A Pilot Study on Co-located EMG and FMG Transient-phase Signal Fusion Approach for Hand Motion Characterization

Multimodal sensor combinations can improve the potential of human-machine control systems. However, the majority of studies focus on a single modality, either Electromyography (EMG) or Force myography (FMG). In such studies, more effort is made towards utilizing the steady phase of the signal ignoring the transient-phase due to its non-stationarity. Thus, in this work, we attempt to leverage the combination of EMG and FMG in characterizing hand gestures by examining the potential of transient-phase signal. To effectively combine both signals, an attention mechanism was introduced to fuse time-domain features from EMG and FMG. The fused representations were subsequently classified using three classification algorithms. Experimental results of 8 gestures recorded from 6 subjects show that the EMG and FMG combination improved the performance compared to single modality. In addition, system redundancy analysis showed that reducing the number of channels by 25% has no significant influence on the performance of the system. This result could influence the optimization of human-machine control systems by reducing complexity and cost while maintaining robustness and performance.Clinical Relevance- This work can be clinically applicable in stroke patients' rehabilitation and restoration of upper limb fine movement to amputees.