Prediction of Limb joint angle using EMG signals and artificial neural network

• Main Authors: Min-Feng Cai, 蔡旻峰
• Other Authors: Po-Lei Lee
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/89902732322372755702

碩士 === 國立中央大學 === 電機工程研究所 === 99 === Due to the coming of aging society and high labor cost, modern society has increased demands on medical instruments and assistive prostheses. One novel prosthesis which has drawn great attention is the use of electromyography (EMG) as control signal to control the movements of a prosthesis. Nevertheless, one key element to the success of these intelligent prostheses is the reliability and stability of articular angle during movements. This study aims to develop a method for predicting the changes of articular angles during movements. The predicted articular angle can be used to facilitate the rotation stability of motor control in an artificial prosthesis. In this study, the articular angles were measured by a thin bending flex sensor. The measured articular angles were used as ground truth for performance evaluation. By recording the EMG signals as inputs and measured articular angles as outputs to an artificial neural network (ANN), the ANN predicts the angle position at next time point based on past information. The root Mean Square Error of predicted hand, hip ,and knee joint angles is 1^o~2^o , 7^o~8^o , and 5^o~7^o , respectively.