A biosignal quality assessment framework for high-density sEMG decomposition

Abstract

Surface electromyography (sEMG) signal is superimposed by hundreds of motor unit (MU) discharge waveforms during muscle contraction. With the development of high-density sEMG (HD-sEMG) acquisition techniques, exploring the motoneuron discharge events, which is called sEMG decomposition, has shown advantages over traditional analysis. However, experimentally recorded HD-sEMG signals are nonetheless inevitably contaminated by several kinds of noise, including baseline noise (BLN), electrocardiogram interference (ECI), powerline noise (PLN), and white Gaussian noise (WGN). These noises directly degrade the signal quality and affect the performance of decomposition, but the effect of noise type on the performance of sEMG decomposition remains to be determined. To evaluate the impact of noise, this paper presents a biosignal quality assessment framework for HD-sEMG decomposition. Firstly, the four most prominent types of noise were analyzed via simulation modeling and experimental verification. The results showed that WGN was the dominant noise factor influencing sEMG decomposition, while the other types of noise had no significant influence. Secondly, a signal-to-white Gaussian noise ratio (SWGR) estimation method based on canonical correlation analysis (CCA) was proposed, and SWGR was further used as a metric to evaluate the biosignal quality for HD-sEMG decomposition. The experimental results indicated that SWGR accurately (R2≥0.72 for decomposed MU number) reflected the performance of the sEMG decomposition. The outcomes provide guidance for the preprocessing of HD-sEMG with regard to decomposing MU discharge, potentially paving the way for a non-invasive neural interface.