A Flexible ICA-Based Method for AEC Without Requiring Double-Talk Detection

In this paper, we propose the use of the flexible independent component analysis (ICA) algorithm for the acoustic echo cancellation (AEC). The flexible ICA algorithm has a parametric score function which is controlled by the Gaussianity of the source signal (speech in our case). The probability density function of the speech signal is not always the super Gaussian, hence the inflexible score function used in the existing ICAbased AEC system is not always suitable. On the other hand, the proposed method is able to always provide the suitable score function depending on the present Gaussianity of the speech signal. The experimental results indicate that the proposed method significantly outperforms the existing ICA-based AEC algorithm and the variable step-size (VSS)-NLMS algorithm.