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: INTERSPEECH (2019)
Keyphrases
- hidden markov models
- phoneme recognition
- completely unsupervised
- discriminative training
- sequential data
- speech recognition
- speaker dependent
- conditional random fields
- markov models
- hidden state
- generative model
- sequence classification
- automatic speech recognition
- machine learning
- baum welch
- continuous hidden markov models