Face detection and recognition in mobile devices designed for visually impaired persons

Abstract

Severe visual impairment is a type of disability that more affects the Brazilian population currently reaching approximately 7 million people, according to the latest statistics of the IBGE (Instituto Brasileito de Geografia e Estatística). This large number of visually impaired persons justifies the development of technologies that can help these people in their daily activities to improve their integration into society, giving them a better life quality. Among technologies that have developed over the past years include those related to mobile devices such as smartphones and tablets. This undergraduate research project proposes the development of a system to identify people through facial recognition on mobile devices, for use by persons who are visually impaired. To this end, the system will interface with adequate resources for this type of user. Facial recognition was selected since it is one of the most accepted biometric techniques. Besides, it can perform remotely and the input data (the faces) can be collected by the cameras embedded into the mobile devices. To implement this system it will be evaluated some techniques for face detection based on skin color and Haar features, as proposed by Viola-Jones. For face representation and description it will be evaluated techniques based on Principal Component Analysis (PCA), because of its wide use, and Local Binary Pattern (LBP), a technique also known as Census Transform, due to its higher invariance to lighting changes. The proposed system is an extension of the BioMobile system, recently developed by the Recogna research group, Department of Computing, School of Sciences, UNESP, Bauru. (AU)