Demographic Classification with Local Binary Patterns

LBP (Local Binary Pattern) as an image operator is used to extract LBPH (LBP histogram) features for texture description. In this paper, we present a novel method to use LBPH feature in ordinary binary classification problem. Given a restricted local patch, the Chi square distance between the extracted LBPH and a reference histogram is used as a measure of confidence belonging to the reference class, and an optimal reference histogram is obtained by iteratively optimization; real AdaBoost algorithm is used to learn a sequence of best local features iteratively and combine them into a strong classifier. The experiments on age, gender and ethnicity classification demonstrate its effectiveness. Key words: real AdaBoost, LBPH, demographic classification

• Type: Conference
• Year: 2007
• Where: ICB
• Authors: Zhiguang Yang, Haizhou Ai
• Keywords: Biometrics, ICB 2007, LBP Histogram, Real Adaboost, Reference Histogram