Studies Of Face Recognition Based On Machine Learning Methods

Face recognition is an important subject in artificial intelligence field, and it has gained extensive attention from researchers in the past decades. Due to the popularity of machine learning methods, face recognition technology based on machine learning methods is popular. Face recognition is an intersecting subject, and this paper is focusing on feature extraction algorithms of face recognition, and face recognition/tracking system.Firstly, this paper posed the definition of feature extraction and the basic principles of Principle Component Analysis(PCA),Linear Discriminant Analysis(LDA),Eigenfaces and Fisherfaces, discussed the general framework and study backgrounds of Generic Alogrithm and proposed a new algorithm called Bagging Evolutionary Feature Extraction Algorithm(BEFE) to combine Evolutionary Feature Extraction with Bagging method and applied it to face recognition. Compared with other GA based feature extraction algorithm, this algorithm has low space complexity and better search efficiency. Another advantage of this algorithm is that it provided a simple but effective way to study the difference between different subspaces from one initial data space. Besides, thanks to Whitened Principle Component Analysis (WPCA), weighted fitness function and the introduction of Bagging method, we can apply BEFE to classification with good and steady performances for the cases where the training set is small and the number of classes is big.Then, this paper explained what face recognition system is, Adaboost algorithm along with the improvement and the definition of Non-negative Matrix Factorization (NMF) together with the algorithms and improvements. Then it introduced the guideline of face recognition performances and discussed a face recognition system based on Adaboost and NMF. Also, we testified and discussed NMF using the standard face databases, thus showing that NMF is suitable for face recognition due to its partialness and sparseness.