Fig. 7From: Joint training of DNNs by incorporating an explicit dereverberation structure for distant speech recognitionDNN-JT1 results: a static LMFB features of clean speech; b static LMFB features of reverberant speech with T60 = 0.61 s, distance = 3 m; c static LMFB features of DNN-PP dereverberated speech. d The output of front-end regression DNN: before joint training (estimated clean LMFB features). e The output of front-end regression DNN: after joint trainingBack to article page