Highly Robust and Wearable Facial Expression Recognition via Deep-Learning-Assisted, Soft Epidermal Electronics

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Wearable Technology

Background:

• Facial expression recognition (FER) is vital for understanding emotions but limited by computer vision requirements and rigid devices.
• Existing systems struggle with dynamic, curvilinear faces and environmental constraints like lighting and occlusion.

Purpose of the Study:

• To develop a robust, highly wearable FER system using soft epidermal electronics and deep learning.
• To overcome the limitations of traditional FER systems by enabling high-fidelity biosignal acquisition without hindering natural expressions.

Main Methods:

• Utilized deep-learning-assisted soft epidermal electronics that conform to the face.
• Enabled high-fidelity biosignal acquisition, freeing constraints of movement, space, and light.
• Employed deep learning to enhance recognition accuracy with small sample sizes.

Main Results:

• The wearable FER system demonstrates high accuracy and robustness across various conditions (light, occlusion, poses).
• Achieved superior performance compared to traditional computer vision FER, especially for individual, dynamic recognition.
• Successfully applied to human-avatar emotion interaction and verbal communication disambiguation.

Conclusions:

• The proposed wearable FER system offers wide applicability and high accuracy for individual emotion recognition.
• It complements computer vision FER by providing a solution for dynamic, unconstrained environments.
• Presents promising advancements for human-computer interaction applications.