Biometric identification system

Abstract

In today's world, most of the daily routine processes are moving towards digitalisation and automation. For most of the systems assisting in this transition, their security is based heavily on validation routines which utilise information to identify and give access to the client. But these methods prove insufficient in the current times. Most of the identification is carried out either through a physical token or through remembered information. But physical identification is easy to replicate and the remembered information has high chances of being forgotten or being cracked. This leads to the high chances of unauthorised access to the system. Hence the interest for a more dependable security framework to protect sensitive data. The answer to all the above is security based on Biometric Identification.

Biometric identification refers to the automatic recognition of a person through his/her physiological or behavioural traits such as face, fingerprint, DNA and palm print. Several types of identification schemes such as analysis of fingerprint and palm print, capillary vessel of retina, voice analysis, vein matching and even gait analysis. These traits are unique to an individual and are extremely difficult to replicate.

The current system uses two separate biometrics for the identification process -the hand vein and palm print matching. The system has the following subsections: Image Acquisition, Image Processing, Feature Extraction, Data Retrieval and Identification. The Vein Images have been obtained using external vendor device, producing NIR (Near Infra Red) and FIR (Far Infra Red) Images. The Palm print Images have been acquired using any image capturing device.

Although the application is extremely robust, it is not error free. Unimodal biometric systems have to contend with a variety of problems such as noisy sensor data, spoof attacks by imposters and unacceptable error rates. This issue can be alleviated by the deployment of Multimodal Biometric System, which integrates the multiple biometric information of an individual to determine its identity.

This project tackles this issue via implementing Multimodal Biometrics System using Rank Based Fusion by combining the information at the matching rank level using Borda Count Rank Fusion. The fusion is performed on ranks from Benchmark Palm Print Database (Poly_U), Mobile Phone Palm Print Database and Hand Vein Database. It can be concluded that the identification rate of the system improved. This report aims to provide the details on the strategy to thus assist the biometric identification system to be more robust and efficacious.