From: Discriminative likelihood score weighting based on acoustic-phonetic classification for speaker identification
SNR
ML-GMM
MCE-GMM (mixture weights only)
SVM
DSW-GMM
Error reduction over ML-GMM
Clean
10.90
9.00
8.30
23.85
20 dB
17.70
17.05
25.43
15.80
10.73
10 dB
30.02
29.35
42.45
26.45
11.89
0 dB
72.13
68.53
69.53
59.22
17.90