基于BP神经网络的说话人识别技术的实现

说话人识别就是从说话人的一段语音中提取出说话人的个性特征,通过对这些个人特征的分析和识别,从而达到对说话人进行辨认或者确认的目的。神经网络是一种基于非线性理论的分布式并行处理网络模型,具有很强的模式分类能力及对不完全信息的鲁棒性,为说话人识别技术提供了一种独特的方法。BP(Back-propagation Neural Network)是一种非循环多级网络训练算法,有输入层,输出层和N个隐含层组成。首先概述了语音识别技术,介绍了BP神经网络训练过程的7个步骤及其模型,如何建立BP神经网络模型。同时介绍了与其相关的特征参数的提取,神经网络的训练和识别过程,最后,通过编程在Linux系统下实现说话人身份的识别。

The Realization of Speaker Recognition Technology based on BP Neural Network

Speaker recognition is to extract the speaker’s personality traits from vioce of the speaker,to identify or confirm the analysis and identification of these personal characteristics,so as to achieve the right speaker.The neural network is precisely based on the nonlinear theory of distributed parallel processing network model,with a strong pattern classification capability and robustness of incomplete informa-tion,which provides an unique approach to speaker recognition technology.BP(Back-propagation Neural Network) is a non-loop,mul-ti-level network training algorithm,containing input layer,output layer and hidden layer composition.This paper presents an overview of speech recognition technology,describes the seven steps of the BP neural network training process,how to create a BP neural network mod-el,and then discusses the extraction associated with feature parameters,neural network training and recognition process.Finally by pro-gramming in a Linux system,speaker identity recognition is realized.