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,利用文化算法的全局搜索性能,将文化算法用于语音识别系统的RBF网络的训练过程中,基于实验数据,指出该方法的识别结果较k-均值聚类算法有了明显的改善。
In the light of the disadvantages of K-means clustering algorithm used for obtaining the central vectors of hidden layers in RBF neural network, and by using the global search capability of cultural algorithm, this paper applies culture algorithm into the training process of RBF neural network of speech recognition system, and based on the test data, points out that the recognition results of cultural algorithm have been improved obviously comparing with K-means clustering algorithm.
出处
《科技情报开发与经济》
2011年第10期119-122,共4页
Sci-Tech Information Development & Economy
关键词
文化算法
RBF神经网络
语音识别
cultural algorithm
RBF neural network
speech recognition
