Skip to main content
EURASIP Journal on Advances in Signal Processing
Fig. 9 | EURASIP Journal on Advances in Signal Processing

Fig. 9

From: Front-end technologies for robust ASR in reverberant environments—spectral enhancement-based dereverberation and auditory modulation filterbank features

Fig. 9

ASR performance of BN features for Dev. MFCC(L=R=4) and AMFB features are used as the input to generate BN {MFCC} and BN {AMFB}, respectively. For comparison, LDA-MLLT is also applied to MFCC(L=R=4) and AMFB features to reduce their dimensions to 40. GMM- and SGMM-HMMs are used for evaluation. The SE algorithm is employed to SGMM-HMM afterwards. Solid lines are for SimData, and dashed lines for RealData

Back to article page