Impact of reconstruction and special iterative algorithm on Non-Stationary sEMG decomposition

To assess the impact of signal reconstruction and specialized iterative algorithms on the decomposition of non-stationary surface electromyography (sEMG) signals, this study proposes a Segmentation and Re-ordering based on energy for KmCKC decomposition (SROK) method. SROK first partitions the sEMG signal into short epochs and reorders these epochs by energy. The reordered signal is then divided into a small number of longer segments, and the KmCKC algorithm is iteratively applied to each segment. The resulting innervation pulse trains (IPTs) are then regrouped into their original temporal sequence, ensuring precise signal decomposition. Validated on both synthetic and experimental biceps brachii sEMG data from seven adults, SROK effectively decomposed a broader temporal range in simulations, identifying an average of 23.7 IPTs with 97.9% accuracy at a signal-to-noise ratio (SNR) of 20 dB. For experimental data, SROK identified an average of 17.0 motor units (MUs) across subjects, with a mean pulse-to-noise ratio (PNR) of 38.6. Compared with standard KmCKC, SROK demonstrated a substantial improvement in MU yield, with an average increase of 12.9 MUs, confirming its reliable decomposition performance. Overall, SROK exhibits superior capability in handling non-stationary signals and provides deeper insight into muscle activation, force generation, and potential applications in injury assessment, fatigue monitoring, and rehabilitation.