| Summary: | 博士 === 國立交通大學 === 電控工程研究所 === 99 === Electromyography (EMG) signal, as a physiological signal generated during muscle contraction, implicates several important messages, such as the muscular force level and operator’s intention. It is very suitable to serve as the control signal for the manipulation of the rehabilitation device, human-assisting robot and others. To develop an effective robot motion governing based on using EMG signal, this dissertation proposes a so-called initial point detection method to discriminate the up limb motion onset by detecting the instant when the magnitude of the extracted EMG feature reaches the upper critical value and offset when that descends to the lower critical value from onset state. Consequently, the mapping between the limb EMG signals and the corresponding robot arm movements can be established very quickly. Meanwhile, due to the individual fuzziness, the tuning of the system parameters for the individual user is not that straightforward. Thus the concept of the fuzzy system is employed so that the tedious process encountered in the trial-and-error method can be avoided. While the proposed system is shown to be effective for robot motion governing, it is not appropriate to serve as a classifier for more than 1-DOF (degree of freedom) limb motion as it has larger muscle mutual interference. To tackle this, the EMD is applied to decompose the EMG signals into several intrinsic mode functions (IMFs). Each IMF represents different physical characteristic, so that the major muscular movements can be recognized. Meanwhile, for multi-DOF limb motion, the fuzzy system adopted for 1-DOF motion is not efficient enough for the tuning of the critical values for each individual user. For its excellence on adaptation, the adaptive neuro-fuzzy inference system (ANFIS) is employed to realize the fuzzy system. Because neither complicated computation nor training and learning processes are needed, the proposed scheme not only simplifies system complexity, but also increases the efficiency in motion governing.
|
|---|
