K-means approach to facial expressions recognition

A method is proposed to recognize facial expressions. The method used two simple features to recognize the expressions which are the density of pixels and the ratio of height to width of cropped boundary regions. The system first applies some preprocessing stages to enhance the input image and reduce the noise. The face boundary will then be detected. The region of interest (i.e. mouth and eyes) will be determined, from which, features will be extracted. Finally based on the features extracted, the face will be classified into one of three different classes using the K-means method. The method was applied and tested on a dataset of 200 images of faces and the success rate obtained was 76.5%.