Method for recognizing user-context by using mutimodal sensors

The present invention relates to a method of recognizing user-context by using a multimodal sensor comprising an acceleration sorting step in which mobility feature candidates are extracted from user acceleration data collected through an acceleration sensor, at least one mobility feature is selected from the extracted mobility feature candidates based on relevancy and redundancy of the extracted mobility feature candidates, and a mobility type of a user is inferred from the selected mobility feature by using a first time sequential stochastic model.In an audio sorting step, surrounding environment features are extracted from audio data concerned about user surrounding environment and collected through an audio sensor and a type of surrounding environment is inferred from the extracted surrounding environment features by using a second time sequential stochastic model.Further, the method of recognizing user-context comprises a user-context recognizing step by which the user context is recognized by using at least one of the types of user motion and surrounding environment. [Reference numerals] (AA) Start; (BB) End; (S111) Collect acceleration data through acceleration sensor; (S113) Extract mobility feature candidate; (S115) Select mobility feature; (S117) Infer mobility type; (S131) Collect audio data through audio sensor; (S133) Extract surrounding environment feature; (S135) Infer surrounding environment feature; (S150) Recognize user context; (S170) Confirm validity; (S190) Valid user context is recognized?