| 针对说话人识别进行的加权小波去噪方法 |
| |
| 引用本文： | 邱政权,尹俊勋.针对说话人识别进行的加权小波去噪方法[J].电声技术,2005(9):53-56. |
| |
| 作者姓名： | 邱政权 尹俊勋 |
| |
| 作者单位： | 华南理工大学,电信学院,广东,广州,510640;华南理工大学,电信学院,广东,广州,510640 |
| |
| 摘 要： | 采用改进的小波去噪方法对含噪语音进行了前端处理，并针对说话人识别的特点，在小波重构之前对各小波系数进行加权处理；识别过程采用GMM识别算法。实验结果显示，相比纯粹使用MFCC作为识别特征的说话人识别提出的方法对于含噪说话人识别有明显的优越性。该方法对实时说话人识别有很好的指导作用。
|
| 关 键 词： | 说话人识别 小波变换 DWT-MFCC GMM |
| 文章编号： | 1002-8684(2005)09-0053-03 |
| 收稿时间： | 2005-05-17 |
| 修稿时间： | 2005年5月17日 |
Speaker Recognition Applied with Wavelet Transform |
| |
| QIU Zheng-quan,YIN Jun-xun.Speaker Recognition Applied with Wavelet Transform[J].Audio Engineering,2005(9):53-56. |
| |
| Authors: | QIU Zheng-quan YIN Jun-xun |
| |
| Abstract: | This paper adopts a novel wavelet de-noising method for front-end processing of speech with noise. Moreover, according to the character of speaker recognition, the wavelet coefficient is weighted before the reconstruction of the wavelet. In the process of recognition, GMM recognition arithmetic is adopted. The experiment shows that the proposed method has more advantages for speaker recognition in noisy circumstance than that with simple MFCC. The proposed method provides very good guidance for real time speaker recognition. |
| |
| Keywords: | DWT-MFCC GMM |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
