Contamination in Electromyography Signals and Noise Removal Techniques

Abstract

The electromyography (EMG) signal can be contaminated with noise during data collection. For example, when the EMG signal is acquired from muscles in the torso, the electrocardiography (ECG) signal coming from heart activity can interfere. In this thesis, we proposed a novel method on noise removal and the signal- to-noise ratio (SNR) estimation algorithms. For the noise removal method, a technique based on discrete stationary wavelet transform (DSWT) is proposed to remove ECG interference from the EMG signal while taking into account the SNR. The contaminated EMG signal is decomposed using 5-level DSWT with the Symlet wavelet function. A clean EMG signal can then be obtained by inverse DSWT mapping of the new thresholded coefficients. The performance based on mean absolute error, correlation coefficient, and relative error shows that the DSWT method is better than a high-pass filter. For the SNR estimation method, we present a novel SNR estimation in the EMG signal contaminated with the ECG interference. We calculate the features from the EMG signals. Then, the features are used as an input of a neural network (NN). The NN output is an SNR estimate. The results showed that the waveform length was the best feature for estimating SNR. It gave the highest average correlation coefficient at 0.9663. These results suggested that the waveform length was able to be deployed not only in an EMG recognition system but also in an EMG signal quality measurement when the EMG signal was contaminated with the ECG interference.