Learning hidden unit contributions for unsupervised speaker adaptation of neural network acoustic models
- 1 December 2014
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 171-176
Abstract
This paper proposes a simple yet effective model-based neural network speaker adaptation technique that learns speaker-specific hidden unit contributions given adaptation data, without requiring any form of speaker-adaptive training, or labelled adaptation data. An additional amplitude parameter is defined for each hidden unit; the amplitude parameters are tied for each speaker, and are learned using unsupervised adaptation. We conducted experiments on the TED talks data, as used in the International Workshop on Spoken Language Translation (IWSLT) evaluations. Our results indicate that the approach can reduce word error rates on standard IWSLT test sets by about 8-15% relative compared to unadapted systems, with a further reduction of 4-6% relative when combined with feature-space maximum likelihood linear regression (fMLLR). The approach can be employed in most existing feed-forward neural network architectures, and we report results using various hidden unit activation functions: sigmoid, maxout, and rectifying linear units (ReLU).Keywords
This publication has 33 references indexed in Scilit:
- An investigation of deep neural networks for noise robust speech recognition2013
- KL-divergence regularized deep neural network adaptation for improved large vocabulary speech recognition2013
- Speaker adaptation of context dependent deep neural networks2013
- Revisiting hybrid and GMM-HMM system combination techniques2013
- Linear hidden transformations for adaptation of hybrid ANN/HMM modelsSpeech Communication, 2007
- Catastrophic forgetting in connectionist networksTrends in Cognitive Sciences, 1999
- Maximum likelihood linear transformations for HMM-based speech recognitionComputer Speech & Language, 1998
- Speaker adaptation using combined transformation and Bayesian methodsIEEE Transactions on Speech and Audio Processing, 1996
- Maximum likelihood linear regression for speaker adaptation of continuous density hidden Markov modelsComputer Speech & Language, 1995
- Connectionist probability estimators in HMM speech recognitionIEEE Transactions on Speech and Audio Processing, 1994