Human Gender Detection from Facial Image using Global and Local Feature

Abstract

Identifying gender is frequently difficult when physical characteristics show different facts, which can lead to public controversy. This paper describes a method for determining a person's gender solely based on physical characteristics, particularly facial shape and features. The method used combines global and local facial features. Face landmark detection is a global feature that displays a face's shape or contour in 2D. While the Histogram of Oriented Gradient (HOG) is a local feature that shows the detailed shape of facial components like eyebrows, eyes, nose, and mouth. The combined information from global and local features is fed into the Support Vector Machine (SVM) and classified as male or female. Based on the experimental results, our system was able to identify gender with an accuracy of 80% with an average F-measure is 0.8.