ABSTRACT
This chapter describes face recognition algorithm based on the combination of discriminative feature extractors and fusion of local and global classification results. Decision fusion of classification results on local histograms and global decision on the concatenated coefficients, enhances the recognition accuracy. Gabor and Centrally Symmetric Local Binary Pattern (G-CS-LBP) are combined to extract distinctive features insensitive to appearance variations. Through a block-based strategy, subdivided the G-CS-LBP images into small sub blocks. The chapter discusses the combination of Gabor filter bank and CS-LBP as a local texture descriptor. The proposed classifier in this paper is extreme learning machine (ELM) which is applied locally on the local histograms of concatenated block histograms for different G-CS-LBP images. Despite the iterative tuning of the weights and biases in gradient decent to minimize the cost function, ELM assign random values to hidden layer biases and input weights.
