Beyond the Camera: Epidermal Sensor Platforms for Intelligent Facial Expression Recognition and Interactive Human Monitoring

Camera‐based computer vision remains the dominant approach to facial expression recognition (FER), but its performance depends on controlled conditions, including stable lighting, unobstructed views, and fixed camera angles, while continuous recording raises privacy concerns. Soft, skin‐conformable sensors offer an alternative by detecting facial muscle activity directly through mechanical deformation or biopotential signals. This review organizes recent epidermal FER hardware into four sensing mechanisms: piezoresistive, capacitive, electrophysiological, and self‐powered sensing. Multimodal systems are considered a complementary architecture that integrates two or more mechanisms. We examine how nanostructure and device design address key trade‐offs, including stretchability and conductivity in strain sensors, sensitivity and linear range in capacitive sensors, and power autonomy and signal fidelity in self‐powered transducers. We also survey applications in facial expression and emotion recognition, avatar and robotic control, and physiological and transportation safety monitoring. Although multimodal and self‐powered systems have improved recognition performance, practical deployment remains limited by poor cross‐user generalization, long‐term instability, and the lack of standardized benchmark datasets. Future progress will require more stable materials, lightweight on‐device algorithms, and learning strategies that adapt to individual users.