Face recognition using contourlet-based features and hybrid PSO-neural model

This paper presents a hybrid technique for face recognition. The proposed technique consists of four stages: feature extraction, dimensionality reduction, feature selection, and classification. In the first stage, the features related to face images are obtained using Contourlet Transformation (CT). The dimension of features is reduced using Principal Component Analysis (PCA) to form more essential features. For the feature selection, Particle Swarm Optimization (PSO) algorithm is used to search the feature vector space for the optimal feature subset. Finally, in the classification stage, a classifier based on feedforward back-propagation artificial neural network is used with PSO-optimized hidden layer size and learning rate. The classification is performed with 90% accuracy on the ORL database. This result shows that the proposed technique is robust and effective as compared with hybrid model based on discrete wavelet transform and PCA (DWT-PCA).