Ameliorating the Accuracy & Dimensional Reduction of Multi-modal Biometrics by Deep Learning

Abstract

Enhancement of an individual s security and protection is the major concern in the current era. Biometric security becomes a positive solution, adopted for many identification problems in various sectors, including Aviation. As the avionics business keeps growing worldwide, air terminals face the difficulties of increasingly more passenger volumes; with that, travelers desire for a secure and consistent pass through the airport terminals is just expanding. COVID-19 has managed a hit in all parts of our lives. Maintaining social distance and wearing masks are very important in maintaining a person s health in this pandemic situation. In this scenario, facial recognition for passenger identification is difficult and tends to false acceptance of passengers with identical faces. To address this issue, contactless biometric traits like iris and fingerprints are considered for security growth. The paper deals with the fusion of iris and fingerprint to identify passengers without removing the masks and promotes safe travel. The iris and fingerprint features are extracted using Gray Level Co-occurrence Matrix(GLCM) and cross number technique. Maintaining these templates requires more space for storage. To make this simple, Sparse autoencoder, an unsupervised deep learning technique is incorporated on the fused template for attaining the dimensionality reduction. A minimum cost matcher(MCM) is employed to increase the multimodal biometric system s accuracy and performance. Thus, the proposed system honors passenger security builds confidence in air travel, and also strives for economic growth.