Monte Carlo method for evaluating the effect of surface EMG measurement placement on motion recognition accuracy.

Surface electromyogram (SEMG) is one of the most important biological signal in which the human motion intention is directly reflected. Many systems use SEMG as a source of a control signal. (We call them "SEMG system"). In order to develop SEMG system, constructions of discriminant function and SEMG measurement placement are important factors for accurate recognition. But standard criterions for selection of discriminant function and SEMG measurement placement have not been clearly defined. Almost all of the conventional SEMG system has decided to select measurement placements of SEMG according to standard general anatomical structure of the human body and that mainly focused on signal processing method. However, SEMG measurement placement is also critical for recognition accuracy and evaluating the effect of SEMG measurement placement is important. In this study, we investigate the effect of SEMG measurement placement in hand motion recognition accuracy. We use a 96-channels matrix-type surface multielectrode and four channels are selected as the SEMG measurement placements from the channels that compose multielectrode. 5,000 configurations of SEMG measurement placements are generated by randomly selected number and each configuration is assessed by motion recognition accuracy (i.e. Monte Carlo method). In order to consider the influence of discriminant analysis, our system employs the linear discriminant analysis and nonlinear discriminant analysis. Each selected SEMG measurement placement is evaluated by those two types of discriminant analysis and the results are compared with each other. The experimental results show that motion recognition accuracy differs between these two analyses even if the same SEMG measurement placement is used. Not all optimal measurement placements for linear discriminant function suit for nonlinear discriminant function. The outcome of these investigations, the SEMG measurement placement should be taken into consideration and it suggests the necessity of evaluating the optimal measurement placement depending on a discernment analysis.