ICASSP2022 TTS&VC Summary

总结了ICASSP2022中TTS和VC相关论文，主要是TTS。统计了sessions，topics，tasks/motivation以及对应的论文数量。

• TTS相关工作最多的是acoustic model，其次是expressiveness和prosody。
• Acoustic model主题中关于模型结构的论文居多，其次是关于AM中的时长建模。
• Expresiveness主题中关于disentanglement的论文居多。
• Prosody主题中关于control的论文居多。
• Front-end主题中关于G2P，多音字消歧的论文居多。
• Vocoder主题中大多是提升效率的论文。
• Multimodal主题中关于配音的论文居多。

---

Content

TTS

• Sessions
• Topics
• Tasks
  • Acoustic model
  • Expressiveness
  • Prosody
  • Front-end
  • Vocoder
  • Adaptation
  • Multimodal
  • Multi-lingual/Cross-lingual TTS
  • Singing voice synthesis
  • Speech editing
  • Others

VC

• Sessions
• Tasks

---

TTS

TTS Sessions

	Sessions	#Sessions	#Papers
1	Expressiveness/Adaptation	4	24
2	General topic	2	12
3	Novel acoustic model	1	6
4	Front-end	1	6
5	Vocoder and evaluation	1	6
6	Multi-lingual/Multimodal	1	6
7	Singing Voice and others	1	5
Total	7	11	65

TTS Topics

	Topics	#Papers
1	Acoustic model	12
2	Expressiveness	10
3	Prosody	9
4	Front-end	8
5	Vocoder	6
6	Adaptation	5
7	Multimodal	5
8	Multi-lingual/Cross-lingual	4
9	Singing voice synthesis	2
10	Speech editing	2
11	Others	2
Total	11	65

TTS Task/Motivation

AM (Acoustic model)

	Tasks	#Papers
1	Model	3
2	Duration	2
3	Analysis	2
4	Input	1
5	Speaker	1
6	Efficiency	1
7	Noisy data	1
8	Incremental TTS	1
Total	8	12

Expressiveness

	Tasks	#Papers
1	Disentanglement	3
2	Emotion	2
3	Low-quality	1
4	Adaption	1
5	Reference selection	1
6	Conversational TTS	1
7	Low-resource	1
Total	7	10

Prosody

	Tasks	#Papers
1	Control	4
2	Rich prosody	2
3	Cross-sentence context	2
4	Word-level prosody	1
Total	4	9

Front-end

	Tasks	#Papers
1	G2P	2
2	Polyphone disambiguation	2
3	Prosodic structure prediction	1
4	POS model compression	1
5	End-to-end text normalization	1
6	Mathematical formulas	1
Total	6	8

Vocoder

	Tasks	#Papers
1	Efficiency	4
2	New method	2
Total	2	6

Adaptation

	Tasks	#Papers
1	Speaker generation	1
2	VC for postprocessing	1
3	Multimodal	1
4	Low-quality data	1
5	New structure	1
Total	5	5

Multimodal

	Tasks	#Papers
1	Dubbing	4
2	Speech-to-animation	1
Total	2	5

Multi-lingual/Cross-lingual TTS

	Tasks	#Papers
1	Data augmentation	1
2	Lifelong learning	1
3	Triple loss	1
4	Improved structure	1
Total	4	4

Singing voice synthesis

	Tasks	#Papers
1	End-to-end	1
2	Melody unsupervision	1
Total	2	2

Speech editing

	Tasks	#Papers
1	Speech editing	2
Total	1	2

Others

	Tasks	#Papers
1	MOS net	1
2	Phase reconstruction	1
Total	2	2

---

VC

VC Sessions

	Sessions	#Sessions	#Papers
1	Conversion	2	12
2	Representation	1	6
3	Singing voice and others	1	6
Total	3	4	24

VC Tasks

	Topics	#Papers
1	VC	7
2	One-shot/Representation	6
3	Singing VC	3
4	Dysarthric speech	2
5	Noise robust	2
6	Prounication robust	1
7	Streaming VC	1
8	Data augmentation	1
9	Tool	1
Total	9	24