Fig. 3From: Speech dereverberation for enhancement and recognition using dynamic features constrained deep neural networks and feature adaptationThe architecture of the DNN coefficient mapping using a sequential training cost functionBack to article page