Completely Unsupervised Phoneme Recognition By A Generative Adversarial Network Harmonized With Iteratively Refined Hidden Markov Models.
Kuan-Yu ChenChe-Ping TsaiDa-Rong LiuHung-yi LeeLin-Shan LeePublished in: CoRR (2019)
Keyphrases
- hidden markov models
- phoneme recognition
- completely unsupervised
- discriminative training
- speech recognition
- speaker dependent
- generative model
- markov models
- conditional random fields
- sequence classification
- audio signal
- hidden states
- discriminative classifiers
- automatic speech recognition
- hidden state
- image processing
- noise reduction
- feature set
- continuous hidden markov models