In this paper the linguistic design, speaker selection, recording procedures and transliteration instructions for three Dutch speech corpora are described. The first corpus is the Dutch POLYPHONE, which can be used for several aplications, among which speaker identification and verification. A special addition to the POLYPHONE corpus is then described, which aims specifically at problems in forensic speaker identification. Finally, a corpus is described that is specifically aimed at speaker verification. All three corpora are limited to telephone speech, recorded of a digital line.
Cite as: Boves, L., Bogaart, T., Bos, L. (1994) Design and recording of large data bases for use in speaker verification and identification. Proc. ESCA Workshop on Automatic Speaker Recognition, Identification and Verification, 43-46
@inproceedings{boves94_asriv,
author={Louis Boves and Tineke Bogaart and Leonie Bos},
title={{Design and recording of large data bases for use in speaker verification and identification}},
year=1994,
booktitle={Proc. ESCA Workshop on Automatic Speaker Recognition, Identification and Verification},
pages={43--46}
}