Experimental Analysis of Face and Iris Biometric Traits Based on the Fusion Approach

Abstract

Objectives : To develop an efficient algorithm for face and iris multimodal traits on ORL and CASIA dataset and to increase the performance rate and decrease the error rate of the model. The main goal is to increase the performance rate and decrease the error rate of the model. Methods: The proposed algorithm utilizes a fusion of face and iris modalities using Stationary Wavelet Transform (SWT) and Local Binary Pattern (LBP) techniques. The Principal Component Analysis (PCA) is applied to reduce the dimensionality of each sample, improving efficiency while preserving the most relevant information. The relevant characteristics from both face and iris modalities are fused to create a comprehensive pattern for an individual. Findings: The obtained features are compared with the features of the database images using a Euclidean Distance classifier. The performance of the proposed model is evaluated using the ORL and CASIA iris datasets. The accuracy achieved by the proposed algorithm is 99.42%, demonstrating robustness. Novelty: The algorithm introduces feature-level fusion, combining the characteristics of both face and iris modalities. The model encompasses the training and recognition phases within a biometric system. During the training phase, the biometric modality is captured and processed using the fusion of SWT+LBP+PCA techniques to form a template for each user. These templates are later stored in the database for recognition purposes.