Recognition of Facial Patterns Using Surface Electromyography—A Preliminary Study

Abstract

Facial expressions are considered as a universal language method with non-verbal communication between humans. The surface Electromyography (sEMG) is a technique to acquire and processing signals obtained from the electrical activity of the contraction of voluntary muscles. From that, this work presents a preliminary study of recognition of facial expression obtained from sEMG signals. Signals from 4 subjects were acquired for six basic facial expressions that expresses the following emotions: happiness, surprise, sadness, angry, disgust, and fear. The sEMG signals were acquired from two muscles: zygomaticus major and corrugator of the eyebrow. Three feature sets for sEMG and four different classifiers were evaluated. The best combination was found with the TD9 feature set and SVM with radial basis function kernel classifier, reaching 79.8% of accuracy. Among the expressions, the best recognition was happiness, with individual accuracy of 99.2%. The most difficult expression to classify was the expression of disgust.