Appearance and Shape Based Face Recognition Using Backpropagation Learning Neural Network Algorithm with Different Lighting Variations

Abstract

This paper presents an approach of face recognition system using Backpropagation learning neural network algorithm introducing appearance and shape based facial features to enhance the efficiency with different lighting variations. To extract the appearance and shape based facial feature, Active Appearance Model (AMM) has been applied. Appearance based facial feature is useful when the lighting condition is uniform. On the other hand when the environmental lighting condition is different, shape based facial features can perform better in comparison with appearance based feature because shape based structure is not changed with lighting variations. In this work, both appearance and shape based facial features are combined to enhance the recognition efficiency for various light variant system. For dimensionality reduction of appearance and shape based facial features, Principal Component Analysis (PCA) method has been used. Finally, error Backpropagation learning feed forward neural network algorithm has been used to classify the facial features. To measure the performance of the proposed appearance and shape based facial recognition system, VALID database has been used where each face has been captured with four different lighting variations. Experiments have been performed with Appearance-Only, Shape-Only and combined Appearance-Shape based feature and performance of the proposed system shows the superiority of the face recognition system.