Adaptive Parallel Subgradient Projection Techniques with Input Sliding Technique for Stereophonic Acoustic Echo Cancellation

In this paper, we present a novel adaptive filtering scheme for Stereophonic Acoustic Echo Cancellation (SAEC). The proposed scheme is based on the adaptive parallel subgradient projection techniques (Yamada et. al., 2002) combined with the input sliding technique (Joncour and Sugiyama, 1998, Sugiyama, Joncour, Hirano, 2001). The scheme uses information both in the current and previous periods of the input sliding technique simultaneously while the original technique just uses information in the current period. Robustness to noise is achieved by introducing the stochastic property sets. Moreover it requires only linear computational complexity because it is free from solving systems of linear equations. The numerical example shows that the proposed scheme can achieve, even in noisy situation, much faster convergence as well as lower miss-identification of echo paths and higher ERLE than the conventional technique.