Enhanced speaker recognition based on intra-modal fusion and accent modeling

Speaker recognition systems, even though they have been around for four decades, have not been widely considered as standalone systems for biometric security because of their unacceptably low performance, i.e., high false acceptance and rejection. Research has shown that speaker recognition performance can be enhanced through hybrid fusion (HF) of likelihood scores generated by arithmetic harmonic sphericity (AHS) and hidden Markov model (HMM) techniques. Performance improvements of 22% and 6% true acceptance rate (TAR) at 5% false acceptance rate (FAR) were observed, when evaluated on two different datasets - YOHO and USF multi-modal biometric dataset respectively. In this paper, we present a model that combines accent information from an accent classification (AC) system with HF system in order to further increase the speaker recognition rate. The proposed system achieved performance improvements of 17% and 15% TAR at an FAR of 3% when evaluated on SAA and USF datasets. The accent incorporation method discussed in this work can also be applied to any other speaker recognition system.