Fig. 4From: Speech dereverberation for enhancement and recognition using dynamic features constrained deep neural networks and feature adaptationTemporal filtering of feature trajectories versus linear transformation of feature vectorsBack to article page