A broadband speech enhancement technique based on frequency invariant beamforming and GSC

Speech enhancement is one of the key and challenging techniques for a real-time speech recording system working in a meeting room, where cancellation of the competing speech and improvement of the speech intelligibility are two major concerns. In this paper, the application of frequency invariant beamforming to the speech enhancement problem based on the generalized sidelobe cancellation (GSC) structure is studied. It leads to a novel hybrid broadband speech enhancement technique, which combines the ideas of subband processing, frequency invariant beamforming and voice activity detection under the GSC framework. Simulation results show that the proposed method outperforms two previously proposed ones in terms of speech distortion, background noise and competing speech reduction.