基于麦克风阵列的语音增强算法研究  

Research on Speech Enhancement Algorithm Based on Microphone Array

作　　者：闫姝[1] 权建军[1] YAN Shu;QUAN Jianjun(School of Electronic and Electrical Engineering,Lanzhou Petrochemical College of Vocational Technology,Lanzhou 730060,China)

机构地区：[1]兰州石化职业技术学院电子电气工程学院

出　　处：《自动化仪表》2019年第9期59-62,共4页Process Automation Instrumentation

基　　金：国家自然科学基金资助项目(61461024)

摘　　要：麦克风阵列信号处理技术的语音增强方法,能够充分利用语音信号的时空信息,其波束控制能力、抗干扰能力和信号增益均优于传统的方法。对于广义旁瓣抵消(GSC)的自适应滤波算法,在噪声相干的情况下具有很好的噪声抑制作用,但并不适用于噪声为非相干情形;反之,维纳滤波算法在噪声非相干的情况下对噪声有很好的抑制作用,而又不适用于噪声相干情形。为防止出现在高信噪比的情况下信号相消的现象,首先对GSC的阻塞矩阵进行了改进,其次对维纳滤波算法中信噪比取值的不确定性进行了改进,最后尝试将两种算法进行融合。仿真结果表明:融合算法在两种噪声情况下都具有较好的噪声抑制能力,在复杂噪声环境中具有更高的可靠性,因而更具实用价值。The speech enhancement method of the microphone array signal processing technology can fully utilize the spatiotemporal information of the speech signal,and its beam steering capability,anti-interference ability and signal gain are superior to the traditional methods.For the adaptive filtering algorithm of generalized sidelobe cancellation(GSC),it has good noise suppression in the case of noise coherence,but it is not suitable for non-coherent noise.On the other hand,the Wiener filtering algorithm has a good suppression of noise in the case of noise incoherent,but it is not suitable for noise coherence.First,in order to prevent signal cancellation in the case of high signal-to-noise ratio,the GSC blocking matrix is improved.Then,the uncertainty of the value of SNR in Wiener filtering algorithm is improved.Finally,the two algorithms are fused.The simulation results show that the fusion algorithm has better noise suppression ability under two kinds of noise conditions and higher reliability in complex noise environment,so it is more practical.

关 键 词：麦克风阵列 维纳滤波 广义旁瓣抵消 波束形成 时延估计 相干噪声 语音增强 

分 类 号：TH841[机械工程—仪器科学与技术]