Abstract—In this paper, we analyze two unknown speakers identification with having a sample of speech from a finite set of speakers and text-independent speech .Then refers to conventional methods in speaker identification after expression of the existing problems. Cepstral coefficients are evaluated as the most successful feature vectors. Gaussian mixture models, adapted gaussian mixture model and different normalization techniques have been introduced. Classification method of a telephone conversation to the speaker homogeneous segments is expressed. A speaker identification system is implemented and with various experiments was evaluated.

Index Terms—Speaker identification, gaussian mixture model, normalization, DET curve.

Mohsen Bazyar is with the Department of Communication, Bushehr Branch, Islamic Azad University, Iran (e-mail:mohsenbazyar114@yahoo.com).
Ahmad Keshavarz and Reza Dyanat are with the Persian Gulf University of Bushehr, Iran.

Cite: Mohsen Bazyar, Ahmad Keshavarz, and Reza Dyanat, "Speakers Identification in a Telephone Conversation Using Gaussian Mixture Model," International Journal of Computer and Communication Engineering vol. 1, no. 2, pp.  96-99 , 2012.