The wrist is among the most flexible joints in the human body.Decomposing the surface Electromyography (sEMG) signals enables the estimation of the deep neural drives underlying human body movements.Current research on wrist moments primarily focuses on isometric contractions,leaving the decoding of neural drives under dynamic conditions an area requiring substantial exploration.This study investigated Motor Unit (MU) decomposition during wrist movements under varying resistance levels,which were precisely controlled using a magnetorheological damper.The collected continuous electromyography signals were segmented into short intervals where Motor Unit Action Potential (MUAP) waveforms remained relatively stable.Then,a classic decomposition algorithm was applied to each short interval to obtain the Motor Unit Spike Train (MUST),and the MUs were tracked across consecutive intervals via their overlapping parts to reconstruct complete firing sequences.This paper studied the decomposition of MUs during wrist extension and flexion under three resistance levels: 20%,40%,and 60% of the Maximum Voluntary Contraction (MVC).Results showed that the proposed dynamic decomposition algorithm effectively decomposed MUs from forearm sEMG signals,although performance somewhat declined with increasing resistance.During wrist extension,up to 10±1 MUs were decomposed,with Pulse-to-Noise Ratio (PNR) and Silhouette Coefficient (SIL) reaching 19.87±1.42 dB and 0.91±0.03,respectively.While during wrist flexion,up to 22±3 MUs were decomposed,with PNR and SIL values of 20.69±2.14 dB and 0.92±0.03,respectively.This study confirms the feasibility of neural drive decoding from sEMG signals during wrist movements under different resistances,highlighting significant potential for the application of High-Density sEMG (HD-sEMG) under dynamic muscle contractions.