基于遗传矢量量化和遗传神经网络的说话人识别系统  

Speaker recognition using genetic vector quantization and genetic neural network

作　　者：白玉[1] 陈立伟[1] 

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出　　处：《应用科技》2005年第12期45-47,50,共4页Applied Science and Technology

基　　金：黑龙江省自然科学基金资助项目(F2004-09)

摘　　要：提出一种基于遗传神经网络的说话人识别系统.将遗传算法和矢量量化技术结合建立说话人模型,然后利用遗传神经网络进行识别.实验结果表明,这种方法既降低了用户的语音数据采集量,有利于话者模板的建立,又提高了系统的识别性能及鲁棒性,较传统方法有明显的优越性.A speaker recognition system using genetic neural network is brought forward. Genetic algorithm and vector quantization technique are combined to constitute speaker identification model, and then genetic neural network is used for recognition. Experimental results show this method can not only reduce the quantity of the speech data collection, in favor of speaker＇s model, but also improve the recognition performance and robustness, setting up showing obvious superiority over traditional methods.

关 键 词：遗传神经网络 矢量量化 说话人识别 鲁棒性 

分 类 号：TN911.7[电子电信—通信与信息系统]