Using edge orientation histograms in face-based gender classification

In this paper we present our evaluation of the Edge Orientation Histograms (EOH) as feature descriptors in an automatic face-based gender classification application. The feature descriptors extracted from an input image are evaluated using estimated arithmetic means of accuracies to select the feature descriptors that play the most important role in classification success. Our experiments show that features corresponding to the jawline of the subject play the most important role, yielding an average classification accuracy of up to 86%.