Unified platform for domain adaptable human behaviour inference

A system for obtaining unified domain adaptable human behaviour inference using a unified platform comprising:

• a memory storing instructions and one or more modules;
• a database;
• one or more communication or input/output interfaces;
• and
• one or more hardware processors coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by the instructions to execute the one or more modules comprising:
• a input module for receiving a plurality of sensor data using a plurality of sensors, wherein the plurality of sensor data is in one or more sensor data formats, wherein the plurality of sensors includes wearable sensors, infrastructure sensors, near field sensors, infrared sensors, Near-Field Communication (NFC) sensors and soft sensors and wherein the plurality of sensors comprises databases comprising details regarding one or more users;
• a pre-processor for processing the received plurality of sensor data to a time-series format;
• a low level inference module for analyzing the processed plurality of sensor data using cross-sectional analysis to obtain low level inference, wherein the cross-sectional analysis for the low level inference includes analyzing details of location, activity, temperature and physiology;
• a high level inference module for analyzing the processed plurality of sensor data using a plurality of longitudinal analysis and Fourier analysis techniques based on the low level inference, domain knowledge and standard human metadata including mobility, proximity, activity of a human, wherein the domain knowledge includes health, shopping, retail and driving that are dynamically updated in a domain database, wherein the plurality of longitudinal analysis and Fourier analysis techniques includes welch analysis, cross-spectrum analysis, Feature of interest (FOI) identification and time-series clustering, autocorrelation-based distance estimation and exponential smoothing, seasonal & non-seasonal models identification, ARIMA modelling, Hidden Markov models, Long short term memory (LSTM) along with the low level inference, the standard human metadata and the domain knowledge to provide insights for longitudinal tracking, prediction, and anomaly detection, and selecting a longitudinal analysis technique from the plurality of longitudinal analysis techniques for obtaining high level inference by the high level inference module connected to the domain database and a human metadata database, wherein the longitudinal analysis technique is selected based on expectation maximization techniques;
• obtaining the domain adaptable human behavior inference using the unified platform based on the obtained low level inference and the high level inference,
• wherein for the health domain, the plurality of sensors including accelerometer, gyroscope, barometer, photoplethysmograph (PPG) obtains the domain adaptable human behavior inference corresponding to cardiac care pathway for patients, the low level inference obtained are Metabolic Equivalents (MET) values, heart rate, breathing-rate, breathing power and the high level inference obtained is computational model of cardiac fatigue per session of activity or cardiac care,
• wherein for the health domain, the plurality of sensors including passive infrared (PIR) sensors, door contact sensors and pill-box sensors obtains the domain adaptable human behavior inference corresponding to ambient assisted living for geriatric subjects, the low level inference obtained are daily and weekly routine of the geriatric subjects and the high level inference obtained is mild cognitive impairment detection among the geriatric subjects,
• wherein for a transportation domain, the plurality of sensors including Global Positioning System (GPS), inertial sensors obtains the domain adaptable human behavior inference corresponding to driving behavior understanding for transportation, the low level inference obtained are a lateral velocity and a jerk and the high level inference obtained is rash driving details,
• wherein for the transportation domain, the plurality of sensors including photoplethysmograph (PPG), accelerometer, gyroscope, barometer obtains the domain adaptable human behavior inference corresponding to ubiquitous journey recognition, the low level inference obtained is a vehicle recognition and the high level inference obtained is ubiquitous journey recognition;
• and
• a unified display module for displaying the obtained human behavior inference based on a user query as only the low level inference or only the high level inference or both the low level inference and the high level inference, wherein for the transportation domain, the user query corresponds to if a driver is driving rash, then the system displays in confirmation along with the lateral velocity and the jerk,
• wherein the unified platform provides a unified low level inference and a high level inference of the domain adaptable human behaviour inference across multiple domains including the health, the retail and the transportation, and
• wherein an output module is configured to a set of Application Programming Interface (API) or service for user applications to interface with the unified platform to connect with the unified display module.

One or more non-transitory machine readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause:

• receiving a plurality of sensor data using a plurality of sensors, wherein the plurality of sensor data is in one or more sensor data formats, wherein the plurality of sensors includes wearable sensors, infrastructure sensors, near field sensors, infrared sensors, Near-Field Communication (NFC) sensors and soft sensors and wherein the plurality of sensors comprises databases comprising details regarding one or more users;
• processing the received plurality of sensor data to a time-series format;
• analyzing the processed plurality of sensor data using cross-sectional analysis to obtain low level inference, wherein the cross-sectional analysis for the low level inference includes analyzing details of location, activity, temperature and physiology;
• analyzing the processed plurality of sensor data, by a high level inference module, using a plurality of longitudinal analysis and Fourier analysis techniques based on the low level inference, domain knowledge and standard human metadata including mobility, proximity, activity of a human, wherein the domain knowledge includes health, shopping, retail and driving that are dynamically updated in a domain database, and wherein the plurality of longitudinal analysis and Fourier analysis techniques includes welch analysis, cross-spectrum analysis, Feature of interest (FOI) identification and time-series clustering, autocorrelation-based distance estimation and exponential smoothing, seasonal & non-seasonal models identification, ARIMA modelling, Hidden Markov models, Long short term memory (LSTM) along with the low level inference, the standard human metadata and the domain knowledge to provide insights for longitudinal tracking, prediction, and anomaly detection;
• selecting a longitudinal analysis technique from the plurality of longitudinal analysis techniques for obtaining high level inference by the high level inference module connected to the domain database and a human metadata database, wherein the longitudinal analysis technique is selected based on expectation maximization techniques;
• obtaining the domain adaptable human behavior inference using a unified platform based on the obtained low level inference and the high level inference,
• wherein for the health domain, the plurality of sensors including accelerometer, gyroscope, barometer, photoplethysmograph (PPG) obtains obtaining the domain adaptable human behavior inference corresponding to cardiac care pathway for patients, the low level inference obtained are Metabolic Equivalents (MET) values, heart rate, breathing-rate, breathing power and the high level inference obtained is computational model of cardiac fatigue per session of activity or cardiac care,
• wherein for the health domain, the plurality of sensors including passive infrared (PIR) sensors, door contact sensors and pill-box sensors are used for obtains the domain adaptable human behavior inference corresponding to ambient assisted living for geriatric subjects, the low level inference obtained are daily and weekly routine of the geriatric subjects and the high level inference obtained is mild cognitive impairment detection among the geriatric subjects,
• wherein for a transportation domain, the plurality of sensors including Global Positioning System (GPS), inertial sensors obtains the domain adaptable human behavior inference corresponding to driving behavior understanding for transportation, the low level inference obtained are a lateral velocity and a jerk and the high level inference obtained is rash driving details,
• wherein for the transportation domain, the plurality of sensors including photoplethysmograph (PPG), accelerometer, gyroscope, barometer obtains the domain adaptable human behavior inference corresponding to ubiquitous journey recognition, the low level inference obtained is a vehicle recognition and the high level inference obtained is ubiquitous journey recognition;
• and
• displaying the unified domain adaptable human behavior inference on a unified display module based on a user query as only the low level inference or only the high level inference or both the low level inference and the high level inference, wherein for the transportation domain, the user query corresponds to if a driver is driving rash, then a system displays in confirmation along with the lateral velocity and the jerk,
• wherein the unified platform provides a unified low level inference and a high level inference of the domain adaptable human behaviour inference across multiple domains including the health, the retail and the transportation, and
• wherein an output module is configured to a set of Application Programming Interface (API) or service for user applications to interface with the unified platform to connect with the unified display module.