Adaptive beamformer for hands-free communication system in noisy environments

The paper proposes a new adaptive algorithm for speech enhancement in hands-free communication systems. The scheme aims to enhance a target signal corrupted by background noise and by acoustic echo. The main idea is the incorporation of both the source and the interference short-term spectral amplitudes in the solution. These spectral components are estimated iteratively to provide a spectrally optimized constraint while the background noise statistics are estimated and updated continuously. Evaluations in a hands-free car environment show that the proposed algorithm is superior in suppressing both the noise and the interference even in a double-talk situation.