Maximum a posteriori Based Decoding for CTC Acoustic Models.
Naoyuki KandaXugang LuHisashi KawaiPublished in: INTERSPEECH (2016)
Keyphrases
- maximum a posteriori
- acoustic models
- speech recognition
- markov random field
- map estimation
- maximum likelihood
- image reconstruction
- discriminative training
- bayesian framework
- broadcast news
- em algorithm
- automatic speech recognition
- speech recognizer
- prior model
- expectation maximization
- hidden markov models
- generalized gaussian
- energy function
- label field
- high quality
- training data
- super resolution
- multi view
- pattern recognition
- training process
- mixture model
- graphical models