Skip to main content

Advertisement

Table 12 Upper bound and baseline for the experiments using the simulated reverberant data

From: Single-channel dereverberation by feature mapping using cascade neural networks for robust distant speaker identification and speech recognition

Method Acoustic model Digit accuracy (%)
Set A Set B Avg. set A Avg. set B Avg.
Office EvHall InCar LivingR Lounge JPstyR ConfR BathR
Close (Ideal) Clean 99.5 99.4 99.5 99.3 99.5 99.4 99.5 99.3 99.4 99.4 99.4
Multi 96.2 96.5 96.1 96.2 96.2 96.5 96.1 96.2 96.2 96.2 96.2
Baseline Clean 97.5 57.9 95.6 84.4 74.0 89.5 89.8 78.0 83.8 82.8 83.3
Multi 94.3 90.5 94.7 91.5 79.6 93.5 93.4 84.2 92.8 87.7 90.2
CMN Clean 97.8 66.0 98.7 83.5 87.3 92.2 93.3 81.7 86.5 88.6 87.6
Multi 92.8 91.9 92.5 90.0 92.6 91.9 93.1 81.3 91.8 89.7 90.8