麦克风阵列的自适应算法通过迭代运算获取波束形成的最优权矢量时,噪声模型的估计是一个非常关键的因素。它的好坏直接影响着系统波束形成的性能。系统地分析了最小均方(LMS)自适应语音增强算法,并针对阻塞矩阵在估计噪声时存在的缺陷,在该算法的基础上提出了一种利用最小值控制递归平均(MCRA)来估计噪声的方法。将此方法应用于波束形成,并用Maflab软件进行仿真。仿真实验结果表明,MCRA估计出的噪声使LMS自适应语音增强的效果更好和抗噪性更强。
When adaptive algorithm of the microphone array uses iteration to obtain the optimal weight vector about beamforming, the noise model estimation is a critical factor. It will have a direct impact on the performance of beamforming system. Analyse systematically the least-mean-square(LMS ) adaptive speech enhancement algorithms, and propose a method using a minima controlled re.cursive aver- aging (MCRA) to estimate the noise in allusion to the defects of the blocking matrix. This method is applied to beamforming, and using Matlab software to simulate. Simulation results show that,LMS-MCRA adaptive has robust speech enhancement.