基于深度学习神经网络的孤立词语音识别.DOC

基于深度学习神经网络的孤立词语音识别的研究*( 王山海，景新幸，杨海燕+ （桂林电子科技大学 信息与通信学院，广西 桂林 541004） 摘要：为了提高语音识别系统性能，研究提出将自编码器深度学习神经网络应用于语音识别中。该网络结构引入贪婪逐层预训练学习算法，通过预训练和微调两个步骤，提取出待识别语音信号的本质特征，克服传统多层人工神经网络模型在训练时存在易陷入局部极小值且需要大量标签数据的问题。然后经过规整网络，将任意长度帧的语音特征参数规整到某一特定帧，输入到分类器中进行语音识别。对反向传播神经网络和自编码神经网络分别进行了仿真实验，结果表明深度学习神经网络识别准确度较传统神经网络提升了26.1%，是一种优良的语音识别模型。 关键字：语音识别；人工神经网络；深度学习；自编码器；规整网络 中图分类号:TN912 Study of isolated speech recognition based on deep learning neural networks WANG Shan-hai, JING Xin-xing, YANG Hai-yan (School of Information and Communication, Guilin University of Electronic Technology, Guilin Guangxi 541004, China) Abstract: To improve the?performance of the conventional?speech recognition system, the research introduced the autoencoder deep learning neural networks which was applied to speech recognition. The neural networks based on deep learning introduced greedy layer-wise learning algorithm by pretraining and fine-tuning. It could extract the essential features of speech signal which was needed to recognition. It could overcome?the shortcomings of the conventional multilayer artificial neural networks which easily trapped into local optimum when training the model. And they need a large number of labeled data. Then the structured alignment networks could align arbitrary frames of features to fixed frames. And these features were input to a classifier to speech recognition. This paper did some experiment with back propagation neural networks and autoencoder neural networks respectively. The results illustrate that the deep learning neural networks can outperform the conventional neural networks by 26.1% in accuracy. It is an excellent speech recognition model. Key Words: speech recognition; artificial neural networks; deep learning; autoencoder; alignment networks 引言 语音识别是为了让计算机理解人类语言的命令，达到人机交互智能化的目的。随着计算机技术的不断发展和广泛应用，语音识别技术得到了迅速发展，语音识别的研究越来越受到人们的关注。 目前常用的识别方法主要有：动态时间规整（Dynamic Time Warping, DTW）方法、 矢量量化（Vector Quantization, VQ）方法和隐马尔科夫模型（Hidden Markov Model, H