Harris Hawk Optimized CLAHE and Novel Score Level Fusion of Lightweight CNNs for Multimodal Biometric Recognition

Biometric authentication is critical for secure access in diverse domains, emphasizing the need for high accuracy to prevent security breaches. Multimodal biometrics fuse traits to fortify security, vital across diverse domains, surpassing unimodal systems' vulnerability. This paper introduces a novel approach that combines Contrast Limited Adaptive Histogram Equalization (CLAHE) and Harris Hawk Optimization(HHO) (CLAHE-HHO) for image preprocessing. It is followed by an innovative score-level fusion designed for the classification of Iris and Palm-print Biometric Modalities, using Transfer Learning based Lightweight ShuffleNet and MobileNetV3 Architecture. Initially, the Iris and Palmprint images undergo preprocessing using CLAHE, with the clip limit optimized through Harris Hawk Optimization. After preprocessing, Iris images are classified using ShuffleNet, while Palmprint images undergo classification using MobileNetV3. Scores from Palmprint and Iris modalities are fused using Weighted sum Score-level Fusion, with fusion weights determined via the novel Cauchy Search-based Golden Ratio Optimization (CGRO) technique. The performance of the system is evaluated on accuracy, Detection Error Tradeoff (DET) Curve, Equal Error Rate (EER), Total Training time. The proposed methodology proves that the multimodal authentication outperformed unimodal identification with a recognition accuracy of 100 % surpassing results obtained from individual iris and palmprint modalities.