Study Of Adaptive Noise Cancellation In Enhancement Of Speech

The performance of an adaptive noise canceller (ANC) is influenced by reverberation. This paper presents an algorithm for reducing the reverberation in the output of ANC. According to the reason that the excess mean-squared error increases linearly with desired signal power, the modified LMS algorithm is proposed on the basis of S.Ikeda algorithm. The greater input SNR is, the smaller the step size is. So the reverberation can be further reduced. The performances of S.Ikeda algorithm and the modified algorithm are compared in four cases of input SNR. The results of computer simulation show that the performances of the modified algorithm are better than those of S.Ikeda algorithm ,and the reverberation is reduced to a considerable degree using the modified algorithm by hearing.