This paper presents TexonMask, a facial expression recognition system using lightweight electrode-augmented commodity facemasks. With a matrix of textile electrodes carefully deployed on a commodity mask, our edge computing system recognizes the wearer's facial expressions with machine learning based on the capacitive sensor readings, provides a wearable affective display and communicates with external devices using low bandwidth. Results from user studies show that the system is effective and efficient at recognizing five or ten facial expressions with an accuracy of around 90%, using a personalized classifier trained with only six data points per expression. The system's performance is stable across the use sessions and further improves when more data points are collected. We further developed two LiveEmoji applications for facilitating online and face-to-face communication of facemask wearers, demonstrated them in user interviews, and obtained positive participant feedback. Based on the results and findings of the study, we discuss implications and future research directions for facilitating emotional communication between facemask wearers and others.
https://doi.org/10.1145/3544548.3581295
The ACM CHI Conference on Human Factors in Computing Systems (https://chi2023.acm.org/)