The present paper proposes a new probabilistic neural network (NN) that can estimate a posteriori probability for a pattern classification problem. The structure of the proposed network is based on a statistical model composed by a mixture of log-linearized Gaussian components. However, the forward calculation and the backward learning rule can be defined in the same manner as the error backpropagation NN. In this paper, the proposed network is applied to the electroencephalogram (EEG) pattern classification problem. In the experiments, two types of a photic stimulation, which are caused by eye opening/closing and artificial light, are used to collect the data to be classified. It is shown that the EEG signals can be classified successfully and that the classification rates change depending on the number of training data and the dimension of the feature vectors.

The present paper proposes a method to estimate the motion intended by a subject from his EMG signals using error back propagation typed neural networks. Estimation of the motion from the EMG signals is useful for means of human interface in such fields as control of multi-functional powered prosthesis, teleoperation of robot manipulators, virtual reality. The neural network used in the method can learn a mapping from the EMG patterns measured from four pairs of electrodes to six motions of forearm and hand intended by the subject. The experimental results for several subjects including an amputee show the following: 1) the method can discriminate six motions with the accuracy about 90 percent for several electrode locations, 2) ill-discrimination can be decreased by suspending discrimination using entropy of network output, 3) the neural network can adapt to some dynamic variations of the EMG patterns using on-line learning, and 4) utilizing frequency characteristics as well as amplitude characteristics of the EMG signals reduces the number of iterations required for learning convergence.

No abstract