Skip to main content

Table 5 ASR performance in WERs of MFCC + FDLP features modeled by hybrid (HMM/DNN)- and BN-HMM/GMM-based systems on MP-FR evaluation data while taking into account different foreign data for training

From: Exploiting foreign resources for DNN-based ASR

System

HMM/DNN

BN-HMM/GMM

(training)

Supervised

Semi-supervised

Supervised

Semi-supervised

 

(%)

(%)

(%)

(%)

MP-FR (19 h)

14.4

13.9

– final

12.9

 

12.5

 

+ESTER-57 (57 h)

13.6

14.2

12.9

13.5

– final

12.7

13.0

12.1

12.6

+MP-GE (18 h)

14.0

14.1

13.2

13.2

– final

12.7

12.8

12.4

12.6

  1. Systems denoted as “final” exploit enhanced algorithms, as described in Section 5.5