A robust approach to facial ethnicity classification on large scale face databases

In this paper, we propose a novel approach using Kernel Class-dependent Feature Analysis (KCFA) combined with facial color based features to tackle the problem of ethnicity classification on large scale face databases. In our approach, a new design of multiple filtered responses of the Kernel Class-dependent Feature Analysis is used for ethnicity classes. In order to improve the robustness of our system, the facial color based features are also employed to incorporate with the filtered responses to extract the stable ethnicity features. Compared to previous approaches, our proposed method achieves the best accuracy of ethnicity classification on large industrial based face databases and the standard MBGC face database.