ESCA Tutorial and Research Workshop on
Speech Input/Output Assessment and Speech Databases

Noordwijkerhout, The Netherlands
September 20-23, 1989

Design and Recording of a Large Speech Database Over the Local Telephone Network in English and in French

Raymond Descout (1), Pierre Dumouchel (2), Pierre Hamel (1), Louis Vrooment (2)

(1) Canadian Workplace Automation Research Centre (CWARC), Department of Communications of Canada, Laval, Quebec, Canada
(2) Centre de Recherche Informatique de Montreal Inc. (CRIM), Montreal, Quebec, Canada

To provide an assessment of different speech recognition systems, a speech database was recorded over the local telephone network in English and in French. In this base, the same data is available for testing either algorithms on mainframe computers or commercial PC-based speech recognition boards using an analog input. An automatic PC-based server was designed for recording the speech materials (the signal and the MFCC coefficients). The corpus is composed of 49 words (isolated, connected digits and control words). 600 persons were recorded. The data was transferred in both formats: digital files on WORM and DAT recordings. Furthermore, a CD-ROM version will be mastered for an easier dissemination and manipulation of the database.

Full Paper

Bibliographic reference.  Descout, Raymond / Dumouchel, Pierre / Hamel, Pierre / Vrooment, Louis (1989): "Design and recording of a large speech database over the local telephone network in English and in French", In SIOA-1989, Vol.2, 241-244.