摘要
提出了一种 HMM框架下的神经网络分类器 ,它既克服了普通神经网络不能有效地识别时变信号的缺点 ,又解决了 HMM识别时变信号时不能突出不同信号的差异性问题 .用网络权的遗传算法进化学习解决了 Baum-Welch及 BP网络学习中易陷入局部极小点的问题 .
A neural network classifier based on HMM framework was proposed.It can be used to identify timevarying signal, which an ordinary neural network lacks of, and can stress the differences of different signals. Evolutionary learning of the neural network weights using genetic algorithm solves the problem of falling into local minimum point which BP and Baum Welch algorithmnace. An example of recognizing radar return signal successfully by the neural netwrok was presented.
出处
《红外与毫米波学报》
SCIE
EI
CAS
CSCD
北大核心
2001年第2期107-110,共4页
Journal of Infrared and Millimeter Waves
