摘要
为了克服在Matlab中语音处理工具箱的不足,设计出基于Matlab图形用户界面(GUI)的说话人识别测试平台。系统框架设计:特征参数采用美尔倒谱系数及差分美尔倒谱系数,识别模型采用矢量量化模型;人机交互实现:设置各控件属性,通过回调函数实现系统功能;测试实例:测试人数50人,识别率为96%,表明了该测试平台的有效性。该平台的特点是:功能较为完整,可设置参数,如:人数、字数、帧长等,方便用户实验,具有良好的实用性、可交互性等。
In order to overcome the insufficiency of voice processing tool box in Matlab, a speaker recognition test platform based on graphic user interface (GUI) in Matlab was designed. In the system framework design, Mel-Frequency Cepstral Coeffi- cient and difference Mel-Frequency Cepstral Coefficient were adopted for feature parameters, and vector quantization model was adopted for recognition model. In implementation of human-computer interaction, the CheekBoxs' attribute was set, and system functions were achieved with callback function. 50 persons had been tested, and the recognition ratio is 96%. The result shows the validity of the test platform. The platform has the features of complete function, parameter regularity, such as number of peo- ple, number of words, frame length and so on. It is convenient for the clients to test and has very good applicability and interactivity.
出处
《现代电子技术》
2013年第8期59-60,62,共3页
Modern Electronics Technique
基金
西南交通大学峨眉校区科技发展基金(20090118)