Parameter Settings for Speaker Identification using Gaussian Mixture Model

In this paper, the impact of the number of Gaussian mixtures, the duration of training and testing sessions, and the number of speakers on speaker identification has been investigated using clean speech (TIMIT) and telephone speech (NTIMIT) databases. Employing the parameters that provide the maximum performance, 100% and 85.71% identification rates have been obtained for the TIMIT and NTIMIT databases, respectively.