Age Invariant Face Recognition Using Quadratic Support Vector Machine – Principal Component Analysis

Age Invariant Face Recognition (AIFR) is a new research area in the domain of face recognition that lately received a lot of interest due to its tremendous potential with relevant applications in real-world. It is very challenging research problem to extract robust features describing aging facial information, usually when broad age difference between face images are perceived. In this paper, by using the quadratic support vector machine and principal component analysis technique of feature selection that is robust to aging we address this challenge and achieved the highest accuracy as compared to other methods such as K-Neighbor Neural Network (KNN), Probabilistic Neural Network (PNN), Backpropagation Neural Network (BPNN). FGNET aging dataset is used for the implementation of proposed method which includes a total of 82 separate subjects comprising 1002 images and each subject includes 6-18 images per subject obtained primarily from scanning photographs of subjects between the age ranging from newborn to 69-year-old subjects. From the images, LBP, Gabor and shape features have been extracted and PCA is used as a method of feature reduction. Four stages have been explored in this work: (a) Face recognition (b) Identify the gender of the object (c) Identify the age of the subject using the backpropagation neural network (BPNN) (d) Age invariant face recognition using KNN, PNN, NN and QSVM-PCA. The highest accuracy is achieved using the Quadratic Support Vector Machine – Principal Component Analysis (QSVM-PCA) classifier.