Convolutional Neural Network and Kernel Extreme Learning Machine for Face Recognition

Due to its good generalization capability and excellent performance, kernel Extreme learning Machine (KELM) has gained attention in face recognition. However, the efficiency of KELM for face classification depends on the effective extraction of the features from the inputted data. Convolutional neural network (CNN) performs good in finding the invariant features. KELM performs well in approximating target function, but does not perform well in learning complicated invariance. Contrarily, CNN cannot be always considered as optimal choice for classification. In this work, a novel learning framework for face recognition based on CNN and KELM (CNN-KELM) is introduced. In the proposed technique, KELM is used as a classifier in CNN architecture for enhancing the performance of face recognition. Due to integration of CNN and KELM, CNN-KELM has the advantage of KELM and CNN and it is easy to learn. Comprehensive experiments on the face databases have verified the enhanced performance of the presented technique when compared with the various state-of-art approaches.