基於肌電圖之機器臂控制系統

Abstract

肌電圖（Electromyography，EMG)是一種在肌肉收縮過程所產生的生理訊號，含有使用者移動的意圖，因此以肌電圖做為義肢或機器的控制命令，是非常直接且直覺的方式。基於此，我們對肌電圖與相對應之手部運動的關係進行分析，據以建構出一套基於肌電圖的機器臂控制系統，操作者可藉由此系統利用人手運動來操控機器臂。為了分析前臂的運動，此系統利用兩個通道的表面電極，從肱二頭肌與肱三頭肌得到肌肉訊號，接著擷取其四種特徵（Mean absolute value, Variance, Zero crossing, Willison amplitude），以此評估肌肉活動的狀況，之後透過分類器的設計來辨識出前臂的運動，並將之轉換成機器臂的控制命令，機器臂則根據此命令產生相對應的運動；為了達到即時控制的目的，我們也發展一套人機介面來整合訊號擷取、分析以及機器臂的控制。

EMG is a physiology signal generated during muscle contraction. For being the indication of the human motion intention, it is very intuitive and direct to use the EMG as the command for robot or porsthesis control. In the thesis, we aim to analyze the relationship between EMG and corresponding arm movement, and then establish an EMG-based robot control system. With the system, the operator can manipulate robot via the arm movement directly.To analyze the forearm movement, the system uses two surface electrodes to measure the EMG signal from the biceps and the triceps muscles. Four kinds of features, Mean absolute value, Variance, Zero crossing, and Willison amplitude, are adopted to evaluate the muscle activities. Accordingly, a classifier is developed to tell different forearm movements. Finally, for achieving real-time control, a human-machine interface is then developed to for signal extraction, signal analysis, and robot arm control.