EMOTION CLASSIFICATION USING MACHINE LEARNING METHODS

This article studies the problem of automatic classification of emotions using human facial expressions, and compares various machine learning algorithms, in particular k-NN, SVM, Random Forest, MLP and CNN, with the widespread FER2013, KDEF RAFDB and AffectNet databases and a single database created on their basis. Geometric and image-specific features were formed and recognition was performed using facial features and visual cues. The results obtained showed that the CNN model has high accuracy. It was also proved that data augmentation, ensuring class imbalance and selecting effective architectures are important factors in emotion classification. The results of this research can be used in areas such as human-computer interaction, education, and psychology.