| Summary: | 碩士 === 清雲科技大學 === 電子工程所 === 100 === According to the spirit of e-Home, e-Health, “offers the ubiquitous and adaptation service for the specified family”, this thesis presents a novel bonded foil strain floor and activities signal processing technique, it is used to sense the single actions signal and identifying those living activities in living house. This identification technique can make the smart home-cared environments system to understand the activity situation of someone family, and to offer intelligence and conformable services further. The pressure-sensed signals processing in this identification system is composed by four major parts: activity signal processing, feature extraction, model building, and activity recognition. In this thesis, use Hidden Markov Model and Gaussian Mixture Model to construct the activity recognition model, and based on those models to identify the continuous living activities of family. This design makes the smart home-cared serving systems more accurately to care family keep far away from accidents. The correct rate based on those presented models is up to 92.17% on average of the specified activities identification.
|
|---|
