Facial action coding-based facial sub-structures for anxiety emotion classification

Abstract

Most stroke patients usually have problems in communication and body movement, such as speaking, sitting, walking, and picking up items. Moreover, the number of caregivers is smaller than the number of stroke patients. Nevertheless, these patients need 24-hour caregivers for the patients’ safety. Therefore, the objective of this research is to determine patterns of anxiety emotion via facial expressions from the sub-structures on the face, such as the inner brow raiser, brow lower, lid raiser, and lip part. Random samples of 360 facial images from horror-thriller movies based on the internet movie database (IMDb) website were selected. Then, 68 facial landmarks for classifying the emotions were applied to each facial image. The differences in these 68 positions before and after the changed emotions were used as the emotional indicators. Furthermore, these different values are applied to implement a decision tree with all the boundaries of the sub-structures disclosed as a suitable classification model for emotion detection. Consequently, the accuracy when applying this decision tree with other facial images is 83.33%.