首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于输出的语音质量客观评价方法研究
引用本文:陈国,胡修林,杨灵,朱耀庭.基于输出的语音质量客观评价方法研究[J].华中科技大学学报(自然科学版),2001,29(5):86-88.
作者姓名:陈国  胡修林  杨灵  朱耀庭
作者单位:华中科技大学电子与信息工程系
基金项目:中国人民解放军总装备部科研基金资助项目,华中科技大学研究生基金资助项目
摘    要:在分析语谱图基础上提出了一种基于输出的语音质量客观评价方法-度量语亦图密度分布特征MDCSD(Measuring Distribution Characteristics of Spectrogram Density) 方法。该方法通过计算功率谱二维联合概率分布特征来描述不同质量语音信号的语谱图特征,并以此来对语音质量进行了客观评价,实验结果表明该方法是一种性能良好的基于输出的客观评价方法,其评价结果与主观评价的相关度达到0.8左右。

关 键 词:语谱图  语音质量  客观评价  质量语谱图密度分布特征法  特征参量
文章编号:1000-8616(2001)05-0086-03
修稿时间:2000年12月29

Output-based Objective Measure of Speech Quality
Chen Guo,Hu Xiulin,Yang Ling,Zhu Yaoting.Output-based Objective Measure of Speech Quality[J].JOURNAL OF HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY.NATURE SCIENCE,2001,29(5):86-88.
Authors:Chen Guo  Hu Xiulin  Yang Ling  Zhu Yaoting
Abstract:The measuring Distribution Characteristics of Spectrogram Density(MDCSD) method, based on the analysis of spectrogram, is presented as an output based objective measure of speech quality. The feature parameters, indicating the characteristics of spectrogram of speech signals with different quality, are extracted in the MDCSD method by calculating the two dimensions associated with probability distributions of power spectrum in spectrogram. The speech quality is evaluated based on these feature parameters. The experimental results show that the MDCSD method is a preferable output based objective measure of speech quality which correlative degree with subjective measure can reach around 0.8.
Keywords:spectrogram  speech quality  objective measure
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号