Classification Of Hand Images by Person, Age and Gender with The Median Robust Extended Local Binary Model

Abstract

Biometric technologies try to automatically recognize individuals by considering the physiological and behavioral characteristics of individuals. Although the methods used here are very diverse, the personal qualities used also vary. Facial features, finger and vein prints, iris, retina, ear, hand, and finger recognition are only some of the physiological features. It may be preferred to use one or more of these personal features to reduce the margin of error that may arise depending on the security level in the applications used. Biometric recognition systems have varying requirements in security systems applications. Fingerprint and iris recognition work well in applications that require high security levels, while applications that require low security levels are not suitable due to privacy concerns. On the other hand, identification from hand images is more accepted based on the idea that it does not have a very high distinctiveness. But it is sufficient for medium security applications. Apart from these, palm images have many advantages such as reliability, stability, user-friendliness, non-intrusiveness, and flexible use. In this study, it is aimed to identify people, determine their ages, and determine their gender by using both upper surface and inner surface images of right-left hand data of hand shape. For this purpose, images of both the inner surface of the hand (10) and the outer surface of the hand (10) of 100 different people were collected. This was done separately for the right and left hands, and a total of 3955 images were obtained. The features of these images were extracted using the Median Robust Extended Local Binary Model (MRELBP). Images are classified for person, age and gender. The results were 91.4%, 85.9% and 92.6%, respectively.