Feature Selection for Efficient Gender Classification

The study presents an efficient gender classification technique. The gender of a facial image is the most prominent feature, and improvement in the existing gender classification methods will result in the high performance of the face retrieval and classification methods for large repositories. In this paper a new efficient gender classification method is proposed. First, the face part of the image is segmented using Viola and Jones face detection technique which excludes unwanted area from the image, so reducing image size. Histogram equalization is performed to normalize the illumination effect. Discrete Cosine Transform (DCT) is employed for feature extraction and sorting the features with high variance. K-nearest neighbor classifier (KNN) is used for classification. The face images used in this study were obtained from the Stanford university medical student (SUMS) frontal facial images database. The experimental results on the SUMS face database indicate that the proposed approach achieves as high as 99.3% gender classification accuracy.