Research On Children’s Social Emotional Competence Evaluate Based On Facial Expression

Children’s social emotional competence has an important significance to children’s personality training and comprehensive development. The important reference of evaluating children’s social emotional competence can be provided by getting the emotional state of children accurately and in time. As a biometric signal which can be easily acquired, facial expression can reflect the true emotional state of children more accurately. Therefore, facial expression recognition is an important access to acquire the emotional state of children. However, the existing work of facial expression recognition mostly focus on adults but rarely on children. In this paper, to deal with the problems of children’s facial expression recognition, the research of building children’s expression database, expression feature extraction, facial expression recognition was carried out. The main work of this paper can be summarized as follows:(1) Build a children’s expression database. To be able to induce children’s facial expression effectively, two classic psychology experiments were designed. During the experiments, a camera was used to capture children’s facial area, and record the whole process of children’s facial expression changes. Finally, the captured facial expression was marked with expression labels to build a children’s expression database, and the baseline of this expression database was given.(2) Propose a series of methods of spatio-temporal feature extraction in actual scene. These feature extraction methods can effectively deal with the common problems of facial expression recognition in actual scene such as complex illumination and background, non-frontal face, and face occlusion. Finally, the experiments were conducted on several expression databases to verify the effectiveness of the proposed feature extraction methods.(3) Propose a transfer double sparse learning (TDSL) model to deal with the problem of cross database facial expression recognition based on adults and children. TDSL can effectively eliminate the distribution difference between source domain samples and target domain samples by selecting some of the target domain samples as the auxiliary set and the source domain samples to train a model. In particular, during the process of learning TDSL model, TDSL can not only choose the facial sub-regions which contribute more to expression recognition, but also further select the expression feature in these facial sub-regions. The results of experiments showed that TDSL is able to solve the problem of cross database facial expression recognition based on adults and children.(4) Build a children’s emotion monitor system. This system can locate the facial region of children and predict the labels of expressions in actual scene. At the same time, this system can assist the researchers to evaluate children’s social emotional competence by transferring the result to emotional map to monitor children’s emotion in a period of time.