In this modern era, technology is growing very rapidly, one of them is face recognition. Face recognition is an important research in the field of pattern recognition and has achieved successes in recent years [1]. A face recognition system is capable of generating a variety of information about a person's identity quickly and accurately. One of them, face recognition is able to provide information about the gender (male or female) of each person. Gender classification has become an area of extensive research due to its increasing application in existing human-computer interaction (HCI) systems, advertising, biometrics, surveillance systems, content-based indexing and searching [2]. It is therefore important to do research gender classification in order to support its role in various fields and to produce an improved level of accuracy. This paper presents face recognition for gender classification using Support Vector Machine (SVM). Support Vector Machines are a system for efficiently training the linear learning machines which can be used for as powerful classification methodology [3].

 

In this research, we used an image from Computer Vision Science Research Projects Facial Images Collection. The data is held in four directories (faces94, faces95, faces96, and grimace). We using face recognition dataset from directories faces94. The data used amounted to 200 face images. The collection of face images consists of 100 male face images and 100 female face images with 180 x 200 dimensions and JPG image format. This data collection technique is known as image acquisition. Then, do image pre-processing, the face images with dimension 180 x 200 converted into a square shape with the dimensions of the image 125 x 125. After the face images are square, the image of the face with JPG format is converted into a form of the matrix by using the algorithm that built-in Matlab program with the aim to facilitate the classifying process. The goal of this research is to see whether Support Vector Machine (SVM) with polynomial kernel and RBF can produce the best accuracy for gender classification. The result present in the table (please, check the results table on supplementary file we have uploaded).

 

In this research, we have obtained face recognition accuracy rates for gender classification using Support Vector Machine (SVM) with different kernels. As can be seen in Table 1 (please, check the results table on supplementary file we have uploaded), for the training data of 10 to 30 percent, the SVM method with the polynomial kernel has better accuracy than the SVM method with the RBF kernel. Furthermore, for training data 40 to 90 percent, SVM method with RBF kernel and also polynomial has achieved the same accuracy that is 100 percent.