Facial Gender Recognition Using Eyes Images

Abstract: In the last decade, computer vision and pattern recognition field draw more attention; Facial gender classification can significantly improve human identification. It is useful in many applications that make use of the gender information. This paper proposes a novel feature extraction technique that used only the eye and eyebrow region to extract the features will be used in the gender classification. The proposed technique was consisting of several steps. The first step was to crop the eye area from the image as a pre-processing. Then apply one of the feature extraction methods: 2D-Wavelet Transform, Gray Level Co-occurrence Matrix and Discrete Cosine Transform. Finally, use SVM in the classification step to get the results. The proposed method obtained accuracy rate of 99.49 % on gender recognition using 2D-Wavelet Transform, accuracy rate of 98.49 % on gender recognition using GLCM and 99.62 % with DCT on Faces94 database.

Keywords: Gender recognition, feature extraction, 2D-Wavelet Transform, Gray Level Co-occurrence Matrix, Discrete Cosine Transform, Support Vector Machine