Abstract

In recent years, probabilistic features became an integral part of state-of-the-are LVCSR systems. In this work, we are exploring the possibility of obtaining the features directly from neural net with-out the necessity of converting output probabilities to features suit-able for subsequent GMM-HMM system. We experimented with 5-layer MLP with bottle-neck in the middle layer. After training such a neural net, we used outputs of the bottle-neck as features for GMM-HMM recognition system. The benets are twofold: rst, improvement was gained when these features are used instead of the probabilistic features, second, the size of the system was reduced, as only part of the neural net is used. The experiments were performed on meetings recognition task dened in NIST RT'05 evaluation. Index Terms Probabilistic features, bottle-neck features, TRAP-based features, LVCSR, meeting recognition.

Keywords

Computer scienceBottle neckProbabilistic logicHidden Markov modelTask (project management)Artificial intelligenceSpeech recognitionArtificial neural networkPattern recognition (psychology)BottleLayer (electronics)Machine learning

Affiliated Institutions

Related Publications

Publication Info

Year
2007
Type
article
Pages
IV-757
Citations
341
Access
Closed

External Links

View on DOI.org

Social Impact

Altmetric
PlumX Metrics

Social media, news, blog, policy document mentions

Citation Metrics

341
OpenAlex

Cite This

František Grézl, Martin Karafiát, Stanislav Kontar et al. (2007). Probabilistic and Bottle-Neck Features for LVCSR of Meetings. , IV-757. https://doi.org/10.1109/icassp.2007.367023

Identifiers

DOI
10.1109/icassp.2007.367023