KL-divergence regularized deep neural network adaptation for improved large vocabulary speech recognition.
Dong YuKaisheng YaoHang SuGang LiFrank SeidePublished in: ICASSP (2013)
Keyphrases
- speech recognition
- kl divergence
- speaker adaptation
- neural network
- pattern recognition
- speech recognizer
- kullback leibler divergence
- language model
- hidden markov models
- information theoretic
- mahalanobis distance
- speech signal
- automatic speech recognition
- speech recognition systems
- speaker independent
- gaussian distribution
- translation model
- speaker identification
- posterior distribution
- gaussian mixture
- exponential family
- mutual information
- probabilistic model
- maximum likelihood
- mixture model
- feature space