基于分段PCA的BP神经网络语音识别

BP Neural Network For Speech Recognition Based on Segmented PCA

本文提出采用分段PCA的方法提取信号的特征,再通过BP网络完成信号的训练及识别。仿真实验探究了识别网络的输入维数以及隐含层数目对识别率的影响,通过恰当的选择网络的输入维数及隐含层数,实验结果验证了使用分段PCA方法后,语音识别率会有所提高。

This paper puts forward that using segmented PCA method to extract the signal feature. After that, signal is trained and recognized using BP network. At the end of this article the simulation experiment explores the influence of input dimension and the number of hidden layer to the recognition rate. Choose the input dimension and the number of hidden layer properly, the simulation experiment result verifies that it will improve the speech recognition rate.