Score Fusion Strategies in Single-Iris Dual-Probe Recognition Systems

Abstract

Multiple samples can be utilised at the comparison stage of a biometric system in order to increase its biometric performance via information fusion or decision heuristics. It has been shown, that in a single-instance dual-probe setup, fusing the probe scores yields significant biometric performance increase over the single-probe baseline. Additionally, using the probe-probe comparison score was demonstrated to further improve the biometric performance of a fingerprint recognition system in a study by Cheng et al. In this paper, through a benchmark on the CASIA-IrisV4-Interval dataset and on the iris corpus of the BioSecure dataset, the aforementioned method is shown to be viable for an iris recognition system. However, since it requires an additional parameter, which must be estimated empirically, we propose a simpler method which exhibits similar biometric performance, while requiring no additional parametrisation.