面向二元麦克风小阵列改进的广义旁瓣抵消器语音增强算法

GSC-like Speech Enhancement for Dual Small Microphone Array

二元麦克风小阵列在手机、助听器等受空间、成本以及运算能力限制的设备中被广泛研究用以提高目标语音质量。二元麦克风小阵列中语音增强算法主要包括波束形成方法以及相干滤波器方法。波束形成方法的思想是利用目标声源相对阵列的位置关系获取相应的时域和空域信息,可以保留目标声源方向的信号而抑制其他方向的干扰信号;相干滤波器方法则通过阵元间不同信号的相关性进行噪音抑制。考虑这两种类型方法的优点,本文提出一种面向二元麦克风小阵列改进的广义旁瓣抵消器语音增强算法,通过在广义旁瓣抵消器的固定波束形成支路上使用相干滤波器,提高固定波束形成输出信号的信噪比,然后在广义旁瓣抵消器自适应支路利用阵列的时域和空域信息对固定波束形成支路输出的信号中残余噪音进行估计,进而获得增强后目标输出信号。仿真和实际试验表明,本文提出的算法明显优于单独使用小阵列波束形成算法和相干滤波器算法。

Dual small microphone array has been extensively studied in order to increase the quality of the target speech for mobile phones,hearing aids,and other equipments which subject to space,costs and limits of computing power.Dual small microphone array speech enhancement algorithm is mainly including beamforming technique and coherence filter.The idea of beamforming technique is to use the position between desired source and the array to obtain the time-domain and space information to keep the desired direction of the signal and the other direction of the interference signal will be suppressed.Noise reduction by coherence filter is used relative signals information between microphones.Consider the advantages of these two method,this paper proposes an improved generalized sidelobe canceller（GSC） speech enhancement for dual small microphone array.The output noise of fixed beamformer（FBF） of GSC is suppressed by coherence filter to improve signal to noise ratio（SNR）,and the residual noise in the fixed beamformer output can be suppressed by adaptive branch of GSC.Simulations and practical tests show that the proposed algorithm is superior to separate small array beamforming algorithms and coherence filter algorithm.