Facial Expression Recognition Based on Texture Features
DOI:
https://doi.org/10.29194/NJES.24020144Keywords:
Expression Recognition, Feature Extraction, Texture, LBP Variants, Image ProcessingAbstract
Facial expressions are a form of non-verbal communication, they appear as changes on the surface of the facial skin according to one's inner emotional states, aims, or social communications. Classification of these expressions is a normal process for humans, but it is a challenging task for machines.
Lately, interest in facial expression recognition has grown, and many systems have been developed to classify expressions from facial images. Any expression recognition system is comprised of three steps. The first one is face acquisition, then feature extraction, and finally classification. The classification accuracy depends primarily on the feature extraction step. Therefore, in this research we study many texture feature extraction descriptors and compare their results under the same preprocessing circumstances; moreover, we propose two improvements for one of these descriptors, which give better results than the original one. We validate the results on two commonly used databases for expression recognition using Matlab programming language, wishing all of that to be an interesting point for researchers in this field.
Downloads
Downloads
Published
How to Cite
Issue
Section
License
Author(s) Rights
- Each author retains the right to use the work for non-commercial purposes as well as for further research and spoken presentations.
- Each author retains the right to use the illustrations and research data in his/her future work.
- Only one offprint is provided free for each author. The authors can order offprints at the proof stage at certain rates depending on the number of additional copies required and the year of publication.
Publisher Rights
The publisher of the journal has full rights for publication of the submitted manuscripts, electronic and facsimile formats and for electronic capture, reproduction and licensing in all formats now and in perpetuity in the original and all derivative works.