Face locations suitable drunk persons identification

Abstract

Drunk person identification is investigated in this work by means of applying neural networks on infrared images of faces. The images were acquired during controlled alcohol consumption by twenty persons. Since the sample of the persons employed in the experimental procedure was small, the proposed approach gives only indications which locations of the face are prominent to present significant changes with alcohol consumption. Applying each neural networks on a different location of the face we investigate their capability to converge on the specific locations. It is found that the face forehead and nose change thermal behavior with alcohol consumption. Small and large neural structures are tested. The results of this paper are coming to confirm a previous work of the outwards on this subject.