A novel speaker identification algorithm using classifiers fusion

• Published in: 2010 10th International Conference on Information Sciences, Signal Processing and their Applications pp. 145 - 148
• Main Authors: Deriche, Mohamed A, Naseem, Imran A
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.05.2010
• Subjects: Artificial neural networks; Mel frequency cepstral coefficient
• ISBN: 1424471656; 9781424471652
• DOI: 10.1109/ISSPA.2010.5605485

In this paper, a novel speaker identification technique using the Dempster-Shafer evidence theory is discussed. The objective is to fuse the complementary information present from different classifiers into a single decision. Here, we use a decreasing function of the distance (of the classifiers) as the belief function. We show that a combined classifier based on the Dempster-Shafer theory outperforms the individual LPCC and MFCC classifiers when used individually.