Abstract
We think that myoelectric prosthesis can become more functional and easier to control by motion discrimination from EMG signals. Many researchers use neural network for motion discrimination. But, their method is difficult to learn in real time. So, we propose a new motion discrimination method using distance scale in feature space for myoelectric prosthesis. This method can learn in real time. Experimental results show that discrimination rate is 80 percents or more in the method using standarization Euclidean distance, or Mahalanobis' generalized distance.
