| 一种改进的基于GMM-UBM的法庭自动说话人识别系统 |
| |
| 作者姓名： | 王华朋 杨军 吴鸣 许勇 |
| |
| 作者单位： | 1. 中国科学院声学研究所噪声与振动重点实验室, 北京 100190;
2. 中国刑事警察学院, 沈阳 110854 |
| |
| 基金项目： | 国家自然科学基金(11004217,11074279)资助 |
| |
| 摘 要： | 对基于高斯混合模型(GMM)的法庭自动说话人识别系统进行改进.通过参考人群数据库降低了对嫌疑人语音样本数量的需求.以小规模背景人群数据库建立改进的基于高斯混合模型-通用背景模型(GMM-UBM)的法庭自动说话人识别系统.以固定电话信道和移动手机信道的数据库进行了系统的测试.
|
| 关 键 词： | 似然比 法庭自动说话人识别 高斯混合模型-通用背景模型 |
| 收稿时间： | 2012-10-22 |
| 修稿时间： | 2013-03-22 |
A forensic automatic speaker recognition method based on improved GMM-UBM |
| |
| Authors: | WANG Hua-Peng YANG Jun WU Ming XU Yong |
| |
| Institution: | 1. Key Laboratory of Noise and Vibration Research, Institute of Acoustics, Chinese Academy of Sciences, Beijing 100190, China;
2. Department of Forensic Science and Technology, China Criminal Police University, Shenyang 110854, China |
| |
| Abstract: | We improved the forensic automatic speaker recognition(FASR) system based on GMM by applying the reference database to reduce the demands of suspect's recording quantity. We established an improved FASR system based on GMM-UBM, in which a small background population is used. The proposed system was tested in the database of fixed telephone channel and mobile channel, respectively. |
| |
| Keywords: | likelihood ratio forensic automatic speaker recognition GMM-UBM |
|
| 点击此处可从《》浏览原始摘要信息 |
| 点击此处可从《》下载免费的PDF全文 |
