Detecting Noise Reduction in EMG Signals by Different Filtering Techniques
Afsaneh Koohestani, Sajjad Jafari · Advanced Science Letters · 2013
A common problem in processing of EMG signals for assessment of their applications in different biomedical fields such as diagnosing fatigue level, rehabilitation and control of wrist and finger movements in individuals suffering from c5 and c6 spinal traumas is the high susceptibility of the signals to noise. The EMG signal is highly variable in terms of intensity and frequency and has a high frequency range, so its noise also has a high frequency. This paper is focused on filtering the signal in order to reduce a considerable part of the noise. It includes the use of three filters named butterworth, Wiener, and least mean square (LMS). The LMS is a member of the adaptive filters group, while Wiener is a linear filter. Different techniques are used to compare the said filters. A non-linear parameter like correlation dimension and covariance are used to estimate the amount of filtered noise. The other technique includes using a random unbiased parameter like cross-correlation. The non-linear parameter showed better accuracy in showing the amount of non-linear noise reduction. The Wiener filter eliminated larger percents of noise. Data was collected from 10 individuals all being of perfect health and free of any specific disease.