Facial Emotion Recognition Based on Cascade of Neural Networks

The chapter presents a method that uses the cascade of neural networks for facial expression recognition. As an input the algorithm receives a normalized image of a face and returns the emotion that the face expresses. To determine the best classifiers for recognizing particular emotions one- and multilayered networks were tested. Experiments covered different resolutions of the images presenting faces as well as the images including regions of mouths and eyes. On the basis of the tests results a cascade of the neural networks was proposed. The cascade recognizes six basic emotions and neutral expression.