Note from the editor: Special issue on speaker recognition

Abstract

This special issue on speaker recognition has benefited from its exceptional guest editor, Hemant A. Patil, Associate Professor at Dhirubhai Ambani Institute of Information and Communication Technology (DA-IICT), who accepted my invitation to help identify leading researchers in the area of speaker biometrics whose work merited publication in the Journal and to assemble their work in a tightly organized issue that would add some significant new findings to the literature on speaker recognition. Over thirty five researchers contributed to this special issue from among universities and research laboratories in the United States, Germany, France and India. Each paper was rigorously peer reviewed. Eleven papers were chosen for this issue out of forty submissions. Among the outstanding papers contained in this special issue are those that analyze how speaker stress and emotions can compromise the accuracy of performing speaker identification and verification tasks; the investigation of novel approaches to speaker modeling; and the exploration of various source and system features for speaker recognition and/or voice conversion. The issue begins with a discussion of new methods for analyzing stress-induced variations in speech and the design of an automatic stress level assessment scheme that could be used in directing stress-dependent acoustic models or normalization strategies. The authors point out that current stress detection methods typically employ a binary decision based on whether the speaker is or not under stress which can be misleading because stress levels can both vary as well as change gradually. In the real world, stress levels evolve sometimes with a fair degree of subtlety. The authors consider two methods for stress level assessment: The