Gender detection from frontal face images

Abstract

In these modern days, gender recognition from facial image has been a crucial topic. To solve such delicate problem several handy approaches are being studied in Computer Vision. However, most of these approaches hardly achieve high accuracy and precision. Lighting, illumination, proper face area detection, noise, ethnicity and various facial expressions hinder the correctness of the research. Therefore, we propose a novel gender recognition system from facial image where we first detect faces from a scene using Haar Feature Based Cascade Classifier by Paul Viola and Michael Jones with the help of Adaboost technology. The face detection goal is achieved by OpenCV. After the detection of a face and noise is reduced using Histogram equalization. Finally, Deformable Spatial Pyramid (DSP) matching algorithm is used to match the processed facial image with the knowledge base containing classified male and female frontal face images. Our proposed system pulls out better accuracy than most of the modern techniques.