Histogram equalized deep PCA with ELM classification for expressive face recognition
Kanokmon Rujirakul, Chakchai So–In · 2018 International Workshop on Advanced Image Technology (IWAIT) · 2018
In this paper, we propose a novel approach for expressive face recognition with deep learning networks. There are three main components: 1) Histogram Equalization (HE), 2) Principal Component Analysis (PCA), and 3) Extreme Learning Machine (ELM). The first module is used for pre-processing to adjust a histogram curve of input images. Then, a deep learning concept with PCA is applied as a feature extraction, and finally, ELM is used as a baseline classification scheme. Two well-known public databases, LFW and KDEF, are selected to evaluate the proposed method with a comparative performance evaluation against a traditional PCA with several classification methods and state of the art facial recognition technique, i.e., PCAnet, where the experimental results demonstrate our superior performance.