Face recognition based on geometric features using Support Vector Machines

Face Recognition is among the most widely studied problems in computer vision and Pattern Recognition. Face has many advantages like permanence, accessibility and universality. It is still now not solved in literature. Several approaches are proposed to overcome with problems including; changing posed, emotional states, and illumination variation, etc. Geometric approaches which used as example distance between noses, eyes, mouth are still less efficient compared to holistic approaches. However, it provide and additional local information such as shape of local facial parts, face unit action, etc. The major problem of these approaches is to select the most relevant distances that can differentiate human faces. In this paper, we propose a bag of geometrical features based face recognition approaches using Support Vector Machines (SVM), Genetic Algorithm (GA) and minimum redundancy maximum relevance (mRmR) with Mutual Information Difference (MID) and Mutual Information Quotient (MIQ). Support Vector Machine Classifier (SVM) based on linear, radial basis function and multi layer Perceptron kernels is performed on the two benchmarks of facial databases ORL and Caltech Faces. Linear kernel based SVM classification using 10 selected distances by Genetic Algorithm (GA) ranks top the list of kernels conducted in our experimental study.