关键词: 神经网络, 特征提取, 目标分类

Abstract: A Modified Fuzzy Counter Propagation（MFCP） neural network,which is a generalized model of the Counter Propagation（CP） network,is proposed in this paper by defining output of the competitive unit of CP network as a fuzzy membership function and combining Fuzzy C-Mean（FCM） algorithm.The problem,automatic generation of fuzzy membership function has been solved in MFCP network.The classification recognition experiment for three different classes of target noises in the sea was extracted feature from line spectrum and continuous spectrum of DEMON,then performed with neural network,and the results show that MFCP network has higher correct recognition rate and better adaptability to untrained targets than CP network or BP network.

Key words: neural network, feature extraction, target classification