Comparative Analysis of Emotion Detection from Facial Expressions and Voice Using Local Binary Patterns and Markov Models: Computer Vision and Facial Recognition

Abstract

Emotion detection has been achieved widely in facial and voice recognition separately with considerable success. The 6 emotional categories coming out of the classification include anger, fear, disgust, happiness and surprise. These can be infered from one's facial expressions both in the form of micro and macro expressions. In facial expressions the emotions are derived by feature extracting the facial expressions in different facial poses and classifying the expression feature vectors derived. Similarly automatic classification of a person's speech's affective state has also been used in signal processing to give insights into the nature of emotions. Speech being a critical tool for communication has been used to derive the emotional state of a human being. Different approaches have been successfully used to derive emotional states either in the form of facial expression recognition or speech emotional recognition being used. Less work has looked at fusing the two approaches to see if this improves emotional recognition accuracy. The study analyses the strengths of both and also limitations of either. The study reveals that emotional derivation based on facial expression recognition and acoustic information complement each other and a fusion of the two leads to better performance and results compared to the audio or acoustic recognition alone.