A novel approach for face expressions recognition

In this paper a new method for face expression recognition is presented. Haar functions are used for face, eyes and mouth detection; edge detection for extracting the eyes correctly, and finally, Bezier curves to approximate the extracted regions. Then, a set of consecrated distances for each face type is extracted, set that will serve as training input for a multilayer neural network. We analyze the input data using a feed-forward neural network, trained and used for determining the class (Angry, Disgust, Fear, Happy, Neutral or Sad) of an arbitrary facial expression.