Direct word discovery from speech signals based on hierarchical Dirichlet process-hidden language model and deep sparse autoencoder.
Tadahiro TaniguchiRyo NakashimaShogo NagasakaPublished in: ICDL-EPIROB (2016)
Keyphrases
- latent variables
- language model
- speech signal
- speech recognition
- probabilistic model
- latent topics
- n gram
- variational inference
- topic models
- translation model
- language modeling
- mixture model
- automatic speech recognition
- graphical models
- probabilistic graphical models
- statistical machine translation
- latent dirichlet allocation
- information retrieval
- co occurrence
- non stationary
- knowledge discovery
- word recognition
- conditional random fields
- expectation maximization
- high dimensional
- word segmentation
- hidden markov models
- gaussian mixture model
- bag of words
- keywords