Japanese Janken recognition by support vector machine based on electromyogram of wrist

Recent years, biosignal is receiving attention as a tool of human interface. Above all, electromyogram (EMG) has already applied to many researches. In this study, we propose a method which can discriminate hand motions. We measured an electromyogram of wrist by using 8 dry type sensors. We focused on four motions, such as "Rock-Scissors-Paper" and "Neutral". "Neutral" is a state that does not do anything. In the proposed method, we apply fast Fourier transformation (FFT) to measured EMG data, and then remove hum noise. Next, we combine values of sensors based on a Gaussian function. In this Gaussian function, variance and mean are 0.2 and 0, respectively. After that, we apply normalization by linear transformation to the values. Subsequently, we resize the values into range from -1 to 1. Finally, support vector machine (SVM) conducts learning and discrimination. We conducted experiments in three subjects. Discrimination accuracy of the proposed method for three subjects was 96.9%, 95.3%, 92.2%, respectively. Therefore, we think that the Gaussian function is robust to difference of sensor position because this function combines both adjacent channels. In the previous method, the discrimination accuracy rate was 77.1%. Therefore, the proposed method is better in accuracy than the previous method. In future work, we will conduct an experiment which discriminates Japanese Janken of a subject who is not learned.