Fast distributed multichannel speech enhancement using novel frequency domain estimators of magnitude-squared spectrum

The first figure shows that two proposed estimators (MMSE-MSS and MAP-MSS) achieve higher noise reduction and less speech distortion than two existing estimators (Subspace and MMSE-STSA). The second figure shows that increasing the number of microphones not only reduces noise but also has less signal distortion by the proposed MMSE-MSS estimator and MAP-MSS estimator.