First International Conference on Spoken Language Processing (ICSLP 90)

Kobe, Japan
November 18-22, 1990

Eur-Accor: A Multi-Lingual Articulatory and Acoustic Database

William J. Hardcastle (1), Alain Marchal (2)

(1) Speech Research Laboratory, University of Reading, UK
(2) CNRS, Aix-en-Provence, France

A large number of digitized databases of speech items have been compiled recently for various purposes: for training material; benchmark evaluation of speech products; diagnostic evaluation of speech recognizers etc. (e.g. ESPRIT/SAM EUROM.O [1];DARPA/TIMIT [2]). EUR-ACCOR is unique in that it consists of both articulatory and acoustic data for seven different European languages - English, French, German, Italian, Catalan, Irish Gaelic and Swedish. This database is currently being compiled as part of the ACCOR project ("Articulatory/acoustic correlations in coarticulation processes: a cross-language investigation") within the Basic Research Action framework of ESPRIT II (Action 3279). We will describe in this paper the main features of this multi-lingual articulatory and acoustic database.

Full Paper

Bibliographic reference.  Hardcastle, William J. / Marchal, Alain (1990): "Eur-accor: a multi-lingual articulatory and acoustic database", In ICSLP-1990, 1293-1296.