Real-time joint dereverberation and speech enhancement for hearing aid applications using edge devices

In this work, we propose a joint dereverberation and speech enhancement technique for real-time applications. Speech dereverberation is based on the coherent-to-diffuse energy ratio (CDR) measured from the direction-of-arrival (DOA) dependent complex spatial coherence function. A personalized speech enhancement (SE) technique based on joint maximum a posteriori probability (JMAP) estimation that inherently accounts for the effects of changes in acoustic path and reverberation is used. The combination of these masking functions is used to enhance the speech corrupted by reverberation and noise in a frame-by-frame process. The proposed method is implemented on the smartphone (edge device), to illustrate real-time usability. The efficiency of the proposed method is evaluated using speech quality and intelligibility measures and compared with that of other benchmark techniques.