| Summary: | 碩士 === 國立交通大學 === 電控工程研究所 === 107 === In this thesis, surface EMG signal (sEMG) is used to estimate the wrist angle. sEMG signal reflects the effort of the muscle. When measuring sEMG signal, a single electrode can receive EMG signal from multiple muscle group, affected by the cross-talk from muscles Previous research applied ICA algorithm to sEMG signal to separate EMG signal from different muscle group. However, EMG signal source is a highly gaussian signal source, ICA is ineffective. This thesis devised a sEMG signal power separation method, windowed RMS is used to extract sEMG signal power, then two signal separation algorithm (TDSEP and nICA) is applied, avoiding the problem of high gaussianity. Using the proposed wrist angle estimation system, we evaluate the performance improvement of using two signal separation algorithm in wrist angle estimation. In 4-channel sEMG, both algorithms improve the estimation accuracy (quantified by RMSE) by 15~20%; in 6-channel sEMG, the estimation accuracy is improved by 7~16%.
|
|---|
