An EMG-based adaptive algorithm for motion detection in non-stationary noise

Noise is always present in surface electromyographic signals (sEMG) used in bioelectric prosthesis control. To detect sEMG signals against the noise background, a two-threshold algorithm is usually used. It was shown that performance of the algorithm decreases in the case of non-stationary noise. The purpose of this study was to improve the performance of the two-threshold algorithm by adaptive adjusting of the algorithm parameters. Firstly, the noisy conditions were simulated. The sEMG signals were recorded from wrist flexor muscles on one male volunteer with no prior known trauma on his right forearm. The signals were acquired by gel electrodes. The noise was modeled as a Gaussian process with the power increasing with time and added to the recorded sEMG signals. The sum of sEMG and noise was filtered by a whitening filter to make two successive signal samples independent.The proposed algorithm consists of the two-threshold movement-detecting algorithm associated with the noise estimator. The former detects the movements and the latter estimates the noise standard deviation during the pauses between the movements. Then the estimated noise standard deviation is used to adjust the parameters of the movement-detecting algorithm. It was shown that the proposed algorithm has higher performance than the nonadaptive two-threshold algorithm in the case of non-stationary noise.