A myoelectric control interface compensating for electrode displacements

As a control interface for wearable computing devices, we focus on the use of surface electromyography (i.e., EMG), that have been adopted in several human interface research. However, electrode displacements caused by daily attachment/detachment are known as a drawback of the EMGbased interface, since it makes pattern classification unstable. In this study, we carefully investigated the relationship between the electrode displacements and degradation rate of pattern classification. As the result, we found a characteristic of the electrode displacements. According to the finding, we proposed a solution to the problem, which is based on the idea where a wearable device selects an appropriate classification model from the pre-trained classifiers when a user attaches the device. The availability of the proposed model selection method was also confirmed through experiments.