-time sensory timeseries data object…”, “generating, based at least in part on the real-time risk score, an intermediate intervention score”, “performing, responsive to determining that the intermediate intervention score…”. Eleftheriou teaches the limitations of claim 1 regarding “receiving, one or more intermediate intervention response features…” and “generating, using the intermediate risk scoring machine learning model and based at least in part on the one or more intermediate intervention response features…”. Milton teaches the limitations of claim 1 regarding “generating, by one or more processors and using a hierarchical intervention recommendation machine learning framework that comprises…”, “generating… the adjusted risk score…”, “generating, by the one or more processors and responsive to the adjusted risk score…”, and “updating, by the one or more processors, a navigation route…”.
In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Matus et al. (US Patent No. 11,878,720, filed April 2020) teaches systems and methods for adaptive risk modeling in autonomous vehicles with a plurality of sensors. Teaches implementing a responsive action based upon risk analysis. Teaches several thresholds for risks and sensory features. Teaches measuring human driver behavior with biometric monitoring devices to detect physiological or cognitive states of individuals.
Heldman et al. (US Patent No. 11,367,519, filed July 2018) teaches systems and method for instructions of pharmaceutical treatments by measuring physiological or electrophysiological signals from a monitored user. Teaches sensors for measuring the cognitive state and impairment of individuals, including those with Alzheimer’s. Teaches assessing and determining severity scores for the associated cognitive impairment. Users may be performing activities such as driving when having information captured.
Minea et al. (NPL: Advanced e-Call Support Based on Non-Intrusive Driver Condition Monitoring for Connected and Autonomous Vehicles, published Dec. 2021) teaches equipping a semi-autonomous vehicle with a complex sensor structure that provides centralized information regarding physiological signals of drivers/passenger and their location information. Teaches monitoring conditions of driver to reduce risk of accidents. Teaches utilizing deep learning methods for determining the cognitive state of the driver including drowsiness.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LOUIS C NYE whose telephone number is 571-272-0636. The examiner can normally be reached Monday - Friday 9:00AM - 5:00PM.
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, MATT ELL can be reached at 571-270-3264. 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.
/LOUIS CHRISTOPHER NYE/Examiner, Art Unit 2141
/MATTHEW ELL/Supervisory Patent Examiner, Art Unit 2141
Prosecution Insights