A new wavelet network architecture for echo cancellation

We put forward a new wavelet network architecture for echo cancellation. We can efficiently identify the echo path by the proposed network structure based on the theory of multiresolution analysis and functional analysis. The corresponding multiresolutional learning algorithm is also presented. The performance of the proposed wavelet network architecture with the corresponding learning algorithm are compared to that of the direct form FIR NLMS adaptive filter with application to echo cancellation. Simulation results indicate that the proposed method can provide faster convergence than the conventional FIR NLMS algorithm while not increasing the computational complexity.