Published Paper Details:

Dereverberation of speech signals in a hands-free scenario by adaptive algorithms has been a research topic for several years now. Speech is more natural approach of gaining access to information, monitoring things and interacting. However, its intuitiveness is not convenient while interacting with computer. In real time situations, the speech quality is degraded by reverberation occurred in rooms and background noise. These effects causes combine corrupt the speech signal and overall speech level degraded resulting in listening issues in human beings moreover, for hearing-impaired persons. Neural network based speech dereverberation has achieved promising results in recent studies. This algorithm investigates deep learning based multi-channel speech dereverberation. For multi-channel processing, we first denoising the signals using symlet wavelet, and then LSTM for Dereverberation, which is expected to be a filtered version of target signals. The performance of the proposed algorithm is evaluated using performance evaluation metrics such Perceptual Evaluation of Speech Quality (PESQ) and LLR are used to validate our results.