| 说话人识别模型失配下的似然得分补偿变换 |
| |
| 引用本文： | 包永强,赵力,邹采荣. 说话人识别模型失配下的似然得分补偿变换[J]. 电路与系统学报, 2006, 11(4): 51-55 |
| |
| 作者姓名： | 包永强 赵力 邹采荣 |
| |
| 作者单位： | 1. 南京工程学院,通信系,江苏,南京211167
2. 东南大学,无线电工程系,江苏,南京,210096 |
| |
| 基金项目： | 国家自然基金（N0：60472058）,教育部博士点基金（N0：20050286001）和教育部“新世纪优秀人才支持计划” |
| |
| 摘 要： | 基于与文本无关说话人识别最常用的模型一一高斯混合模型（GMM）的输出帧似然概率的统计特性，提出了一种非线性变换方法一一似然得分补偿法。理论分析和实验结果表明：与常用的最大似然（ML）变换相比，该方法可降低误识率达20％。结果还表明：似然得分补偿法基本克服了在与文本无关说话人识别系统中，当说话人的个性特征不断变化、环境对系统识别构成影响从而导致识别模型失配情况下，需要对模型的得分进行补偿的局限。
|
| 关 键 词： | 与文本无关说话人识别 混合高斯模型 似然得分补偿变换 |
| 文章编号： | 1007-0249（2006）04-0050-05 |
| 收稿时间： | 2004-08-12 |
| 修稿时间： | 2004-11-16 |
Speaker recognition likelihood score compensation transformation under model mismatch |
| |
| Bao Yong-qiang,Zhao Li,Zou Cai-rong. Speaker recognition likelihood score compensation transformation under model mismatch[J]. Journal of Circuits and Systems, 2006, 11(4): 51-55 |
| |
| Authors: | Bao Yong-qiang Zhao Li Zou Cai-rong |
| |
| Abstract: | Based on the statistic characteristic of frame likelihood probability output by Gaussian mixture model(GMM)which is the best text-independent speaker recognition model,likelihood score compensation transformation as a non-linear transform method is presented.The theoretical analysis and experiment shows that it could reduce the error recognition ratio 20% as compared with Maximum-Likelihood(ML)transformation.The result shows that likelihood score compensation transformation should be adopted for canceling the influence of variations in speech characteristics,noise and model mismatch.The result also shows that process on frame likelihood probability output by GMM is effectual way of decreasing the influence of noise and improving the recognition ratio. |
| |
| Keywords: | text-independent speaker recognition Gaussian mixture model likelihood score compensation transformation |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
