Support Vector Machine Based Smile Recognition

Facial expression recognition is one of the most challenging research areas in the image recognition field and has been studied actively for a long time. However, it has not achieved enough performance under practical environments yet. Among all the expressions, smile is the most important facial expression used to communicate between the human beings. Therefore, if we can detect smile and also estimate its intensity at low calculation cost and high accuracy, it will raise the possibility of providing many new potential applications in the future. However, there exist many problems to be solved, such as problems caused pose change, scale change, appearance variations, illumination changes, and occlusions etc. This thesis proposes several algorithms to addressing these problems, such as 2D cascaded Boosting based face detection using Haar features, illumination preprocessing via histogram equalization, support vector machine-based smile recognition etc. The main work and contributions of this thesis are as follows.(1) This thesis presents a 2D cascaded Boosting based face detection method using Haar features. The method uses the integral image techniques to efficiently extract Haar features. Meanwhile, the method achieves face detection via a 2D cascaded Boosting algorithm, which makes the trained classifier can handle large-scale training data set.(2) This thesis proposes a smile recognition algorithm based on support vector machine. A face region is firstly divided into a set of cells. Then, the combined features from local intensity histogram and center-symmetric local binary pattern are used to train the support vector machine for recognizing smiles. Meanwhile, we estimate the smile intensity by using the posterior probability obtained by the support vector machine. Thus, the proposed method can effectively handle the pose change, scale change, appearance variations, illumination changes, and occlusions etc.