Optimizing Fused Face-Iris Biometric Recognition Accuracy and Timing Using Improved Mayfly Algorithm

Abstract

A multimodal biometrics system is presented in order to improve the
recognition performance, system complexity, security, and applicability
of current biometrics applications. In this study, an improved Mayfly
optimization algorithm was used as a feature selection method to
improve recognition accuracy and timing for a fused face-iris biometric
recognition system. The improved Mayfly algorithm is an enhancement
to the original Mayfly optimization algorithm. The Mayfly algorithm is
an optimization method based on the behavior of mayflies that provides
a powerful hybrid algorithm structure. It combines the best features of
particle swarm optimization, genetic algorithms, and the firefly
algorithm. Simulation experiments demonstrated that it is capable of
optimizing both benchmark functions, but with significant limitations.
Due to the random selection procedure used, which allows the existing
algorithm to exploit specific areas in the search space, notable
shortcomings included slow convergent rate, premature convergent,
and potential imbalance between exploration and exploitation. As a
result, the Mayfly algorithm has found it difficult to solve highdimensional problem spaces such as feature selection. The Mayfly
algorithm is enhanced in this study with the roulette wheel selection
method, which replaces the random selection method used in the
existing Mayfly algorithm. Both the existing Mayfly algorithm and the
newly developed improved Mayfly algorithm were used as feature
selection on a fused face-iris recognition system in order to improve
recognition accuracy and time complexity. The results of simulation
experiments revealed that the Improved Mayfly algorithm increased the
recognition accuracy and time complexity of the fused face and iris
biometrics recognition system.