3D Nose shape net for human gender and ethnicity classification

摘要

Gender and ethnicity are significant characteristics of human beings. Using human facial data to classify gender and ethnicity of people is important in facial analysis research. A novel method is proposed to address this issue. The method is based on a 3D nose shape organization structure called "3D nose shape net". To construct the 3D nose shape net, a nose measurement method to determine the distances between different noses and to use the results to cluster noses is proposed. Using the nose clustering results, the 3D nose shape net is constructed. The proposed method uses only the nose data from the 3D face; it is robust to facial expressions and facilitates removal of the poses effect. The 3D nose shape net does not consider the texture information in the nose region; therefore it is robust to illumination and cosmetics on faces. Gender and ethnicity classification results are achieved in 3D nose shape net simultaneously. The experimental 3D nose shape nets are built and tested using the FRGC2.0 and Bosphorus3D datasets. (C) 2018 Elsevier B.V. All rights reserved.