An automatic SSA-based de-noising and smoothing technique for surface electromyography signals

Abstract The surface electromyography (sEMG) signal is a low amplitude signal that emanates from contracting muscles. It can be used directly to measure muscle activity (once noise has been removed) or it can be smoothed for some other application, e.g., orthoses or prostheses control. Here, an automatic heuristic procedure is presented which applies singular spectrum analysis (SSA) and cluster analysis to de-noise and smooth sEMG signals. SSA is a non-parametric technique that decomposes the original time series into a set of additive time series in which the noise present in the acquired signal can be easily identified. The proposed approach constitutes an alternative to the traditional smoothing procedures, such as moving average (MOVAG), root mean square (RMS), or low-pass Butterworth filtering that are used to extract the trend of the signal. To assess the quality of the method, the results of its application to a non-stationary sEMG signal are compared with those of other step-wise filtering and smoothing techniques.