Emotion Recognition with Forearm-based Electromyography

TL;DR

This study demonstrates that forearm-based electromyography signals can effectively classify human emotions, achieving 88.1% accuracy, opening new avenues for emotion-aware human-computer interaction and assistive technologies.

Contribution

The paper introduces a novel approach using forearm EMG signals for emotion recognition, with a new dataset and classification method achieving high accuracy.

Findings

Forearm EMG signals can distinguish between Relaxed and Angry states.

Support Vector Machine achieved 88.1% accuracy in emotion classification.

Potential applications include gaming and e-learning systems that adapt based on detected emotions.

Abstract

Electromyography is an unexplored field of study when it comes to alternate input modality while interacting with a computer. However, to make computers understand human emotions is pivotal in the area of human-computer interaction and in assistive technology. Traditional input devices used currently have limitations and restrictions when it comes to express human emotions. The applications regarding computers and emotions are vast. In this paper we analyze EMG signals recorded from a low cost MyoSensor and classify them into two classes - Relaxed and Angry. In order to perform this classification we have created a dataset collected from 10 users, extracted 8 significant features and classified them using Support Vector Machine algorithm. We show uniquely that forearm-based EMG signal can express emotions. Experimental results show an accuracy of 88.1% after 300 iterations.This shows significant opportunities in various fields of computer science such as gaming and e-learning tools where EMG signals can be used to detect human emotions and make the system provide feedback based on it. We discuss further applications of the method that seeks to expand the range of human-computer interaction beyond the button box.