Robust Speaker-Adaptive HMM-Based Text-to-Speech Synthesis

By Junichi Yamagishi, Takashi Nose, Heiga Zen, Zhenhua Ling, Tomoki Toda, Keiichi Tokuda, Simon King, S. Renals. 2009

210 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. Frank K. Soong
2. Xu Tan
3. Tao Qin
4. Tie-Yan Liu
5. Junichi Yamagishi
6. Lei He
7. Lei Xie
8. Heiga Zen

In Signal Processing

1. Junichi Yamagishi
2. Frank K. Soong
3. Lei Xie
4. Zhizheng Wu
5. Lei He
6. Xu Tan
7. Zhenhua Ling
8. Tao Qin

Citations in the corpus

• Deep Voice 3: 2000-Speaker Neural Text-to-Speech (2017, background): “(2010) builds several speaker-adaptive HMM-based TTS systems (Yamagishi et al., 2009) on various ASR corpora with hundreds of speakers.”
• Deep Voice 2: Multi-Speaker Neural Text-to-Speech (2017, methodology): “For instance, in traditional HMM-based TTS synthesis (e.g., Yamagishi et al., 2009), an average voice model is trained using multiple speakers’ data, which is then adapted to different speakers.”
• WaveNet Vocoder with Limited Training Data for Voice Conversion (2018, methodology): “In this paper, we explore the approaches of building WaveNet vocoders with limited training data for VC. Speaker adaptation techniques have been developed for training acoustic models with a few samples in speech synthesis and voice conversion [23, 24, 25].”
• Speech Enhancement of Noisy and Reverberant Speech for Text-to-Speech (2018, methodology): “To train the adapted model we adopted a two stage adaptation, a CSMAPLR stage followed by a maximum a posteriori (MAP) update (CSMAPLR-MAP) [41], [43].”
• Multi-speaker modeling and speaker adaptation for DNN-based TTS synthesis (2015, background): “Index Terms— statistical parametric speech synthesis, deep neural networks, multi-task learning, transfer learning”
• CSS10: A Collection of Single Speaker Speech Datasets for 10 Languages (2019, background): “Not surprisingly, for the English language, there are many public speech datasets such as Blizzard [11], VCTK [12], LibriSpeech [13], TED-LIUM [14], VoxForge [15], and Common Voice [16].”
• Deep Voice 3: Scaling Text-to-Speech with Convolutional Sequence Learning (2017, methodology): “Our model with the WORLD vocoder achieves a comparable MOS of 3.44 on VCTK in contrast to 3.69 from Deep Voice 2, which is the state-of-theart multi-speaker neural TTS system using WaveNet as vocoder and seperately optimized phoneme duration and fundamental frequency prediction models.”
• On the training of DNN-based average voice model for speech synthesis (2016, methodology): “In [14], a speaker adaptive system was built, which employed several kinds of effective adaptation methods such as CSMAPLR+MAP and featurespace adaptive training.”
• From Speaker Verification to Multispeaker Speech Synthesis, Deep Transfer with Feedback Constraint (2020, background): “proposed feature-space adaptive training for speaker-adaptive TTS [21].”
• Statistical parametric speech synthesis using deep neural networks (2013, background): “This approach has various advantages over the concatenative speech synthesis approach [2], such as the flexibility to change its voice characteristics, [3–6], small footprint [7–9], and robustness [10].”