基于GMM说话人分类的说话人识别方法研究  

Speaker Recognition Research Based on GMM Speaker Clustering Technology

作　　者：赵振东[1] 张静[1] 李圆[1] 胡喜梅[1] 

机构地区：[1]华北电力大学电子与通信工程系,河北保定071003

出　　处：《通信技术》2009年第10期192-193,共2页Communications Technology

摘　　要：提出了基于高斯混合模型(GMM)说话人分类的分级说话人识别系统,同时将小波神经网络(WNN)引入到子识别系统中。分别对未分级说话人识别系统和分级说话人识别系统进行了比较。仿真实验结果表明,分级网络在保证正确识别率的同时,不仅改善了网络训练速度,亦大大提高了识别响应速度。Hierarchical speaker recognition system based on Gauss Mixed Model （GMM） speaker clustering technology is proposed, and the model of wavelet neural network （WNN） is introduced into the sub-recognition system. The hierarchical speaker recognition system is compared with the non-hierarchical system. The simulation results show that the hierarchical speaker recognition system is improved in network training speed and recognition response speed.

关 键 词：说话人识别 小波神经网络 高斯混合模型 

分 类 号：TN392.3[电子电信—物理电子学]