Acoustic Echo Cancellation Algorithm Based on Kalman Filtering of Skewed Observation Noise

Acoustic echo cancellation (AEC) is one of the key technologies in the field of speech enhancement. AEC based on Kalman filter (KF) not only ensures the effectiveness of the dual-talk scenario but also improves its ability to converge quickly. However, the performance of AEC is significantly compromised due to the influence of near-end speech and sudden noise, as the noise distributions does not completely adhere to the assumptions of the standard KF. In this article, a novel AEC algorithm based on KF of skewed observation noise is proposed to improve the effectiveness of the algorithm in complex signal channel scenarios. Based on the assumption that the state noise follows Gaussian distributions while the observed noise follows skewed Student’s
${t}$
distributions, the proposed algorithm is presented. The simulation and experimental results demonstrate the advantages of the proposed algorithm as compared with the existing state-of-the-art filters in terms of echo cancellation ability and convergence performance.