Skip to main content

Table 7 Training time per epoch (hours) vs. the length of local window

From: Advanced recurrent network-based hybrid acoustic models for low resource speech recognition

Model The length of local window  
  10 20 30 40 50
LW-BLSTM-fbank 3.92 4.71 5.03 5.34 5.68
BLSTM-fbank    8.40  
LW-BLSTM-MBN 3.92 4.71 5.01 5.34 5.70
BLSTM-MBN    8.40  
\