Electromyogram refinement using muscle synergy based regulation of uncertain information

Electromyogram signal (EMG) measurement frequently experiences uncertainty attributed to issues caused by technical constraints such as cross talk and maximum voluntary contraction. Due to these problems, individual EMGs exhibit uncertainty in representing their corresponding muscle activations. To regulate this uncertainty, we proposed an EMG refinement, which refines EMGs with regulating the contribution redundancy of the signals from EMGs to approximating torques through EMG-driven torque estimation (EDTE) using the muscular skeletal forward dynamic model. To regulate this redundancy, we must consider the synergistic contribution redundancy of muscles, including “unmeasured” muscles, to approximating torques, which primarily causes redundancy of EDTE. To suppress this redundancy, we used the concept of muscle synergy, which is a key concept of analyzing the neurophysiological regulation of contribution redundancy of muscles to exerting torques. Based on this concept, we designed a muscle-synergy-based EDTE as a framework for EMG refinement, which regulates the abovementioned uncertainty of individual EMGs in consideration of unmeasured muscles. In achieving the proposed EMG refinement, the most considerable point is to suppress a large change such as overestimation attributed to enhancement of the contribution of particular muscles to estimating torques. Therefore it is reasonable to refine EMGs by minimizing the change in EMGs. To evaluate this model, we used a Bland-Altman plot, which quantitatively evaluates the proportional bias of refined signals to EMGs. Through this evaluation, we showed that the proposed EDTE minimizes the bias while approximating torques. Therefore this minimization optimally regulates the uncertainty of EMGs and thereby leads to optimal EMG refinement.