|
|||||
|
|
| 一种模糊高斯混合说话人识别模型 | |
| 引用本文： | 王金明,张雄伟. 一种模糊高斯混合说话人识别模型[J]. 解放军理工大学学报(自然科学版), 2006, 7(3): 214-219 |
| 作者姓名： | 王金明 张雄伟 |
| 作者单位： | 解放军理工大学,通信工程学院,江苏,南京,210007;解放军理工大学,通信工程学院,江苏,南京,210007 |
| 摘 要： | 为了研究模糊聚类算法在高斯混合模型（GMM）参数获取方面的应用,采用模糊C均值算法（FCM）进行语音特征矢量的聚类,并结合Tabu搜索算法得到全局最优的聚类结果,进一步用EM算法得到GMM模型参数.使用TIMIT数据库中的语音进行测试,开集和闭集说话人辨认实验都表明,该方法获取的GMM参数比普通EM算法获得的GMM模型参数性能更优,能有效降低说话人辨认系统的误识率. |
| 关 键 词： | 高斯混合模型 说话人识别 模糊C均值算法 Tabu搜索算法 Mel倒谱系数 |
| 文章编号： | 1009-3443（2006）03-0214-06 |
| 收稿时间： | 2005-05-10 |
| 修稿时间： | 2005-05-10 |
Fuzzy Gaussion mixture models for speaker recognition |
|
| WANG Jin-ming and ZHANG Xiong-wei. Fuzzy Gaussion mixture models for speaker recognition[J]. Journal of PLA University of Science and Technology(Natural Science Edition), 2006, 7(3): 214-219 | |
| Authors: | WANG Jin-ming and ZHANG Xiong-wei |
| Affiliation: | Institute of Communications Engineering, PLA Univ. of Sci. &. Tech. .Nanjing 210007. China;Institute of Communications Engineering, PLA Univ. of Sci. &. Tech. .Nanjing 210007. China |
| Abstract: | Expectation maximization (EM) algorithm is usually used to estimate parameters of Gaussion mixture models (GMM). The method using fuzzy clustering algorithm to get GMM parameters was researched, the tabu search algorithm was fused to estimate more accurate parameters, and the training process of GMM was optimized. Closed set test and open set test show that this method performs better compared with EM algorithm, and can obviously reduce the recognition inaccuracy of the speaker identification system. |
| Keywords: | GMM (Gaussion mixture models) speaker recognition FCM (fuzzy C-means) algorithm Tabu search algorithm MFCCXmel-frequency cepstrum coefficients) |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ | |
| 点击此处可从《解放军理工大学学报(自然科学版)》浏览原始摘要信息 | |
| 点击此处可从《解放军理工大学学报(自然科学版)》下载全文 | |