Hybrid Approach of Haar Cascade Classifiers and Geometrical Properties of Facial Features Applied to Illumination Invariant Gender Classification System

This paper proposes a fast and efficient approach for gender classification under non uniform illumination variations. Haar Cascade Classifiers are used for face detection from an image. Facial feature extraction from detected face is done by using combined approach of Haar Cascade Classifiers and geometrical properties of facial features. Preliminary facial features viz. eyes, nose and mouth are extracted using Open CV Haar Cascade Classifiers. Further, geometrical properties of facial features are used for eyebrow detection. To further make our approach faster and reduce time complexity, we have used regions which contains moustache and beard. Weber illumination normalization technique is employed to compensate non uniform illumination variations from detected facial features. Support Vector Machine (SVM) is used as classifier for gender classification. Experimental results on Color FERET database and Caltech database show that the proposed approach improves gender classification rate upto 98.75 % along with significantly reduced computing time.