论文部分内容阅读
人的听觉对语音信号中相位的感知比较迟钝,因而对语音信号进行处理和编码时常常不关心相位失真。实际上,相位失真到一定程度时会明显导致语音质量的下降。为了取得高质量的声码器,语音谱分量的相位信息是不能不考虑的。本文通过主观听觉测试实验研究了语音信号的短时Fourier变换相位谱对人的听觉感知的影响。测试结果表明: (1)如果完全舍弃原相位信息,则得到的重建语音含有很强的噪声且自然度很差; (2)不论舍弃高频段还是低频段的相位信息,均能导致听觉感知差异; (3)当相位的量阶小于π/7时,人的听觉系统将分辨不出重建语音和原始语音之间存在的差异.
Human auditory perception of speech signals in the relatively slow, so the speech signal processing and coding often do not care about phase distortion. In fact, the phase distortion will obviously lead to the decline of voice quality. In order to obtain a high quality vocoder, the phase information of speech spectral components can not be ignored. In this paper, subjective auditory experiments were conducted to study the effect of short-time Fourier transform phase spectrum of speech on human auditory perception. The test results show that: (1) If the original phase information is completely discarded, the reconstructed speech will have strong noise and poor naturalness; (2) Whether the phase information of high frequency band or low frequency band is discarded will lead to the difference of auditory perception ; (3) When the magnitude of the phase is less than π / 7, the human auditory system will not be able to tell the difference between reconstructed speech and original speech.