Face Recognition using Multi-class BSVM with Component Features

We present a fast and simple method that detects and extracts local components of face. The method is based on Haar wavelets and integral projections. It automatically locates facial components, extracts them and combines them into a single feature vector which is classified by multi-class bias support vector machine (BSVM). Multi-class BSVM translates the multi-class SVM classification problem to the single-class SVM problem, it is more convenient for optimization than the other multi-class SVM methods. Our experiments indicate our component-based recognition system is faster than other methods