COMPARING MULTILAYER PERCEPTRON TO DEEP BELIEF NETWORK TANDEM FEATURES FOR ROBUST ASR
Robust ASR
Přednášející: Oriol Vinyals, Autoři: Oriol Vinyals, Suman Ravuri, University of California Berkeley, United States
In this paper, we extend the work done on integrating multilayer perceptron (MLP) networks with HMM systems via the Tandem approach. In particular, we explore whether the use of Deep Belief Networks (DBN) adds any substantial gain over MLPs on the Aurora2 speech recognition task under mismatched noise conditions. Our findings suggest that DBNs outperform single layer MLPs under the clean condition, but the gains diminish as the noise level is increased. Furthermore, using MFCCs in conjunction with the posteriors from DBNs outperforms merely using single DBNs in low to moderate noise conditions. MFCCs, however, do not help for the high noise settings.
Informace o přednášce
| Nahráno: | 2011-05-26 17:15 - 17:35, Panorama |
|---|---|
| Přidáno: | 15. 6. 2011 19:11 |
| Počet zhlédnutí: | 39 |
| Rozlišení videa: | 1024x576 px, 512x288 px |
| Délka videa: | 0:21:00 |
| Audio stopa: | MP3 [7.10 MB], 0:21:00 |
Příbuzné přednášky
0:24:58
ROBUST SPEECH RECOGNITION USING DYNAMIC NOISE ADAPTATION
Robust ASR
Přidáno: 15. 6. 2011 19:02
0:18:54
MODEL-BASED COMPRESSIVE SENSING FOR MULTI-PARTY DISTANT SPEECH RECOGNITION
Robust ASR
Přidáno: 15. 6. 2011 19:18
Komentáře