Joint Dereverberation and Residual Echo Suppression of Speech Signals in a Noisy Environment

Hands-free devices, such as mobile phones, are perhaps one of the most prominent applications of acoustic echo cancellation systems. They are often used in a noisy enclosed environment. Therefore, the recorded near-end speech will also contain reverberation and background noise, which degrade the fidelity and intelligibility of speech. In the last two decades postprocessors have been developed for suppressing residual echo and background noise in single microphone acoustic echo cancellation systems. Residual echo and noise suppression are usually performed using a simple spectral enhancement technique, which can distort the desired near-end speech signal. Furthermore, dereverberation of the near-end speech signal has never been addressed in this context. In this paper we present a novel post-processor which suppresses late reverberation of the nearend speech signal, residual echo, and background noise. In our approach the acoustic echo path is divided into three distinct parts, each parts is handled in a different way. Moreover, we determine room characteristics such as the reverberation time from the estimated acoustic echo path. These room characteristics are used to estimate the residual echo related to the tail of the acoustic echo path and the late reverberant energy of the near-end speech signal. A single post-filter is applied to the error signal of the acoustic echo canceller to obtain an estimate of the dereverberated near-end speech signal. Experimental results demonstrate the beneficial use of the proposed system for reducing reverberation, residual echo and background noise.