Maximum a posteriori estimation of spectral gain with harmonic-structure-based phase reconstruction for phase-aware speech enhancement
Maximum a posteriori estimation of spectral gain with harmonic-structure-based phase reconstruction for phase-aware speech enhancement
复制标题
DOI：
10.23919/apsipa.2018.8659759
复制
发表时间：
2018-11
期刊：
影响因子：
--
通讯作者：
Yukoh Wakabayashi;Nobutaka Ono
中科院分区：
文献类型：
--
作者：
Yukoh Wakabayashi;Nobutaka Ono
与长期以来的看法相反，最近的工作表明，相位谱修改对语音质量和清晰度起着重要的作用。在本文中，我们提出了一种相位感知语音增强方法，它由先前的相位重构方法和基于最大后验准则的相位估计的谱增益估计方法组成。基于相位估计的谱增益估计提高了目标干净语音信号的估计精度。实验结果表明，该方法提高了语音感知质量的客观指标。
Contrary to a long-held belief, recent work shows that phase spectrum modification plays an important role in speech quality and intelligibility. In this paper, we propose a phase-aware speech enhancement method that consists of our previous phase reconstruction method and a spectral gain estimation method using the phase estimate with a maximum a posteriori criterion. The spectral gain estimation based on a phase estimate improves the estimation accuracy of the target clean speech signal. Experimental results demonstrate that the proposed method improves objective measures for speech perceived quality.