Abstract
This paper presents a general framework for hybrids of hidden Markov models (HMM) and neural networks (NN). In the new framework called hidden neural networks (HNN) the usual HMM probability parameters are replaced by neural network outputs. To ensure a probabilistic interpretation the HNN is normalized globally as opposed to the local normalization enforced on parameters in standard HMMs. Furthermore, all parameters in the HNN are estimated simultaneously according to the discriminative conditional maximum likelihood (CML) criterion. The HNNs show clear performance gains compared to standard HMMs on TIMIT continuous speech recognition benchmarks. On the task of recognizing five broad phoneme classes an accuracy of 84% is obtained compared to 76% for a standard HMM. Additionally, we report a preliminary result of 69% accuracy on the TIMIT 39 phoneme task
Original language | English |
---|---|
Title of host publication | IEEE International Conference on Acoustics, Speech and Signal Processing |
Volume | 4 |
Publisher | IEEE |
Publication date | 1997 |
Pages | 3233-3236 |
ISBN (Print) | 0-8186-7919-0 |
DOIs | |
Publication status | Published - 1997 |
Event | 1997 IEEE International Conference on Acoustics, Speech and Signal Processing - Munich, Germany Duration: 21 Apr 1997 → 24 Apr 1997 Conference number: 22 http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=4635 |
Conference
Conference | 1997 IEEE International Conference on Acoustics, Speech and Signal Processing |
---|---|
Number | 22 |
Country/Territory | Germany |
City | Munich |
Period | 21/04/1997 → 24/04/1997 |
Internet address |