Embedding a Differentiable Mel-Cepstral Synthesis Filter to a Neural Speech Synthesis System.
Takenori YoshimuraShinji TakakiKazuhiro NakamuraKeiichiro OuraYukiya HonoKei HashimotoYoshihiko NankakuKeiichi TokudaPublished in: ICASSP (2023)
Keyphrases
- speech synthesis
- speech recognition
- hidden markov models
- vocal tract
- network architecture
- text to speech
- prosodic features
- automatic speech recognition
- language model
- neural network
- filtering algorithm
- loss function
- objective function
- pattern recognition
- program synthesis
- noisy environments
- vector space
- speech corpus
- information retrieval
- neural model
- information hiding
- bio inspired
- texture synthesis
- speech signal
- associative memory
- noise reduction
- color images