Skip to main content
EURASIP Journal on Advances in Signal Processing

Table 5 Comparison of acoustic feature (WER (%))

From: Reverberant speech recognition combining deep neural networks and deep autoencoders augmented with a phone-class feature

 

Feature

SimData

RealData

DNN-HMM (mc)

MFCC

10.97

32.97

DNN-HMM (mc) + DAE

MFCC

11.80

32.51

DNN-HMM (mc)

LMFB

8.74

29.67

DNN-HMM (mc) + DAE

LMFB

10.62

24.93

Back to article page