| 基于不变集多小波的语音特征参数提取 |
| |
| 引用本文： | 莫家玲,胡维平.基于不变集多小波的语音特征参数提取[J].电声技术,2009,33(7):63-67. |
| |
| 作者姓名： | 莫家玲 胡维平 |
| |
| 作者单位： | 1. 梧州学院,计算机与电子信息工程系,广西,梧州,543002
2. 广西师范大学,物理与电子工程学院,广西,桂林,541004 |
| |
| 基金项目： | 广西自然科学基金，梧州学院院级项目 |
| |
| 摘 要： | 在研究不变集多小波理论的基础上，借鉴Mel频率倒谱系数（MFCC）的提取算法，用多小波交换代替傅里叶变换及Mel滤波．构造了一种新的语音特征参数MWBC。汉语数字识别实验结果表明，提出的新语音特征参数MWBC的识别性能和抗噪性能均优于MFCC，为提高语音识别系统的噪声鲁棒性提供了一条新途径。
|
| 关 键 词： | 语音识别 不变集多小波 特征参数 Mel频率倒谱系数 |
Speech Features Extraction Based on Invariant Sets Multi-wavelet |
| |
| MO Jia-ling,HU Wei-ping.Speech Features Extraction Based on Invariant Sets Multi-wavelet[J].Audio Engineering,2009,33(7):63-67. |
| |
| Authors: | MO Jia-ling HU Wei-ping |
| |
| Affiliation: | MO Jia-ling1,HU Wei-ping2(1.Department of Computer and Electronic Information Engineering,Wuzhou University,Wuzhou Guangxi 543002,China,2.College of Physics and Electronic Engineering,Guangxi Normal University,Guilin Guangxi 541004,China) |
| |
| Abstract: | According to a research in invariant sets multi-wavelet theory and denining approach of MFCC(Mel-Frequency Cepstrum Coefficient),the multi-wavelet transform is used to replace the short-time Fourier transform and Mel filter,and then a new speech feature parameters MWBC is proposed.The Chinese numeral recognition experiment results indicate that the recognition performance and anti-noise performance of the new feature are better than MFCC.A new way is provided to improve the noise robustness of speech recogn... |
| |
| Keywords: | speech recognition invariant sets multi-wavelet features parameters MFCC |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
