Efficient Multimodal Biometric Recognition for Secure Authentication Based on Faster Region-Based Convolutional Neural Network
Ponugoti Kalpana, Surendar Rama Sitaraman, Sunil Swamilingappa Harakannanavar, Zaid Ajzan Alsalami, S. Nagaraj · 2024
In recent years of technology, securing the personal information is an important thing to avoid financial loss, criminal attacks. So, in this regard demand for security regulations, biometric recognition technology, and information security have been increased tremendously. To overcome the security issues in this research, Multimodal Biometric recognition system was implemented by using a proposed method called as Faster Region-based Convolutional Neural Network (FRCNN). The main advantage of using the proposed method is it will minimize the training process time as it consisting of two fully connected layers namely classifier and regression layer. Initially, the proposed method implemented on SDUMLA-HMT dataset that contains the 3 biometric traits - face, finger vein, and fingerprint. In data preprocessing stage by using the TopHat and Soble filtering method the quality of the image is improved by Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and K-means techniques. CNN which is significantly used to extract the features of the images and it also can be viewed as an automatic feature extractor. In classification stage, FRCNN is used as a classifier that achieved high accuracy of 99.53%, when compared with all other existing methods namely VGG-16, IrisConvNet, SVM and residual networks.