Proceedings of the 11 th INDIACom; INDIACom-2017; IEEE Conference ID: 40353 2017 4 th International Conference on “Computing for Sustainable Global Development”, 01 st - 03 rd March, 2017 Bharati Vidyapeeth's Institute of Computer Applications and Management (BVICAM), New Delhi (INDIA) Study of Various Multibiometric Techniques Divyakant T. Meva, Associate Professor, Faculty of Computer Applications, MEFGI, Rajkot Email ID: divyakantmeva@gmail.com Abstract- Biometrics is becoming necessity for security issues in Information Technology trend. Authentication and authorization purposes can be served efficiently and effectively with biometric techniques. But sometimes unimodal biometric systems (systems using one biometric trait) may not serve purpose because of various reasons. Mutibiometric systems resolve the issues which cannot be addressed using unimodal biometric systems. The author has presented comparison of various multibiometric systems in this paper. Keywords— Biometrics, Multibiometrics, Multimodal Biometrics I. INTRODUCTION Authentication and authorization of user has always been a challenging issue in the domain of information technology. Various techniques like password protection, smart cards, tokens were popular till last decade. But, these techniques are not the proved one. They have prompted various challenges for the said purpose. Usage of physical and behavioral traits, in other word, Biometric systems have proven their ability address the challenges proposed by traditional techniques. Unimodal biometric systems like fingerprint, face, palmprint, hand geometry, gait, keystroke have following limitations [1]: 1. Noise in the sensed data 2. Intra class variation 3. Inter class similarities 4. Non universality 5. Interoperability issues The limitations of unimodal biometric systems can be addressed with the usage of multiple samples, multiple algorithms, multiple sensors or multiple biometric traits. This approach is called Multibiometric system approach. II. CLASSIFICATION OF MULTIBIOMETRIC SYSTEMS Multibiometric systems can be categorized on the based on sources of evidence used for authentication or authorization. The fusion of information can be there from either single trait or multiple traits of biometric system. The following classes can be identified based on usage of singe biometric trait[1]: 1.Multi sensor system - These systems use multiple images of singe biometric trait captured with multiple sensors 2.Multi algorithm system - These systems use images of the same biometric trait but processed with multiple algorithms 3.Multi instance system - These systems use images of multiple instances of the same biometric trait 4.Multi sample system - These systems use multiple samples of same instance of same biometric trait with captured with single sensor The following class of system uses multiple biometric traits : 1.Multi modal system - These systems uses images of multiple biometric traits Sometimes hybrid systems are also used by combining above mentioned classes of systems. Here we have considered an example of a 3G network which consist of one HLR, Three VLRs and simultaneously 3 MSCs as well as 3 LAs. The whole network can be graphically shown in figure - 2. III. VARIOUS FUSION TECHNIQUES The fusion of the information can be applied before matching or after matching. Here is the tabular representation of fusion techniques : TABLE I. FUSION TECHNIQUES Fusion prior matching Feature level fusion Sensor level fusion Fusion after matching Match score level fusion Rank level fusion Decision level fusion IV. RELATED WORK AND IMPLEMENTATIONS Multibiometric systems are in implementation since a long time and these implementation have been proved in performance also. Here are the details of various multibiometric system implementations. 4.1 Multi-sensor systems Multi-sensor systems extract information from registered images computers with multiple sensors. Here the images are captured for single trait. E.g. facial images of the person can be captured using CCD camera and range sensors for 2 D and 3D images for the purpose of authentication. Thermal IR offers a decent alternative to visible imagery to handle variations in face appearance. It is invariant to changes in ambient illumination. It supports almost all kind of lighting conditions. Copy Right © INDIACom-2017; ISSN 0973-7529; ISBN 978-93-80544-24-3 6612