HS 4: Sprachsynthese I

Speech synthesis is an essential component of any system relying on "natural" and intuitive human-machine communication. Speech synthesis systems are also used in phonetic research to gain further insight into speech production and acoustic properties of speech. This seminar offers an introduction to speech synthesis systems and strategies. Various approaches to speech synthesis are presented, including formant synthesis, concatenative synthesis, and corpus-based unit selection. Participants will carry out hands-on exercises and experiments using speech synthesis software available at the IMS (KPE, Festival). In the compact seminar the participants will build a speech synthesis system for a restricted application domain.

Contact:

Dr. Bernd Möbius

Zimmer 3.15

Tel 121-1381

moebius@ims.uni-stuttgart.de

Sprechstunde: Mi 10.30-11.30 und jederzeit n.V.

Dates and topics:

17.10.	Introduction
24.10.	History and evolution of speech synthesis
31.10.	Formant synthesis, Introduction to the formant synthesis tool KPE
7.11.	Exercises in formant synthesis using KPE: vowels
14.11.	Exercises in formant synthesis using KPE: consonants
21.11.	Articulatory synthesis
28.11.	Concatenative synthesis (1)
5.12.	Concatenative synthesis (2)
12.12.	Corpus-based synthesis, unit selection (1)
19.12.	Corpus-based synthesis, unit selection (2)
9.1.	Linguistic text analysis: structure
16.1.	Linguistic text analysis: tasks and components
23.1.	Prosodic components
30.1.	Concept-to-speech and document-to-speech, TTS markup
6.2.	(no class)
13.2.	Evaluation of speech synthesis systems and quality
16.-20.2.	Compact seminar (all day)

Theory parts of the compact seminar:

16.2.	Compact seminar task and scenario
19.2.	Acoustic inventory design
19.2.	Speaker selection criteria
19.2.	Recording conditions
20.2.	Unit candidate selection

Talks and presentations by participants may be arranged anytime, e.g., by email to moebius@ims.uni-stuttgart.de or during the "Sprechstunde." Exercises may be carried out by students individually or in groups. For each topic specific literature will be indicated, and suggestions and support for the presentations will be provided.

Exercises:

• 07.11. -> 14.11. Exercises in formant synthesis with KPE: vowels
• 14.11. -> 21.11. Exercises in formant synthesis with KPE: consonants
• 09.01. -> 16.01. Exercise in linguistic text analysis
• 23.01. -> 30.01. Exercise with the Festival data structure

Instructions:

• KPE formant synthesis tool
• KPE overview slides (thanks to UCL)
• launching KPE
• KPE parameters and buttons
• Audio recording on SGI
• Festival
• Fringe

Literature:

• Richard Sproat (Hg.) (1998): Multilingual Text-to-Speech Synthesis. Kluwer, Dordrecht.
• Thierry Dutoit (1997): An Introduction to Text-to-Speech Synthesis. Kluwer, Dordrecht.
• Bernd Möbius (1999): "The Bell Labs German text-to-speech system". Computer Speech and Language 13, 319-358.
• BibTeX entries for all publications used or referred to in the seminar.

• Demo systems by previous synthesis seminars:
• IMS Pizza Service (WS 1999/2000)
• IMS S-Bahn Timetable and Connections Service (WS 1998/1999)
• Interactive German TTS systems:
• IMS Festival
• Bell Labs GerTTS (local access)
• Bell Labs GerTTS (official site)
• Hadifix (IKP Bonn)
• SVOX (ETH Zürich)
• Elan Informatique
• Audio demos of many speech synthesis systems, Dennis H. Klatt (1987).
• Audio demos of further TTS systems in various languages, Gregor Möhler (1999).
• LDC server for online side-by-side comparison of TTS systems
• TTS evaluation workshop in Jenolan Caves (Australia)
• SABLE Markup Language
• KPE formant synthesis tool