摘要： 提出以模拟听觉感知模型的非均匀滤波器组为基础的多通道语音增强系统，与基于均匀滤波器组的语音增强系统相比，该方法达到相同频率分辨率所需的通道数较少。采用Itakura-Saito距离对系统进行客观评价，仿真结果表明，该系统增强后的语音比均匀多通道系统增强后的语音具有更好的改善效果。

关键词: 全通变换, 加权叠接相加听觉感知模型, 语音增强, Itakura-Saito距离

Abstract: This paper proposes a multichannel speech enhancement system based on non-uniform filterbank which can animulate heuring perceptual model. To achieve the same frequency resolution, the system uses less channels than that based on uniform filterbank. Simulations show the system achieves better quality in terms of Itakura-Saito(IS) distance than uniform filterbank-based speech enhancement system.

Key words: allpass transformation}Weighted OverLap-Add(WOLA), heuring perceptual model, speech enhancement, Itakura-Saito(IS) distance

中图分类号: 

• TP391