动态神经元模型在水下目标识别中的应用  

APPLICATION OF THE CONSTRICTED DYNAMIC NEURAL NETWORKS IN THE CLASSIFICATION OF UNDERWATER TARGETS

作　　者：蔡悦斌[1,2,3] 黄采金[1,2,3] 史习智[1,2,3] 杨毓英[1,2,3] 

机构地区：[1]上海交通大学振动国家重点实验室,200030 [2]上海交通大学冲击国家重点实验室,200030 [3]上海交通大学噪声国家重点实验室,200030

出　　处：《模式识别与人工智能》1998年第2期199-205,共7页Pattern Recognition and Artificial Intelligence

基　　金：国家自然科学基金;中国科学院声学所声场声信息国家重点实验室资助课题

摘　　要：本文提出了两种改进型动态神经元模型—CIIRDNU(Constricted Infinite Impulse Response Dy-namic Neural Unit)和FIRDNU(Finite Impulse Response Dynamic Neural Unit).给出了基于改进型动态神经元模型的神经网络结构,并推导了其BP学习算法公式.将FIRDNU模型应用于水声目标识别中,实验结果表明在此基础上建立的单层前馈网络学习训练速度很快,推广性能良好.其平均正确识别率明显高于静态BP神经网络.In this paper, the two types of constricted dynamic neural units- CIIRDNU(Constrict,ed Infinite Impulse Response Dynamic Neural Unit) and FIRDNU(Finite Impulse Response Dynamic Neural Unit) are proposed. The structure, learning algorithm and implementation of CIIRDNU and FIRDNU networks axe presented. Their applications in the classification of underwater targets are sucessful and the results have shown that FIRDNU networks have a advantage over the static BP neural networks according to the abilities of learning and popularizing.

关 键 词：动态神经元模型 水声目标识别 模式识别 

分 类 号：TB56[交通运输工程—水声工程] TP18[理学—物理]