| Summary: | 碩士 === 銘傳大學 === 電腦與通訊工程學系碩士班 === 99 === Due to the fast development in MEMS, human activity recognition can be done not only by the video recorder through image processing method but also by the wearable sensors through wireless sensor network. In this paper, the accelerometer, gyroscope, wireless sensor network and fuzzy system are used to perform the human activity recognition. The two system result demonstrates that the system can recognize two postures (static and dynamic) and 5 activities, such as sitting, lying, standing, running and walking. We designed two fuzzy inference systems with different parameters called System I and System II. System I utilized the magnitudes of five parameters to define the membership functions of the fuzzy system, and System II used five parameters from the difference rate of parameters as the inputs of membership functions. From the result, the recognition rate of System II is higher than that of System I. The reason is that the magnitudes of the parameters from System I are easily affected by different users. However, the different rates of the parameters from System II can be varied by different activity and eliminate the use of personal qualities, so the future recognition will use the System II to analyze the human activity. Finally, we apply the results to stroke rehabilitation treatment in the home care system with half-elbow and elbow rehabilitation action. The proposed rehabilitation system can use the parameters to recognize the elbow function through the strength, degree and frequency of the user’s elbow rehabilitation action. The result will apply to clinical rehabilitation treatment in the future.
|
|---|
