Face Recognition and Spoofing Detection System Adapted To Visually-Impaired People

According to estimates by the World Health Organization, about 285 million people suffer from some kind of visual disability, of whom 39 million are blind, resulting in 0.7% of the world population. Computer vision techniques and image analysis can help improve visually-impaired people. In this project, a system that allows for facial recognition and detection of spoofing adapted to the needs of disabled people is proposed, implemented and validated. The architecture has been carefully selected and subsequently implemented following an innovative facial normalization algorithm in order to increase both the recognition rate of facial identification and spoofing detection. The information provided to the user is composed by the name of the person identified and whether it is real or fake image (photograph). This information is provided by means of a text-to-speech tool. This architecture can be integrated into video door-phone installations (videointercom installations), devices with reduced computing capabilities or the users' mobile phones. The architecture has been validated in a real environment with both real users and printed images achieving very good results.