ABSTRACT
Classification of gender can be performed with the textural and geometric information using the ocular region. Although the gender predication with ocular region is encouraging, the performance is significantly challenged when occluded by eyeglasses. In this chapter, we present the influence of eyeglasses on gender classification accuracy based on the ocular region. Unlike the earlier works that focus on gender classification in visible and near-infra-red spectrum, we present in this chapter, an extensive analysis of gender classification using the multi-spectral images of 104 ocular instances corresponding to total of 16,640 samples collected in eight narrow spectrum bands. Further, we introduce a new approach that selects four discriminative spectral band images based on the highest entropy value from the set of spectral images. The features are then extracted using Gabor filters to learn gender classification model using Probabilistic Collaborative Representation Classifier (ProCRC). The results are presented in the form of average classification accuracy using 10-fold cross-validation for random selection of training and testing samples. The highest average classification accuracy obtained based on the proposed approach is 75.72%, and we compare it with other state-of-the-art methods using individual spectral bands and fused spectrum bands.
