Emg pattern classification using hierarchical network based on boosting approach

Abstract. This paper proposes a new electromyogram (EMG) pattern classification method using probabilistic neuralnetworks based on boosting approach [1]. Since the pro-posed method automatically constructs a suitable classification network from measured EMG signals, there is no need to set the structure of network in advance. To verify the feasibility of the proposed method, phoneme classification experiments are conducted using EMG signals measured from mimetic and cervical muscles. In these experiments, the proposed method achieved high classification rates