Machine Learning Based Robust Access for Multimodal Biometric Recognition

For organizations requiring high security clearance, multimodal sources of biometric scans are preferred. Computational models for the unimodal biometric scans have so far been well recognized but research into multimodal scans and their models have been gaining momentum recently. For every biometric we used separately feature extraction techniques and we combined those features in efficient way to get robust combination. In this paper, a novel method for fusion of the scan images from the different modes has been introduced. The method is based on representation of data in terms of its sparsity. Feature coupling and correlation information are obtained from the biometric images. The images from each mode are fused by taking into account a quality measure. The algorithms are kernelised so as to handle nonlinear data efficiently. The result of the proposed system is compared to already existing image fusion methods to show its advantage over them.