Real-time Algorithms for Facial Emotion Recognition: A Comparison of Different Approaches

Abstract

Emotion recognition has application in various fields such as medicine (rehabilitation, therapy, counseling, etc.), e-learning, entertainment, emotion monitoring, marketing, law. Different algorithms for emotion recognition include feature extraction and classification based on physiological signals, facial expressions, body movements. In this paper, we present a comparison of five different approaches for real-time emotion recognition of four basic emotions (happiness, sadness, anger and fear) from facial images. We have compared three deep-learning approaches based on convolutional neural networks (CNN) and two conventional approaches for classification of Histogram of Oriented Gradients (HOG) features: 1) AlexNet CNN, 2) commercial Affdex CNN solution, 3) custom made FER-CNN, 4) Support Vector Machine (SVM) of HOG features, 5) Multilayer Perceptron (MLP) artificial neural network of HOG features. The result of real-time testing of five different algorithms on the group of eight volunteers is presented.