| 基于改进的小波阈值去噪和二级判断模型的说话人识别 |
| |
| 引用本文： | 李战明,林娟,陈若珠.基于改进的小波阈值去噪和二级判断模型的说话人识别[J].兰州理工大学学报,2010,36(1). |
| |
| 作者姓名： | 李战明 林娟 陈若珠 |
| |
| 作者单位： | 兰州理工大学,电气工程与信息工程学院,甘肃,兰州,730050 |
| |
| 基金项目： | 甘肃省教育厅信息技术专项 |
| |
| 摘 要： | 针对目前说话人识别系统中噪声使得识别率严重下降的问题,在特征提取前用小波阈值去噪方法对带噪语音进行去噪处理.对于小波阈值函数以及阈值的选取进行研究,提出一种改进的阈值函数,小波阈值中的噪音方差估计采用基于实际噪音方差估计的谱熵法,将改进的小波阈值去噪结合一种二级判断模型提高噪声环境下的说话人识别率.在不同信噪比条件下进行试验,改进的小波阈值去噪法优于传统小波阈值去噪方法,改进后的阈值函数与二级判断模型相结合的识别率比原始语音DTW模型提高了7.9%,比原始语音GMM提高了4.6%,对于短时语音有较好的识别率.
|
| 关 键 词： | 说话人识别 小波阈值函数 噪音方差估计 二级判断模型 |
Speaker recognition based on the improved of wavelet threshold denoising and two-stage decision model |
| |
| LI Zhan-ming,LIN Juan,CHEN Ruo-zhu.Speaker recognition based on the improved of wavelet threshold denoising and two-stage decision model[J].Journal of Lanzhou University of Technology,2010,36(1). |
| |
| Authors: | LI Zhan-ming LIN Juan CHEN Ruo-zhu |
| |
| Institution: | College of Electrical and Information Engineering;Lanzhou Univ.of Tech.;Lanzhou 730050;China |
| |
| Abstract: | Aimed at the problem that recognition rate of present speaker recognition system declines seriously due to the noise,using the denoising processing was conducted for the noised speech with wavelet threshold denoising before feature extraction,and the wavelet threshold function and threshold selection were studied,an improved threshold function proposed,and the noise variance estimation of wavelet threshold value calculated with spectral entropy method based on actual noise variance estimation,so that the im... |
| |
| Keywords: | speaker recognition wavelet threshold function noise variance estimation two-stage decision model |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
