An assessment of the human performance of iris identification

Biometric iris recognition systems are widely used for a range of identity recognition applications and have been shown to perform with high accuracy. For large-scale deployments, however, system enhancements leading to a reduction in error rates are continually sought. In this paper we investigate the performance of human verification of iris images and compare against a standard computer-based method. Our results suggest that performance using the computer-based system is no better than performance of the human participants. Additionally and importantly, however, performance can be improved through incorporation of the human as a `second decision maker´. This fusion system yields a false acceptance rate of just 9% when disagreements are resolved in line with strengths of each `decision-maker´. The results are presented as an illustration of the benefits that can be gained when combining human and automated systems in biometric processing.