Recognizing Human Emotions Using a Convolutional Neural Network

Improvement of developments in recognizing human emotions is one of the tasks in the field of Internet behavior. Such developments can be used to determine a person's mood, workplace fatigue, and many other factors. This paper aims to study the performance of the convolutional neural network (CNN) in human emotion recognition and define the quality of recognition in relation to similar studies. In the theoretical part of the study, was analyzed the basic algorithms used in determining the emotions on the face of a person, as well as compare approaches to the application of different methods that could be applied to the data set used in solving the problem, where subsequently proposed options for improving accuracy. Algorithms and functions such as convolution, pooling, activation, batch normalization, and optimization are taken as methods of research. The results show that the implemented model has an average accuracy relative to similar models.