Abstract
Facial expressions play a main role in determining the internal impressions of the human. Therefore, dealing with these expressions by machine learning attract the attention of many researches. However, the large number of extracted features of faces is a hard task in this process due to the numbers of irrelative and repeated features among them. This paper tries to decrease the extracted feature of facial expression images by deleting repeated features to increase the performance and decrease the computation time of the recognition task. Therefore, in this paper, a dataset of students' facial expression is used, and the binary Sine-Cosine Algorithm (SCA) is applied to select the most relative features. The experiment results showed that, the SCA reduced the features number by 87% as well as it achieved the high accuracy results and outperformed all other methods.