Local Binary Pattern Based Feature Extraction with KELM for Face Identification

This paper emphasizes a novel methodology for robust face identification system, which is unification of holistic and local feature extraction technique and non-iterative learning algorithm. Here, Local Binary Pattern (LBP) is utilized to locate and summarize the local features corresponding to micro-regions of face image. The obtained features are classified (intra-class) by concatenating and correlating with corresponding local-statistics. After the classification of images with in same class using LBP operator, the inter-class classification is performed using fast and accurate kernel based extreme learning machine (KELM), learning algorithm. The efficacy of proposed LBP based KELM machine learning algorithm is tested and compared with other state-of-art methodologies. The experimental results evaluated on facial image databases evidently confirmed the supremacy of proposed approach.