Maximum-a-Posteriori-Based Decoding for End-to-End Acoustic Models.
Naoyuki KandaXugang LuHisashi KawaiPublished in: IEEE ACM Trans. Audio Speech Lang. Process. (2017)
Keyphrases
- end to end
- computational complexity
- maximum a posteriori
- acoustic models
- speech recognition
- map estimation
- markov random field
- image reconstruction
- maximum likelihood
- hidden markov models
- bayesian framework
- speech recognizer
- em algorithm
- automatic speech recognition
- discriminative training
- expectation maximization
- broadcast news
- generalized gaussian
- energy function
- multipath
- speech signal
- image segmentation
- support vector
- spoken language
- scalable video
- image compression
- high quality
- mixture model