Facial Expression Recognition of Public Speaking Anxiety

Public Speaking Anxiety (PSA) is one of the most universal subtypes of social anxiety, and the facial expression recognition of PSA is an immediate area of research focus. The experiment obtained the facial expression data of 18 postgraduates in their thesis defense for the master's degree. Then, valid data were selected by using self-evaluation of the subject and the average evaluation of 5 audience. Next, the paper used Active Shape Model (ASM) algorithm for feature extraction of 68 feature points in 5 facial zones. Moreover, the paper applied support vector machine algorithm to recognize the social anxiety state according to the displacements of the feature points, and backward selection (BS) algorithm was used to find out the most important features in recognition. The result illustrated the validity of the method and obtained the best correct rate of 99% in the recognition of social anxiety state. Meanwhile, the maximum displacement zone in social anxiety state is jaw, and followed by mouth, eyebrow, eyes and nose. Obvious individual differences were found in the formation of anxious facial expression.