Portable Facial Expression System Based on EMG Sensors and Machine Learning Models.

Area of Science:

• Computer Science
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Interpreting human emotions is a significant challenge for computers.
• Traditional methods like computer vision and electroencephalograms require substantial computational resources.
• Portable sensors offer a localized solution for capturing behavioral data.

Purpose of the Study:

• To develop a portable device for recognizing six primary human emotions using electromyography (EMG) sensors.
• To assess the feasibility of on-device emotion recognition with limited computational power.
• To highlight the importance of naturalistic data collection for emotion recognition.

Main Methods:

• Utilizing electromyography (EMG) sensors placed on specific facial muscles.
• Implementing a deep learning model on a CortexM0 microcontroller for emotion classification.
• Collecting and processing data from naturalistic environments.

Main Results:

• Achieved a 92% classification accuracy for recognizing six primary emotions (happiness, anger, surprise, fear, sadness, disgust).
• Demonstrated sufficient computational capability on the CortexM0 microcontroller for on-device deep learning model storage.
• Validated the necessity of naturalistic data collection and machine learning pipelines for effective emotion recognition.

Conclusions:

• Portable EMG sensor devices can effectively recognize primary human emotions with high accuracy.
• Edge computing with microcontrollers is viable for real-time emotion analysis.
• Naturalistic data collection and processing are crucial for robust emotion recognition systems.