Acoustic echo suppression based on speech presence probability

摘要

This paper proposes a novel acoustic echo suppression (AES) algorithm based on speech presence probability in a frequency domain. Double talk detection algorithm based on two cross-correlation coefficients modeled by Beta distribution controls the update of echo path response to improve the quality of near-end speech. The near-end speech presence probability combined with the Wiener gain function is used to reduce the residual echo. The performance of the proposed algorithm is evaluated by objective tests. High echo-return-loss enhancement and perceptual evaluation of speech quality (PESQ) scores are obtained by comparing with the conventional AES method.