基于等边三角形麦克风阵列的语音增强

Speech enhancement using a small trigonal microphone array

针对小体积应用场合下的语音增强,提出了一种基于等边三角形结构的麦克风阵列与Wiener后置滤波相结合,并由VAD作为控制单元的语音增强方法。该方法克服了自适应零限波束形成只能抑制空间相干噪声的缺点,并得到可在二维平面上旋转的波束主瓣。通过仿真和真实环境的实验,证明算法能够显著地提高输入语音信噪比,且适用于多种噪声场。

This paper describes a speech enhancement technique for low volume and low power applications. The method uses a small trigonal microphone array combined with Wiener filtering and voice activity detection （VAD）. The VAD is the controlling unit. This method more effectively suppresses spatial nonstationary and stationary noise than the adaptive null-forming beamformer method by using a rotatable main lobe in the 2-D plane. Simulations and experiments show that the algorithm significantly increases the SNR and is effective in various kinds of noise fields.