本文提出一种基于小波分析的大词汇汉语连续语音识别的方法,即采用一维小波变换将原始语音信号进行五层小波分解,然后对各层小波系数进行重构,得到五层语音信号,分别对各层语音信号进行训练,得到各层的声学模型,然后结合语言模型对各层声学模型的性能进行测试。通过对纯净语音和带噪语音的各层重构语音数据进行测试。结果表明对于含有高斯白噪声的带噪语音,该方法能使系统性能有所提高,但对于粉红噪声,该方法效果不明显。对于含有真实环境噪声的带噪语音,该方法能获得比基线系统更好的性能。
Abstract
In this paper wavelet decomposition is used to decompose speech signal into five levels. The wavelet coefficients of each part were reconstructed. Because different frequencies of the speech signal have different influence on the performance of the system , the acoustic model of each level was trained and tested. The experimental results show that the method of this paper is effective on gauss white noise and real environmental noise. However it is not effective on pink noise.
关键词
计算机应用 /
中文信息处理 /
大词汇连续语音识别 /
小波分析 /
声学模型
{{custom_keyword}} /
Key words
computer application /
Chinese information processing /
large vocabulary continuous speech recognition /
wavelet analysis /
acoustic model
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 刘鸣,戴蓓倩,李辉,李霄寒,陆伟. 基于离散小波变换和感知频域滤波的语音特征参数[J] . 电路与系统学报,2000 ,5 (1) :21 - 25.
[2] Tufekci Z. and Gowdy J. N , Feature Extraction Using Discrete Wavelet Transformfor Speech Recognition[A] . Southeastcon 2000. In : Proceedings of the IEEE[C] . 7 - 9 April 2000 ,116 - 123.
[3] Yuri Romanyshyn , Wavelet Transforms Applications for Speech Signals Processing[A] . CADSM * 2001 Proceedings [C] ,297 - 298
[4] C. J. Long , S. Datta. Wavelet Based Feature Extraction for Phoneme Recognition[A] . Spoken Language , 1996. ICSLP 96. Proceedings. , Fourth International Conference on , Volume :1 ,3 - 6 ,Oct. 1996 ,Pages :264 - 267 vol. 1.
[5] MallatS. Theory for multi-resolution signal decomposition : The wavelet representation[J] . IEEE Transactions on Pattern Analysis and Machine Intelligence ,1989 ,11 (7) :674 - 693.
[6] Mallat S ,Hwang WL. Singularity detection and processing with wavelets[J] . IEEE Transaction on Information Theory , 1992 ,38 (2) :617 - 643.
[7] Donoho DL. Adapting to unknown smoothness via wavelet shrinkage [J] . J Amer Statist Assoc , 1995 , 90 : 1200 - 1224.
[8] Coifman RR , Donoho DL. Translation-invariant denoising , wavelets and statistics[M] . NewYork : Springer - Verlag ,1995. 125 - 150.
[9] Goodman T N T, Lee S L. Wavelets of multiplicity [J] . Trans of Amer Math Soc ,1994 ,342 (1) :307 - 324.
[10] 胡昌华. 基于MATLAB的系统分析与设计:小波分析[M] . 西安:西安电子科技大学出版社,1999.
[11] 张欣研,等. 基于子带信息的鲁棒语音特征提取框架[J] . 中文信息学报,16 (1) :19 - 24.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
基金
教育部跨世纪人才基金资助项目;教育部重点项目(02029)
{{custom_fund}}
PDF(715 KB)
