Embedding a Differentiable Mel-cepstral Synthesis Filter to a Neural Speech Synthesis System.
Takenori YoshimuraShinji TakakiKazuhiro NakamuraKeiichiro OuraYukiya HonoKei HashimotoYoshihiko NankakuKeiichi TokudaPublished in: CoRR (2022)
Keyphrases
- speech synthesis
- speech recognition
- text to speech
- prosodic features
- vocal tract
- network architecture
- hidden markov models
- neural network
- vector space
- automatic speech recognition
- speech signal
- language model
- noisy environments
- objective function
- speaker identification
- loss function
- pattern recognition
- neural network model
- machine learning
- face recognition
- program synthesis
- linear prediction
- neural model
- filtering algorithm
- bio inspired
- noise reduction
- multiresolution