HIGNN-TTS: Hierarchical Prosody Modeling With Graph Neural Networks for Expressive Long-Form TTS.
Dake GuoXinfa ZhuLiumeng XueTao LiYuanjun LvYuepeng JiangLei XiePublished in: ASRU (2023)
Keyphrases
- text to speech
- neural network
- speech synthesis
- prosodic features
- genetic algorithm
- pattern recognition
- feed forward
- word processing
- connected components
- neural network model
- graph partitioning
- graphical representation
- back propagation
- fuzzy logic
- hybrid models
- directed graph
- recurrent neural networks
- graph theory
- graph structure
- directed acyclic graph
- hopfield neural network