一种基于听觉特性的语音失真测度方法

详细信息

A speech distortion measure based on auditory properties

Department of Electronics & Information Engineering, Huazhong University of Science & Technology Wuhan 430074

摘要: 提出了一种基于听觉特性的语音失真测度方法——感知谱失真 PSD(Perceptual Spectrum Distortion)测度,该测度方法通过模拟人的听觉特性把语音短时频谱转变为符合听觉特性的感知频谱,再以感知谱为基础来度量语音失真程度。经过对不同质量的语音进行仿真实验以及与Itakura测度方法作对比实验,结果表明PSD测度是一种与语音质量主观评价一致性较好的语音失真测度方法。

Abstract: The Perceptual Spectrum Distortion (PSD) measure,based on auditory properties of human being,is presented as a speech distortion measure.The PSD measure calculates the speech distortion distance by simulating the auditory properties of human being and converting short-time speech power spectrum to auditory perceptual spectrum.Preliminary simulative experiments in comparison with the Itakura measure have been done.The results show that the PSD measure is a preferable speech distortion measure that can be consistent with subjective assessment of speech quality well.