自适应语音消噪算法的研究与仿真

针对语音通信中的噪声问题,对最小均方误差(LMS)算法进行研究。研究发现,该算法在收敛速度与稳态误差之间始终存在着矛盾,为此在F-LMS算法的基础上,提出一种改进的LMS算法,该算法通过引入误差加权累加的平均值的方法来更好地解决两者之间的矛盾,并通过计算机仿真证实了该算法具有良好的收敛性能和稳态性能,最后利用传统的LMS算法、F-LMS算法和改进的LMS算法对带有噪声的信号进行了消噪处理,结果表明:在三种算法中,改进的LMS算法的噪声消除效果最好。

ADAPTIVE VOICE DENOISING ALGORITHM AND ITS SIMULATION

Least mean square(LMS) algorithm has been studied for the noise problem in voice communication.The study founds that in the algorithm there is always the contradiction between the convergence speed and the steady-state error.To solve this,a new LMS algorithm is proposed based on the F-LMS algorithm,which introduces the error weighted cumulative mean to solve this contradiction in a better way.Through the computer simulation it has confirmed that this algorithm has good convergence and steady-state performance.Finally,by using traditional LMS algorithm,F-LMS algorithm and the improved LMS algorithm respectively,the signal with noise is undertaken the de-noising processing.Results show that the improved LMS algorithm has best noise cancellation effect among the three.