Facial Expression plays an important role in human communication, allowing people to express themselves without the use of any verbal means but still understanding each other's mood. Thus the interfaces must have the ability to detect any kind of change in the behavior of the user and to communicate based on the information available through interaction rather than the commands given by user. Thus, facial expression recognition is a challenging problem in computer vision. The project retrieves real-time images from a webcam and converts them to gray scale images. In facial feature expression recognition system we calculate 18 feature values from 16 feature points extracted from facial images. Then, these pre-defined feature vectors extracted from the images are sent to multilayer perceptron network for training and classification using back propagation. Using the result, the software will be able to develop human computer interaction and to judge the emotions of the user. The main aim is to work upon five different emotions - neutral, happy, sad, surprised and fear. Using only two-thirds of the total features, our approach achieves a classification rate (CR) which is higher than the CR obtained using all features. The system also outperforms several existing methods, evaluated on the combination of existing and self-generated databases. ",Feature vectors

covariance analysis