| 易混淆语音特征提取方法的研究 |
| |
| 引用本文： | 武玉峰,张玲华,颜永红.易混淆语音特征提取方法的研究[J].南京邮电大学学报(自然科学版),2010,30(2). |
| |
| 作者姓名： | 武玉峰 张玲华 颜永红 |
| |
| 作者单位： | 1. 南京邮电大学,通信与信息工程学院,江苏,南京,210003
2. 中国科学研究院,声学研究所,北京,100190 |
| |
| 摘 要： | 在语音识别系统中,易混淆语音是导致系统识别率下降的重要原因。汉语音节是由声母和韵母组成的,在易混淆语音中,其韵母部分的混淆度很大。针对易混淆语音的韵母部分,通过改进特征提取的方法来提高易混韵母之间的区分度,提出了一种基于小波分解和线性预测(WLPC)的特征提取方法,并用局部保持映射(Locality Preserving Projections)算法对提取的特征进行了特征变换。实验结果显示,与传统的MFCC特征相比,该特征能更好的区分不同的韵母。
|
| 关 键 词： | 小波变换 局部保持映射 易混淆语音 |
A Study of Feature Extraction Method for Confusable Speech |
| |
| WU Yu-feng,ZHANG Ling-hua,YAN Yong-hong.A Study of Feature Extraction Method for Confusable Speech[J].Journal of Nanjing University of Posts and Telecommunications,2010,30(2). |
| |
| Authors: | WU Yu-feng ZHANG Ling-hua YAN Yong-hong |
| |
| Institution: | 1.College of Telecommunications & Information Engineering/a>;Nanjing University of Posts and Telecommunications/a>;Nanjing 210003/a>;China 2.Institute of Acoustics/a>;Chinese Academy of Sciences/a>;Beijing 100190/a>;China |
| |
| Abstract: | In automatic speech recognition(ASR) systems,the existence of confusable speech is one important factor decreasing the recognition rate.One Chinese syllable is consisted of consonant and vowel,and the confusion degree of the vowel is very large in the confusable syllables.A novel approach to feature extraction using Discrete Wavelet Transform(DWT) and Linear Prediction Coefficients(LPC) for the vowel part of confusable speech is presented in this paper.Locality Preserving Projections(LPP) algorithm based tr... |
| |
| Keywords: | discrete wavelet transform locality preserving projections confusable speech |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
