Deep Generative AI-Based Multimodal Biometric Authentication System for Enhanced Security and Accessibility in Healthcare Applications
M. Vijay, P. Nagaraj, N. Sundareswaran, S.R.B. Christopher · 2025
A multimodal biometric authentication system in healthcare strengthens security and accuracy by using multiple biometric traits for user verification, which makes unauthorized access significantly more challenging. By incorporating different biometric modalities, these systems reduce enrolment errors, improve performance, and enhance user accessibility, especially for individuals with cognitive or physical limitations. The suggested system creates an individual&s;s unique identifier by analyzing biometric data from their finger, vein, ear, and iris. The five primary sections of the proposed system architecture are as follows: data collection, preprocessing, feature extraction, feature selection, and user authentication. We begin by pre-processing the input photos for each modality to increase their quality. We perform deep hidden feature extraction from each modality&s;s input photos using the synthetic rabbit optimization (SRO) technique. The Adaptive Omega Q-square (AOQS) technique handles data dimensionality by selecting the most optimal features from the extracted features. The deep neural network (DNN) calculates this recognition score for each biometric modality. Score-level fusion takes scores from the finger vein, ear, and eye and combines them into a single recognition score. A hybrid gradient neural network (GNN) authenticates individuals based on their final recognition score to achieve optimal results. Finally, we validate the performance of the proposed system by using benchmark SDUMLA-HMT and AMI ear database. It has been demonstrated through the simulation results that the proposed multimodal biometric authentication system achieved a maximum accuracy of 98.3% and 98.1% for the two benchmark databases, respectively. Security, and accessibility.