Design of a Neural Network for Recognizing Noised Patterns

잡음이 혼입된 패턴인식을 위한 신경망의 설계

In this paper, a novel method which can recognize noised patterns by using a neural network is proposed. Patern recognition is processed by two networks and the processing mechanism is partially based on the results of neurophysiological studies. Synaptic activitoes of combining neurons between layers are implemented by spatial filters which approximate receptive fields of the modelled optic nerve cells.