作者: Cédric Jaboulet , Philippe Langlais , Jean-Luc Cochard , Andrei Constantinescu , Gérard Chollet
DOI:
关键词:
摘要: Following the demand of speech technology market, a number companies and research laboratories joined their forces in order to produce valuable reusable resources, especially databases. Serving purpose, collected databases are used for developing, testing, enhancing evaluating products, like interactive voice servers, listening typewriter, speaker verification identification systems, etc. Especially capturing intra-speaker variability, PolyVar database was designed recorded at IDIAP, as complement Swiss French PolyPhone database, which adresses inter-speaker variability issues. We will detail following specific problems collection (sampling population, selection vocabulary items, ...), present actual development we carried out IDIAP throught