抄録
This paper proposes a novel neural network enabling estimation of a posteriori probability for learned and unlearned classes. By defining probability density functions of unlearned classes in a Gaussian mixture model, undefined classes can be discriminated via network training using given learning samples. This method can be applied to various pattern discrimination problems such as electromyogram (EMG) classification. In the experiments reported here, the classification ability of the proposed network was demonstrated using artificial data and EMG patterns. The results showed that the method provides a high level of performance for learned and unlearned class discrimination.
