期刊名称：International Journal of Electrical and Computer Engineering
电子版ISSN：2088-8708
出版年度：2012
卷号：2
期号：3
页码：325-332
DOI：10.11591/ijece.v2i3.246
语种：English
出版社：Institute of Advanced Engineering and Science (IAES)
摘要：The basic theme of our paper is to implement a new idea of noise reduction in the real time applications using the concepts of adaptive filters. Our model which is presented as one of the solutions is based on two stages of operation with the first stage based on the ALE (Adaptive Line Enhancer) filters and the second stage on NLMS (Normalized Least Mean Square) filter. The first stage reduces the sinusoidal noise from the input signal and the second stage reduces the wideband noise. Two input sources of voice are used; one for the normal speech and the other for the noise input, using separate microphones for both signals. The first signal is of the corrupted speech signal and the second signal is of only the noise containing both wideband and narrowband noise. In the first stage the narrowband noise is reduced by using the ALE technique. The second stage gets a signal with ideally only the wideband noise which is reduced using the NLMS technique. In both the stages the concerned algorithms are used to update the filter coefficients in such a way that the noise is cancelled out from the signal and a clean speech signal is heard at the output.
关键词：Adaptive line enhancer, Normlized least mean squares, Sinusoidal noise, Wideband noise.
