Multimodal geometric sparse representation for reliable social network communication

In this work we plan to develop a framework called, Multimodal Geometric Sparse Representation (MGSR) for reliable social network communication with Face, and Fingerprint features of human individuals on social networks. Neighbor Flow Feature Extraction (NFFE) algorithm performs feature extraction using Neighborhood flow for both the face and fingerprint features. The most predominant attributes are stored in a geometric sparse vector for both the modalities that obtains the exact features even when integration is made for two different features, addressing, non-aligned features. Biometric Matching algorithm is designed to evaluate true positive rate of test data to the available training data. Experimental evaluation with Biometric Research Repositories conducted with varied model features and fusion template size show that the Biometric Template Matching algorithm is able to significantly improve true positive rate of human biometric samples. Further, we demonstrate that our NFFE framework has a low execution time suited for deployment in real time social network sites.

I. Introduction

Due to the various market (consumer, product, service) entering into the social media, reliable and secure communication between the users are one of the most important problems to be addressed. In recent years, both single and multimodal biometrics user authentication have been explored extensively in addressing this challenge. A collaborative Face Recognition [1] with the aid of social graph addressed the problems related to hit rate, ensuring authentication rate. Game theoretic models were presented in [2] [3] using fingerprinting to minimize the collision rate. At the same time, a multimodal biometric system was proposed in [4] [5] using behavioral traits to measure the presence of students in virtual classroom. An additional factor of gestures for different face characteristics was presented in [6]. Identity authentication was on the other hand ensured in [7] through keystrokes.