用遗传算法改进的RBF神经网络在语音识别系统中的应用  

The Application of Improved RBF Neural Network Based on Genetic Algorithm in Speech Recognition System

作　　者：李高云[1] 张雪英[1] 

机构地区：[1]太原理工大学信息工程学院,山西太原030024

出　　处：《太原理工大学学报》2009年第4期345-347,352,共4页Journal of Taiyuan University of Technology

基　　金：国家自然科学基金(60472094);山西省自然科学基金(20051039)

摘　　要：针对传统的RBF网络求取隐层基函数中心的K-均值聚类算法的缺点,利用遗传算法的全局搜索性能,将遗传算法用于RBF网络的训练过程。由于简单遗传算法存在收敛速度慢及不能收敛到全局最优解等不足,引入自然数编码的自适应遗传算法,加快网络的训练速度。将该算法应用到语音识别系统中,实验结果表明:使用该方法的识别结果优于使用K-均值聚类算法选取质心的RBF网络的识别结果。This paper trained RBF neural network using genetic algorithms, which has global search characteristic, in order to overcome shortage of traditional K-mean clustering algorithm applied in learning the central vectors of hidden layers in RBF neural network. It introduced natu- ral number coding adaptive genetic algorithms to speed up training speed of the network because simple genetic algorithm has the disadvantages of slow convergence speed and being unable to converge to the optimum solution. The experimental results show that the recognition results of improved algorithm are better than those of RBF network using K--mean clustering algorithm in learning the central vectors of hidden layers.

关 键 词：自适应遗传算法 RBF神经网络 语音识别 

分 类 号：TN912.3[电子电信—通信与信息系统]