In this paper we present a new method in 3D face recognition with expression variation images by FRAV3D database. We find a central region of 3D face and extract feature face in nose region. Then extract diagonal profiles of nose region as individual feature face and compare the profiles to recognize 3D faces. The experimental results indicate that diagonal profiles contain diacritic information in this region. Diagonal profiles on nose region achieve the highest performance 78.27% and used diagonal profiles with wide line of face in nose region achieve the 85% in facial expression images with FRAV3D database. Also by used weighting coefficient in each objects of probe set, achieved the highest performance 95.8%. Computational results demonstrator the accuracy and efficiency, which has been used in commercial applications for human recognition.