An Improved Wrapper Based Feature Selection using Hybrid BFO for Multimodal Biometrics

The automatic identification of an individual using physiological/behavioral traits connected to a person is Biometrics. Pattern recognition is used in biometric authentication (identification/verification) systems to recognize an individual based on measurements of certain physiological or behavioral features. Various studies prove that multimodal biometrics is more reliable than unimodal. This research suggests using fingerprint and palmprint biometrics to create a multimodal biometric system. The Gabor filter extracts fingerprint and palm print characteristics, and feature subset selection is performed. With Correlation-based Feature Selection (CFS) - Hybrid Bacterial Foraging Optimization - a novel enhanced feature selection approach for Gabor features is provided (BFO). Results reveal that the new method produces improved performance.