Improvement of face recognition performance evaluation technology by using a display device through color adjustment

Abstract

Recently, with the increase in the number of social robots, there is a growing interest in face recognition technology for natural interaction between humans and robots. As a result, there is a growing demand for technology that can be used to evaluate the performance of face recognition capable of ensuring reproducibility and objectivity in real service environments. The best way to replace real faces for performance evaluation is to use the facial image output from a display device. Previous studies have reported no significant difference between recognition performance for real faces and facial image output from a display device. However, the question remains whether a difference exists between two images at the low levels; if it does, then how does it affect face recognition performance? To answer these questions, the current study shows color differences before and after the adjustment by using color as a representative feature at the low level, and validates the proposed evaluation method by comparing the differences in recognizing two images.