| Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 103 === Due to the incorporation of pervasive computing and machine learning algorithm, a wide variety of applications have been developed. One of these applications in smart environment is to monitor and track the functional status of residents and provide personal support for residents. Those application rely on being able to automatically recognize the activity perform by resident based on the series of streaming sensor data which collected by different sensors. In this thesis, we attempt to use a structure of hierarchy to improve the performance of activity recognition. We adopt a genetic algorithm-based method to automatically search the hierarchy which has the best performance in terms of accuracy. Furthermore, instead of experimenting on scripted or pre-segmented sequence of sensor events related to activities, we adopt a sliding window based approach to perform activity recognition in an on line or streaming fashion. From our experiment, whether in the situation that only considering the predefined activities or coupling with the “other” activity, the accuracy all can slightly be improved.
|
|---|
