Automatic calibration in adaptive filters to EMG signals processing

Abstract

In this work, an adaptive filtering that includes an automatic calibration process to acquire EMG (electromyography) signals has been implemented. We propose a novel technique called "autocalibration" to minimize the noise generated by the contact of the skin with sensors used (electrodes) during physical activities development. Adaptive filtering has been used considering both, physical activity and sweating in persons are factors that could change the measurement conditions. To evaluate the proposed technique, a group of persons have been selected to develop physical activities for different intensities of effort. Relative improvement of the signal to noise ratio (RI-SNR) has been used to compare both, the proposed technique and adaptive filters that use "white noise" as reference signal. This work is focused on Wiener, LMS and RLS estimators, with measurements performed before and after of the physical activities. Applying the autocalibration process in adaptive filtering, an improvement up to 45,49% compared with the corresponding that uses "white noise" for calibration has been obtained.