Design of robust adaptive Volterra noise mitigation architecture for sEMG signals using metaheuristic approach

Surface Electromyogram (sEMG) signals, like other electrophysiological measurements, get corrupted by several artefacts; much critical helpful information regarding a person’s clinical conditions may alter or be lost entirely. Therefore, the sEMG signal must be filtered to minimise such artefacts. The drive of this paper is to propose an efficient adaptive Volterra noise mitigation architecture (AVNMA) for the sEMG signal. Further, the optimal coefficients for designing the Volterra filter architecture are achieved by applying the recently proposed metaheuristic algorithm, gannet optimisation algorithm and compared to the performance of other benchmark optimisation algorithms, namely harmony search optimisation algorithm and teaching learning-based optimisation algorithm to the proposed architecture. The quantitative analysis based on the proposed architecture is measured in terms of several metrics such as mean squared error, normalised root mean square error, peak reconstruction error, mean difference, maximum error and signal-to-noise ratio at various noisy environments in the presence of additive white Gaussian noise, including artefacts such as muscle noise, baseline wandering, electrode misplacements, and electrical interference. The outcomes of experimental research (SNR = 102.236 dB andMSE = 6.71E-09 for healthy-sEMG) ensure the superiority of the gannet optimisation algorithm based-AVNMA over the other metaheuristic algorithm based-AVNMAs. The proposed approach is verified parametrically and non-parametrically in a statistical sense with the help of two sample t-tests and the Mann-Whitney U test, respectively.