Multi-Channel Noise Reduction with Interference Suppression on Mobile Phones

Speech enhancement techniques have been widely incorporated in voice communications for improving the speech quality and intelligibility. However, existing methods are not sufficiently effective in addressing complex noisy conditions in real-world hands-free scenarios. This is especially true for interferences from other competing talkers or point sources. Additionally, the prior work needs more improvements to fully exploit the deployment of multi-channel microphone array on the burgeoning high-end mobile phones. In this paper, we propose a multi-channel noise reduction (NR) method based on a hybrid adaptive noise coherence models without the need for the prior knowledge of the interference direction. The method also adapts to the fast-varying noisy and reverberant conditions. Evaluation results on real-world recordings collected via a mobile phone confirm its superior effectiveness in terms of the background noise/interference suppression and the overall speech quality with lower distortions compared to the state-of-the-art NR techniques.