A wearing orientation-independent electromagnetic self-powered sensor for human activity recognition based on biomechanical energy scavenging

Abstract

Self-powered human activity recognition (HAR) using vibration energy harvesters attracts broad interest in wearable electronics. To address the challenge of making such sensors adaptive to human motions, we design a novel electromagnetic self-powered HAR sensor with an eccentric-rotor-based vibration pick-up structure. Its adaptability to the low-dominant-frequency (<5 Hz) and volatile-direction wrist movement is realized by the low resonant frequency and in-plane rotary responses of this magnetic rotor, promoting the biomechanical motion pick-up. Particularly, its capability of providing repeatable and human-activity-related voltage signals is revealed and proved wearing orientation-independent. On this basis, walking, jogging, and running at 2–8 km/h are recognized precisely by multi-feature-based signal representation and random forest models, achieving comparable classification accuracies over 97% with different wearing orientations. The practicality of this sensor is also enhanced by its light weight and small volume. The proposed sensor and signal analyzing method offer a promising and versatile solution to convenient HAR.