We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Influenced factors reduction for robust facial expression recognition.
- Authors
Sun, Zhe; Hu, Zheng-ping; Wang, Meng
- Abstract
The performance of facial expression recognition (FER) would be degraded due to the influenced factors such as individual differences and limited number of training samples. Therefore, reducing the influenced factors in facial images may be useful for improving the performances of FER. In this paper, we propose to reduce the influenced factors for robust FER. First, we reduce the influences of individual differences by the auxiliary neutral dictionary and obtain the feature space which highlights the expression features. Then we exploit the difference training samples to synthesize the virtual training samples to alleviate the influenced factors of the limited training samples. Third, we combine the difference dictionary with virtual training samples to form the extended dictionary and select the optimal training samples from the extended dictionary. Finally, we exploit the optimal training samples based ℓ2-norm representation algorithm for the classification.
- Subjects
FACIAL expression; HUMAN facial recognition software; IMAGE recognition (Computer vision); INDIVIDUAL differences; FACTOR analysis; STATISTICAL sampling; COMPUTER algorithms
- Publication
Multimedia Tools & Applications, 2018, Vol 77, Issue 13, p16947
- ISSN
1380-7501
- Publication type
Article
- DOI
10.1007/s11042-017-5264-y