Toward Instantaneous Facial Expression Recognition Using Privileged Information

The Facial Expression Recognition is one of determining factors in automatic recognition of humans' emotion. There are many works done in Facial Expression Recognition research, yet it has been difficult to build real-time and robust systems. This paper proposes the use of a classifier trained by Extreme Learning Machine to solve the speed issue, combined with the use of privileged information to improve the testing time and the reduction of the testing error. The study is done on features differently extracted from still facial datasets. The experimental results show that the proposed method improves the testing time making it feasible in real-time applications and more stable than the classical method, with time improvement reaching 25%.