Siri On-Device Deep Learning-Guided Unit Selection Text-to-Speech System

By Tim Capes, P. Coles, Alistair Conkie, L. Golipour, Abie Hadjitarkhani, Qiong Hu, Nancy Huddleston, Melvyn Hunt, Jiangchuan Li, M. Neeracher and 8 more. 2017

67 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Yu Zhang
2. Xu Tan
3. Tao Qin
4. Tie-Yan Liu
5. Yonghui Wu
6. Ye Jia
7. Heiga Zen
8. Ron J. Weiss

In Signal Processing

1. Yu Zhang
2. Shinji Watanabe
3. Heiga Zen
4. Yonghui Wu
5. Jinyu Li
6. Ron J. Weiss
7. Ye Jia
8. Xu Tan

Citations in the corpus

• LibriTTS: A Corpus Derived from LibriSpeech for Text-to-Speech (2019, background): “The properties which are addressed in this paper are as follows: • The audio files are at 16 kHz sampling rate ; 16 kHz sampling is high enough for the ASR purpose but too low to achieve high quality TTS. Modern production-quality TTS systems often use 24, 32, 44.1, or 48 kHz sampling rate [17,18].”
• Speech Processing for Digital Home Assistants: Combining signal processing with deep-learning techniques (2019, methodology): “The Siri Team at Apple developed ondevice, DL-guided, hybrid unit selection concatenative TTS systems [43] to achieve these goals.”
• Neural Text-to-Speech Adaptation from Low Quality Public Recordings (2019, background): “The input is text, which first undergoes text normalization and Grapheme-to-Phoneme conversion [21, 17, 22].”
• Deep Voice 3: Scaling Text-to-Speech with Convolutional Sequence Learning (2017, background): “Deployed TTS systems (e.g., Capes et al., 2017; Gonzalvo et al., 2016) should include a way to modify pronunciations to correct common mistakes (which typically involve proper nouns, foreign words, and domain-specific jargon).”