Effectiveness of Feature Collaboration in Speaker Identification for Voice Biometrics

Abstract

Voice biometrics is a promising solution to online banking that doesn’t need one’s physical presence, unlike fingerprint and retina scanners. Systems for identifying speakers are a crucial component of biometric technologies. Over the past few years, numerous speaker identification systems have been developed and used; these systems rely on various feature extraction methodologies. Due to its capacity to capture’ the repeated nature and effectiveness of signals, short-time characteristics like perceptual linear predictive (PLP) and Mel frequency cepstral coefficients (MFCC) have been used in the majority of studies on speaker identification. The efficiency of MFCC characteristics in accurately identifying speakers has been demonstrated in various research. However, the’ performance of these features degrades in noisy environments. To address this feature, a novel feature fusion of some spectral and time-domain features has been suggested in this paper. Moreover, this study evaluates the effectiveness of feature collaboration for speaker identification. The experimental results show that the suggested feature vector and classifying model can be widely applied to different types of voice biometric systems.