Skip to main content

Advertisement

Table 5 WER (%) comparisons on RealData for distant-talking conditions with multi-channel beamforming front-end

From: Joint training of DNNs by incorporating an explicit dereverberation structure for distant speech recognition

System Room1 Room2 Room3 Avg
Eight-channel + beamforming systems     
Multi-2 16.73 18.83 22.23 19.26
DNN-JT1 15.43 17.81 20.06 17.77
DNN-JT2 15.57 18.11 19.84 17.84
DNN-JT3 16.31 19.15 21.91 19.12
DNN-JT4 15.18 17.74 19.90 17.61
  1. Room1 is a living room, Room2 is a conference room, and Room3 is a classroom