Abstract: This paper proposes a new approach combing beamforming and subspace filter for noise reduction under multi-microphone condition. The directional characteristic of beamforming provides nulls against the noise signal and directional lobe to the target signal. The output of beamforming acts as input signal of subspace filter. Experimental results show that the method outperforms conventional algorithms in noise reduction, while its computation complexity is reduced.

Key words: beamforming, noise reduction, subspace divison

摘要： 提出一种基于波束形成技术与信号子空间滤波器的多麦克风噪声消除方法。利用波束形成器的空域滤波特性，使波束主瓣对准期望信号方向，零陷对准噪声方向，消除观察信号中弱相关与不相关噪声。利用信号子空间滤波器进一步消除波束形成器输出信号中的残余噪声。实验采用NOISEX92语音库，结果表明该方法具有较低的算法复杂度，有效降低了背景噪声。

关键词: 波束形成, 噪声消除, 子空间分解

CLC Number: 

• TP37