Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions
Autor: | Yannis Agiomvrgiannakis, Jonathan Shen, Ron Weiss, Zhifeng Chen, Rif A. Saurous, Rj Skerrv-Ryan, Navdeep Jaitly, Zongheng Yang, Mike Schuster, Ruoming Pang, Yu Zhang, Yonghui Wu, Yuxuan Wang |
---|---|
Jazyk: | angličtina |
Rok vydání: | 2017 |
Předmět: |
FOS: Computer and information sciences
Computer Science - Computation and Language Computer science Mean opinion score Speech recognition 020206 networking & telecommunications Speech synthesis 02 engineering and technology computer.software_genre Reduction (complexity) 030507 speech-language pathology & audiology 03 medical and health sciences 0202 electrical engineering electronic engineering information engineering Feature (machine learning) Waveform Spectrogram 0305 other medical science Computation and Language (cs.CL) computer Decoding methods |
Zdroj: | ICASSP |
Popis: | This paper describes Tacotron 2, a neural network architecture for speech synthesis directly from text. The system is composed of a recurrent sequence-to-sequence feature prediction network that maps character embeddings to mel-scale spectrograms, followed by a modified WaveNet model acting as a vocoder to synthesize timedomain waveforms from those spectrograms. Our model achieves a mean opinion score (MOS) of $4.53$ comparable to a MOS of $4.58$ for professionally recorded speech. To validate our design choices, we present ablation studies of key components of our system and evaluate the impact of using mel spectrograms as the input to WaveNet instead of linguistic, duration, and $F_0$ features. We further demonstrate that using a compact acoustic intermediate representation enables significant simplification of the WaveNet architecture. Accepted to ICASSP 2018 |
Databáze: | OpenAIRE |
Externí odkaz: |