Acoustic Echo Cancellation in Distributed Network Using Improved Diffusion Subband Adaptive Filtering Algorithm

This paper proposes an improved diffusion subband adaptive filtering algorithm for an acoustic echo cancellation (AEC) in distributed network. For the adaptation step, a variable individual step size that takes account of the characteristics of speech input signals is derived by minimizing the upper bound of the mean-square deviation. For the combination step, an error-variance-based combination method is proposed that enables to reflect information on the sudden interferences such as double talk. Simulation results show that the proposed algorithm outperforms the existing diffusion adaptive filtering algorithms in terms of the convergence rate and the steady-state errors in various AEC environment.