Speech enhancement using Affine Projection Algorithm in subband

In this paper, a novel structure is proposed in order to improve the functionality of Adaptive Noise Cancellers (ANC); for this purpose Affine Projection Algorithm (APA) has been used. We exploit the Stationary Wavelet Transform (SWT) to decompose the input signals in subband to achieve a higher convergence rate. The reason of using SWT (an oversampled filter bank) is to reduce the effect of aliasing which produces some disturbance on the adaptive process. Since In the higher frequency subbands, there is no noticeable difference between APA and simpler approaches such as spectral subtraction or Wiener filter, these algorithms are substituted instead of APA for the sake of complexity. Simulation results show that the proposed ANC performance is better than the conventional ANC which uses adaptive algorithm in the time domain. . The results of subjective and objective tests confirm this fact.