MeciFace: Mechanomyography and Inertial Fusion-Based Glasses for Edge Real-Time Recognition of Facial and Eating Activities

The prevalence of stress-related eating behaviors and their profound impact on overall health underscore the urgent need for effective and widespread monitoring systems. To handle these issues, we introduce MeciFace, an innovative wearable technology meticulously crafted to monitor facial expressions and eating activities in real-time on-the-edge (RTE). MeciFace represents a significant advancement in health monitoring by offering a low-power, privacy-conscious, and remarkably accurate solution to promote healthy eating behaviors and facilitate stress management. At its core, MeciFace leverages lightweight convolutional neural networks as backbone models for both facial expression recognition and eating activity monitoring scenarios. One of the standout features of the MeciFace system is its commitment to efficiency, achieved through meticulous optimization resulting in a tiny memory footprint ranging from 11 to 19 KB. This ensures seamless operation without undue strain on resources, making it ideal for continuous usage in real-world scenarios. MeciFace demonstrates outstanding performance in rigorous RTE evaluations, achieving an F1-score of \(\ge 86\%\) for facial expression recognition and an impressive 94% for eating/drinking monitoring, for the RTE of unseen users. These results are obtained in a user-independent case, underscoring the robustness and generalizability of the system across diverse user profiles.