Automatic Facial Expression Analysis

Facial expression analysis is an active area in human-computer interaction. Many techniques of facial expression analysis have been proposed that try to make the interaction tighter and more efficient. The essence of facial expression analysis is to recognize facial actions or to perceive human emotion through the changes of the face surface. Generally, there are three main steps in analyzing facial expression. First, the face should be detected in the image or the first frame of image sequences. Second, the representation of facial expression should be determined, and the data related to facial expression should be extracted from the image or the following image sequences. Finally, a mechanism of classification should be devised to classify the facial expression data. In this article, the techniques for automatic facial expression analysis will be discussed. The attempt is to classify various methods to some categories instead of giving an exhausted review. The rest of this article is organized as follows. Background is presented briefly firstly. The techniques used in the three steps, which are detecting the face, representing the facial expression, and classifying the facial expression, are described respectively. Then some facial expression databases are discussed. The challenges and future trends to facial expression analysis are also presented. Finally, the conclusion is made.