We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
다중 분류기 융합에 기반한 마스크 얼굴 표정인식.
- Authors
이승호
- Abstract
Facial expression recognition has been a core technology in affective computing and used in various applications such as analyzing the concentration of students for personalized education and detecting drowsy driving of drivers in smart cars. Since COVID-19 pandemic started and wearing masks became common, there has been increasing demand for recognizing masked faces. However, because most of the lower facial part is occluded when wearing mask, it is difficult to extract discriminative feature information. Thus, a sophisticated classifier is required. This paper proposes an expression recognition method for masked face, which fuses prediction scores from multiple complementary classifiers, i.e., CNN and optimal facial reconstruction based classifiers. Experiments were performed by synthesizing mask image into the face images of the CK+ and MMI which were the widely used datasets in facial expression recognition. The proposed method achieved the accuracies of 70% and 63%, respectively and 7% improvements were obtained compared to using the classifier adopted in the recently proposed method.
- Subjects
AFFECTIVE computing; FACIAL expression; AUTOMATIC systems in automobiles; COVID-19 pandemic; MEDICAL masks; SURGICAL equipment; AUTOMOBILE license plates; OXYGEN masks; AUTOMOBILE driving
- Publication
Journal of the Korea Institute of Information & Communication Engineering, 2023, Vol 27, Issue 7, p837
- ISSN
2234-4772
- Publication type
Article
- DOI
10.6109/jkiice.2023.27.7.837