The evaluation of the discriminant ability of multiclass SVM in a study of hand motion recognition by using SEMG

Electromyogram (EMG) is a kind of biological signal that is generated because of excitement of muscle according to the motor instruction from a brain. We have been experimentally developing the hand motion recognition system by using 4 channels forearm EMG signals. In our system, in order to classify measured EMG SVM (Support Vector Machine) that has higher discriminability is used. Often SVM is used as a non-linear classifier. But, In the conventional system that we developed, we used a canonical discriminant analysis (CDA) method. CDA method is linear discriminant function, but it has shown good experimental results. Therefore, we have compared the discriminant ability between SVM and CDA. In this report, we will describe about the results of this experiment.