Go To Program 137 A Distortionless Subband Beamformer for Noise Reduction in Reverberant Environments Emanuel Habets 1 1 Department of Electrical and Electronic Engineering, Imperial College London, London, United Kingdom In this paper a distortionless subband beamformer for noise reduction in reverberant environments is described and evaluated. The subband beamformer is based on a recently proposed spatio-temporal prediction model (STPM). An alternative expression is deduced that does not require the explicit estimation of the STPM. In addition, online estimators for the speech and noise power spectral density matrices are proposed that are based on the conditional speech presence probability (SPP). Finally, the SPP is used to enhance the output of the beamformer. The presented results demonstrate the subband beamformer’s ability to achieve significant noise reduction with low speech distortion. View pdf