Method and System for Context-based Activity Recognition | Project and Innovation Support

Method and System for Context-based Activity Recognition

Publication Number: SI24356A
Application Date: 01. 8. 2014
Assignee/Applicant: Jožef Stefan Institute [SI]
Inventor: Matjaž Gams, Hristijan Gjoreski, Mitja Luštrek, Boštjan Kaluža
Title: Metoda in sistem za prepoznavanje aktivnosti na podlagi konteksta [SL], Method and System for Context-based Activity Recognition [EN]
Description: The present invention relates to improved methods for identifying daily activities of a person. It includes a new method for identifying activities that uses the context of the current activity to determine and improve the accuracy of its recognition. The method uses sampled data on linear and gravitational acceleration. This information is provided by a motion sensor, directly or indirectly supported by the subject. The context extractor processor calculates multiple contexts from the data. The processor for context modeling is a machine learning memory that teaches the model for each value of the context. The activity of the subject is recognized by combining, t. j. pooling the results of several context models. The invention provides an application for identifying an activity that determines the activity of the subject. An application may include a notification mechanism with an alarm if a drop is detected. Among other things, it can include: a gym application, an activity-based daily application, an out-of-home help application, etc. The technical effect of the experience allows you to control the smart home and more precisely detect the fall that triggers an automatic emergency call.
Category: Electronics, IT and Telecommunications
Technology application codes: Electronics, IT and telecoms
Market application codes: Communications, Computer related, Other electronics related
Patent offices: UIL RS