| 基于离散小波变换和感知频域滤波的语音特征参数 |
| |
| 引用本文： | 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 |
| |
| 作者姓名： | 刘鸣 戴蓓倩 李辉 李霄寒 陆伟 |
| |
| 作者单位： | 中国科学技术大学电子科学与技术系,合肥,230026 |
| |
| 基金项目： | 国家自然科学基金!（69872036） |
| |
| 摘 要： | 为了提高语音识别系统对环境噪声的鲁棒性，本文通过对几种特征参数的噪声鲁棒性的研究，在感知倒谱分析的基础上，引入了多分辨率小波分析技术，利用小波变换的时频局域性并结合感知频域上的滤波技术，提出了一种新型的语音特征参数，该特征参数对于环境噪声具有很强的耐噪性。
|
| 关 键 词： | 语音识别 感知频域滤波 离散小波变换 |
| 文章编号： | 1007-0249(2000)01-0021-05 |
| 修稿时间： | 1999-09-29 |
A New Speech Feature Extracted by Wavelet Analysis & Mel-Frequancy Filtering |
| |
| LIU Ming,DAI Bei-qian,LI Hui,LI Xiao-han,LU Wei. A New Speech Feature Extracted by Wavelet Analysis & Mel-Frequancy Filtering[J]. Journal of Circuits and Systems, 2000, 5(1): 21-25 |
| |
| Authors: | LIU Ming DAI Bei-qian LI Hui LI Xiao-han LU Wei |
| |
| Abstract: | In order to improve the noise robustness of speech recognition systems , the robustness of several kinds of speech features have been studied extensively. Based on Mel-Cepstrum Analysis and the Multi-scale Wavelet Analysis, a new kind of speech feature (DWT-MFC) is discovered. This kind of feature has great immunity to the environmental noise. The experiment results show that the system built on DWT-MFC has better noise robustness than on LPCC and MFCC. |
| |
| Keywords: | Robust speech recognition Noise robustness Mel-cepstrum analysis Multi-scale wavelet analysis. |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
