An overview of the CAVE project research activities in speaker verification

This article presents an overview of the research activities carried out in the European CAVE project, which focused on text-dependent speaker verific ation on the telephone network using whole word Hidden Markov Models. It do cuments in detail various aspects of the technology and the methodology use d within the project. In particular, it addresses the issue of model estima tion in the context of limited enrollment data and the problem of a posteri ori decision threshold setting. Experiments are carried out on the realisti c telephone speech database SESP. State-of-the-art performance levers are o btained, which validates the technical approaches developed and assessed du ring the project as well as the working infrastructure which facilitated co operation between the partners. (C) 2000 Elsevier Science B.V. All rights r eserved.