Detection of Facial Components Based on SVM Classification and Invariant Feature

Facial features determination is essential in many applications such as personal identification several approaches have been proposed, but an effective method for face detection is still a research problem. In this paper we focus on a recent method called the support vector machines (SVM) has been adapted and applied to the problem of pattern recognition such as face detection. The idea information of the skin color is used to reduce the search region and the main idea based on SVM is to project the data input space (belonging to two different classes) non-linearly separable in a larger space called feature space so that data are linearly separable. About fusion a non-parametric model is applied for the segmentation of the pixels of skin color. This last is used to reduce area of research within the image. However the SVMs help us to find exactly the faces in the segmented area. We implemented the SVM using a RBF kernel as a classification technique for face detection by block" approach of considering the face as a set of components (eyes, nose and mouth). The method succeeds in locating facial features in the facial region exactly and is insensitive to face deformation. The method is executable in a reasonably short time.