Abstract
This paper explores the possibility to replace the usual thresh- olding decision rule of log likelihood ratios used in speaker verification systems by more complex and discriminant de- cision functions based for instance on Linear Regression models or Support Vector Machines. Current speaker verifi- cation systems, based on generative models such as HMMs or GMMs, can indeed easily be adapted to use such decision functions. Experiments on both text dependent and text in- dependent tasks always yielded performance improvements and sometimes significantly.