一种k均值和神经网络集成的语音识别方法

Speech recognition based on k-means clustering and neural network ensemble

提出了一种基于k均值聚类和BP神经网络集成的语音识别方法,该方法以神经网络集成模型为基础,利用k均值聚类算法选择部分有差异性的个体神经网络再进行集成学习,既克服了单个BP网络模型容易局部收敛和不稳定性的缺点,又解决了传统集成方法训练时间长和个体网络差异性不明显的问题。通过对非特定人孤立词的语音识别的实验,证实了该方法的有效性。

In this paper,a method of speech recognition based on k-means clustering and neural network ensemble is proposed.The method is based on neural network model.After a number of individual neural networks are trained,the k-means clustering algorithm is used to select a part of the trained individual networks＇weights and thresholds with small similarity.Many neural networks with the selected weights and thresholds are combined.The method not only overcomes the shortcomings that single BP neural network model is easy to local convergence and lack of stability,but also solves the problems that the traditional method in training lasts for a long time and the differences of individual network are not obvious.The experimental results prove the effectiveness of this method.