Robust Features for Elbow Joint Angle Estimation Based on Electromyography

Abstract

A noisy environment is a major problem which has to be resolved to get a good performance in the estimation. A robust feature is important in order to obtain an accurate position of the elbow joint from the electromyography (EMG) signal. The objective of this research is to modify and assess the time domain features which robust against the white Gaussian noise. In this work, the EMG signal (from biceps) contaminated by artificial white Gaussian noise was extracted using twelve standard time domain features and one modified feature. The threshold of the modified feature (MYOPM) was calculated based on the root mean square (RMS) of the contaminated EMG signal. The linear Kalman filter was used to refine the EMG features and to improve the estimation. The robustness of the features was calculated using the root mean square error (RMSE). Based on the RMSE values, it shows that the proposed feature MYOPM is the most robust feature (the lowest median RMSE of 9º) for the signal to noise ratio (SNR) ranged between 17.96 and 60 dB, compared with the others’ features. The mean RMSE of the MYOPM feature improves by 27.91% from the prior feature (MYOP).