The Parameterized Phoneme Identity Feature as a Continuous Real-Valued Vector for Neural Network Based Speech Synthesis.
Zhengqi WenYa LiJianhua TaoPublished in: INTERSPEECH (2016)
Keyphrases
- speech synthesis
- real valued
- speech recognition
- continuous functions
- text to speech
- action space
- feature vectors
- complex valued
- prosodic features
- vocal tract
- neural network
- integer valued
- latent variables
- real valued data
- data sets
- speech corpus
- interval valued
- speech signal
- feature set
- distributed representations
- computer vision
- automatic speech recognition
- pairwise
- learning algorithm
- machine learning