Facial expression recognition using image orientation field in limited regions and MLP neural network

In this paper, a novel approach for facial expression is introduced by using face orientation field concept. This method is not sensitive to the obstacles such as hair, eyeglasses and surrounding illumination. The computational tools of this paper have been used in previous studies for fingerprint recognition systems. However, we have adopted them for facial expression recognition. The features which extracted from orientation field are applied to the MLP neural network and this network select facial expression among six universal expressions. The presented method, have been tested on the JAFFE database and reached the average accuracy of 90.33% in recognition of facial expressions.