EEG and artifact classification using a neural network

A multilayer perceptron based classifier is proposed for automatic EEG and artifact classification. Conventionally this task has been carried out by a human expert spending a lot of examination time. For efficient network learning, a preprocessor is designed by which expert knowledge is more effectively utilized. A neural network operating characteristic (NOC) with correct and false classification probabilities is introduced for objective performance evaluation, by which an optimal neural network is constructed. From experiments, the neural-network based classifier performs as well as human experts