Using deep bidirectional recurrent neural networks for prosodic-target prediction in a unit-selection text-to-speech system.
Raul FernandezAsaf RendelBhuvana RamabhadranRon HooryPublished in: INTERSPEECH (2015)
Keyphrases
- text to speech
- recurrent neural networks
- text to speech synthesis
- prosodic features
- chaotic time series
- speech synthesis
- prediction accuracy
- neural network
- feed forward
- recurrent networks
- reservoir computing
- echo state networks
- complex valued
- feedforward neural networks
- neural model
- prediction model
- word processing
- artificial neural networks
- programming tool
- long short term memory
- nonlinear dynamic systems
- english text
- writing skills
- noise reduction
- cascade correlation
- regression model
- probabilistic model
- expert systems