Discovering discrete subword units with binarized autoencoders and hidden-Markov-model encoders.
Leonardo BadinoAlessio MeretaLorenzo RosascoPublished in: INTERSPEECH (2015)
Keyphrases
- hidden markov models
- speech recognition
- denoising
- markov model
- sequential data
- hidden states
- markov models
- handwritten word recognition
- gesture recognition
- conditional random fields
- viterbi algorithm
- input image
- continuous hidden markov models
- document images
- discriminative training
- multi stream
- baum welch
- kernel density
- n gram
- learning algorithm
- hidden state
- dynamic probabilistic networks