Optimal feature level fusion based iris and fingerprint multimodal biometric system using improved multi kernel svm

Abstract

The modern society attained secured mechanism to lead their processes in different applications such as airports, hospitals, banks, autonomous and non-autonomous institutions, etc with the improvement of biometric system. Nowadays, biometric technique is employed for human identification process based iris, fingerprint, ear and palm etc. In order to render the effective biometric system we have improved multi-model biometric recognition established on iris and fingerprint. Our work is established on three modules such as recognition module, pre-processing module, and feature extraction module. Then, in the feature extraction module, we processed feature extraction established on changed Local Binary Pattern (MLBP) feature and GLCM features. Fish Swarm optimization algorithm is applied for processing feature level fusion. For recognition, developed Multi Kernel Support vector machine (IMKSVM is inaugurated. In the document, several kernels are integrated to give shape to an innovative hybrid kernel which incredibly improves the classification task of segregating the training data. By way of offering the hybrid kernel, the SVMs gainfully achieve the flexibility to pick the appropriate shape of the threshold, for which it is not essential that it is linear and possesses the identical functional shape for the entire data, in view of its non-parametric function and local operation. We estimated our suggested technique with existing technique therefore; we get better recognition accuracy and successfully implemented our technique in MATLAB platform.