摘要
语音识别作为人工智能研究中不可或缺的一部分已经逐渐渗透到人们的日常生活中。针对传统语音识别方法不能很好地实现并识别复杂多变、非特定人语音的问题,文中提出利用在时间序列上关联性较强的循环神经网络(RNN)建立语音识别模型。考虑到语音信号丰富的时频信息表达,在特征提取环节进行改进,利用具有较好时频分辨率的小波变换(WT)取代快速傅里叶变换(FFT)作为该模型的输入;然后,采用随时间展开的反向传播算法(BPTT)进行特征学习与训练。在实验测试中,首先,对比分析了基于小波变换的特征提取对识别效果的影响;其次,通过与传统的HMM模型及BP神经网络的识别率做对比,验证RNN神经网络可提高语音识别准确率和稳定性。
Speech recognition as an indispensable part of artificial intelligence research has gradually penetrated into peo ple's daily live. In allusion to the problems that the traditional method of speech recognition can not properly identify the com plex and non specific speech,establishing a speech recognition model based on recurrent neural network(R NN) with strong cor relation in time series is propose in this paper. In consideration of the abundant time frequency information of speech signal,the feature extraction process is improved,in which the wavelet transform(W T) with better time frequency resolution is used as the input of the model to replace the fast Fourier transform(FFT). The back propagation time algorithm(B PTT) expanding with time is adopted to conduct the feature learning and training. In the experiment test,the contrastive analysis on the influence of the feature extraction based on wavelet transform on recognition effect was carried out,and the recognition rate of the speech recognition model proposed in this paper was compared with that of the traditional HMM model and BP neural network. By the above measures,the RNN neural network is proved that its accuracy of speech recognition rate and the stability of the recogni tion are improved to a certain extent.
作者
唐美丽
胡琼
马廷淮
TANG Meili;HU Qiong;MA Tinghuai(Nanjing University of Information Science & Technology,Nanjing 210044,China)
出处
《现代电子技术》
北大核心
2019年第14期152-156,共5页
Modern Electronics Technique
基金
国家自然科学基金资助项目(61572259)~~
关键词
语音识别
循环神经网络
反向传播算法
特征提取
小波变换
HMM模型
BP神经网络
speech recognition
recurrent neural network
back propagation algorithm
feature extraction
wavelet trans form
H MM model
B P network
