【作者】 胡奎； 梁维谦；

【Author】 HU Kui ~1,LIANG Weiqian~2 (1.Institute of Microelectronics,Tsinghua University, Beijing 100084,China; 2.Department of Electronic Engineering,Tsinghua University, Beijing 100084,China)

【机构】 清华大学微电子与纳电子学系； 清华大学电子工程系；

【摘要】 针对近讲场景,提出一种双麦克近讲语音增强算法。该算法基于耳间延时差(ITD)、耳间强度差(IID)特征来区分目标语音,利用声学掩蔽效应,实现目标语音对背景噪声的分离降噪。与传统的单麦克增强方法相比,该算法可消除多种类型噪声且对语音造成的损伤较小。实验表明:该算法能将8～33 dB的白噪声、音乐噪声、广播噪声3种噪声类型的带噪语音的信噪比提高到36 dB以上,同时保持较高的目标语音可懂度。对于冲击噪声的带噪语音也具有较好的降噪效果。更多还原

【Abstract】 A two-microphone speech enhancement algorithm was developed for close-talk applications.The algorithm segregates the target speech from the noisy speech using cues such as the interaural time difference(ITD),the interaural intensity difference(ⅡD) and the masking effect.This algorithm suppresses more kinds of noise with less speech distortion than single-microphone speech enhancement methods.Tests show that the algorithm increases the SNR of 3 kinds of noisy signals from 8～33 dB to 36 dB.The enhanced speech is intelligible and the method handles impulsive noise well.更多还原