Abstract: Wiener filtering algorithm is one of the commonly used algorithms to improve the speech intelligibility of hearing impaired patients in the noise environment. In this paper, in order to solve the large deviation problem in the noise spectrum estimation of traditional Wiener filtering algorithm, one speech noise reduction algorithm based on the improved multi-channel Wiener filtering is proposed. Firstly, with the features of human auditory and hearing aid loudness compensation, the speech signal will be decomposed into multi-sub-band signals by the Gammatone filter. Then the Wiener filter based on a priori SNR estimation is used for speech enhancement in each sub-band. Finally, the sub-band signals are integrated to get the enhanced speech. Furthermore, a voice activity detection algorithm based on the envelope estimate is proposed to improve the problem in the noise spectrum estimation of traditional Wiener filtering algorithm. And it is also used to improve the performance of the Wiener filter. Experimental results show that, compared with the traditional Wiener filtering, the proposed method can more effectively suppress the residual noise and improve speech intelligibility, which has high practical value.

Key words: hearing aid, Wiener filtering, priori Signal to Noise Ratio（SNR）, noise spectrum estimation

摘要： 维纳滤波算法是改善噪声环境下听障患者语音理解度的常用算法之一。针对传统维纳滤波算法噪声谱估计偏差大的问题，提出一种基于改进的多通道维纳滤波算法的助听器语音降噪算法。算法首先结合人耳听觉特性和助听器响度补偿的特点，将语音信号进行Gammatone分解为多路子带信号。然后在每个子带内用基于先验信噪比估计的维纳滤波器进行语音增强处理。最后通过综合子带信号，得到增强的语音。此外，为了改善维纳滤波算法噪声谱估计的问题，提出一种基于包络估计的语音活动检测算法，并用于改善维纳滤波性能。实验结果表明，与传统维纳滤波法相比，该方法能更有效地抑制残留噪声，提高语音可懂度，具有较高的实用价值。

关键词: 助听器, 维纳滤波, 先验信噪比, 噪声谱估计