Forearm motion discrimination technique using real-time EMG signals

The objective of this study is to develop a method of discriminating real-time motion from electromyogram (EMG) signals. We previously proposed a motion discrimination method. This method could discriminate five motions (hand opening, hand closing, hand chucking, wrist extension, and wrist flexion) at a rate of above 90 percent from four channel EMG signals in the forearm. The method prevents elbow motions from interfering with hand motion discrimination. However, discrimination processing time of this method is more than 300 ms, and the shortest delay time that is perceivable by the user is generally regarded to be roughly 300 ms. Furthermore, a robot hand has a mechanical delay time. Thus, the discrimination time should be less than 300 ms. Here, we propose a real-time motion discrimination method using a hyper-sphere model. In comparison with the old model, the hyper-sphere models can make more complex decision regions which can discriminate at the state of the motion. Furthermore, this model can learn EMG signals in real-time. We experimentally verified that the discrimination accuracies of this method were above 90 percent. Moreover, elbow motions did not interfere with the hand motion discrimination. The discrimination processing time was less than 300 ms, and was about 30 percent shorter than that of the old method.