GENDER CLASSIFICATION FROM FACIAL IMAGES USING SUPPORT VECTOR MACHINE

The face is a biometric feature of a person. This biometric feature encourages researchers to model it. This unique features modeled and applied in many objects. One of them is used to classified the gender of someone like done in this research. There were four main step, consist of pre-processing using a histogram of equalization, face detection using LBP Haar cascade classifier to get face region, feature extraction, and classification. This extraction used to obtain an unique information named features. The method to get the features named Histogram of Gradient (HOG). After setting the parameters in extraction method we will get many features named HOG features. Meanwhile, The last step is classification using Support Vector Machine (SVM) classify gender form facial images. SVM was chosen because it�s one of classification method that robust in classify two group of data. The type of the kernel that used in this research is the linear kernel. This kernel has an advantage in the robustness to classify the big features. Implementation result show that the best achieved when HOG parameters (cell size = 20x20 pixels, block size = 5x5 pixels, block stride = 5x5 pixels, bin = 5 or 7) and SVM parameters (C=1, epsilon = 1e-6 dan iteration = 10.000). The accuracy of this proposed method is 92%.