DMCA
UNDERSTANDING HOW DEEP BELIEF NETWORKS PERFORM ACOUSTIC MODELLING
Cached
Download Links
| Citations: | 27 - 3 self |
BibTeX
@MISC{Mohamed_understandinghow,
author = {Abdel-rahman Mohamed and Geoffrey Hinton and Gerald Penn},
title = {UNDERSTANDING HOW DEEP BELIEF NETWORKS PERFORM ACOUSTIC MODELLING},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Deep Belief Networks (DBNs) are a very competitive alternative to Gaussian mixture models for relating states of a hidden Markov model to frames of coefficients derived from the acoustic input. They are competitive for three reasons: DBNs can be fine-tuned as neural networks; DBNs have many non-linear hidden layers; and DBNs are generatively pre-trained. This paper illustrates how each of these three aspects contributes to the DBN’s good recognition performance using both phone recognition performance on the TIMIT corpus and a dimensionally reduced visualization of the relationships between the feature vectors learned by the DBNs that preserves the similarity structure of the feature vectors at multiple scales. The same two methods are also used to investigate the most suitable type of input representation for a DBN. Index Terms — Deep belief networks, neural networks, acoustic modeling
Keyphrases
neural network feature vector index term gaussian mixture model many non-linear hidden layer phone recognition performance acoustic modeling multiple scale input representation deep belief network timit corpus hidden markov model belief network similarity structure competitive alternative dbn good recognition performance acoustic input suitable type
