Some study results of speaker verification from Chung-Hwa TL and MJIB of ROC

Abstract

In this paper, we will present some study results from the cooperation of these two units and propose several useful methods to improve the performance of speaker verification. These methods are the discriminative training algorithm, the scoring function and the acoustic segment based probabilistic model. A 100-speaker digit database collected through the public switching telephone network is used for a series of experiments. Besides mentioning the advantages of each proposed method, a series of experiments and analytic steps are also given.