Detection of blood-related signal from a series of fingerprint images acquired during an input action

Abstract

When a finger is pressed against a flat plate and deformed, blood inside the finger moves away from the deformed area. This causes the finger to change its appearance from reddish to white. As the finger leaves the plate, the blood comes back and it looks reddish again. We have proposed to use this color change to distinguish genuine fingers from artificial ones for un-attended fingerprint identification systems. This blood-related signal may reflect the stiffness of the peripheral blood vessels and therefore it may be correlated with some health conditions such as blood pressure. In experiments, we used a fingerprint sensor based on scattered light detection. Because the spectra of the light scattered by the deformed fingers showed large changes mostly in the green portion, an LED emitting at 525 nm at peak strength was used. First, we compared series of fingerprint images acquired during a normal input action and those obtained while a rubber band occluded the blood flow. The occluded finger required a larger force to exhibit a similar change for these pixel values than the finger without the rubber band. Second, we analyzed fingerprint images recorded by six volunteers. We defined some indices based on the pixel values of the fingerprint images and the pressure applied to the fingers. The correlation coefficient of one of such indices and the average blood pressure of the participants was 0.86. Although the number of the subjects is small, this initial result is encouraging.