A Study on Human Emotion Recognition Techniques

Emotions are essential for humans and play an important role in cognition. Emotion is frequently linked to logical decision-making, outlook, interactional behavior, and, to a lesser extent, human intelligence. With a growing interest in ensuring “emotional” interactions amid machines and humans, there is a need for reliable solutions in recognizing human emotional states. Humans typically express their emotions through a variety of non-linguistic and indirect methods. Emotion detection is accomplished primarily through the use of six distinct strategies: facial emotion recognition, physiological signal recognition, speech variation, emotion recognition from audio and videos, and emotion recognition from EEG signals on datasets like RAVDESS, JAFFE, CK+, TESS, DEAP, etc. In general, these methods identify seven basic emotions. In addition, we compared different ways to recognize emotions in humans. Deep convolutional neural networks produced the best results for Facial Emotion Recognition, physiological signals, audio, video, and EEG, emotion recognition through speech with accuracies of 97.83 %, 85%, and 70%, 75%, 80% respectively.