Differentiable pooling for unsupervised speaker adaptation
- 1 April 2015
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 15206149,p. 4305-4309
Abstract
This paper proposes a differentiable pooling mechanism to perform model-based neural network speaker adaptation. The proposed technique learns a speaker-dependent combination of activations within pools of hidden units, was shown to work well unsupervised, and does not require speaker-adaptive training. We have conducted a set of experiments on the TED talks data, as used in the IWSLT evaluations. Our results indicate that the approach can reduce word error rates (WERs) on standard IWSLT test sets by about 5-11% relative compared to speaker-independent systems and was found complementary to the recently proposed learning hidden units contribution (LHUC) approach, reducing WER by 6-13% relative. Both methods were also found to work well when adapting with small amounts of unsupervised data - 10 seconds is able to decrease the WER by 5% relative compared to the baseline speaker independent system.Keywords
This publication has 28 references indexed in Scilit:
- Learning hidden unit contributions for unsupervised speaker adaptation of neural network acoustic models2014
- Multi-level adaptive networks in tandem and hybrid ASR systems2013
- 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
- Maximum likelihood linear transformations for HMM-based speech recognitionComputer Speech & Language, 1998
- Gradient-based learning applied to document recognitionProceedings of the IEEE, 1998
- Connectionist probability estimators in HMM speech recognitionIEEE Transactions on Speech and Audio Processing, 1994
- Backpropagation Applied to Handwritten Zip Code RecognitionNeural Computation, 1989
- Neocognitron: A new algorithm for pattern recognition tolerant of deformations and shifts in positionPattern Recognition, 1982