<p>Intelligent wearable devices equipped with biometric sensing technology enable personnel identification, health management, activity recognition, and location tracking, thereby substantially improving quality of life. However, current devices face challenges in achieving a sustainable energy supply due to the inefficient energy conversion from low-frequency human motion. Furthermore, conventional biometric sensing methods relying on machine learning (ML) often struggle to balance between high recognition accuracy and rapid processing speeds. To address these challenges, this work proposes a self-powered wearable system integrated with a one-way electromagnetic harvester (OW-EH), which simultaneously enables energy harvesting and motion sensing. Employing an integral component extraction (ICE) ML model, the OW-EH device achieves effective recognition of personnel identities and associated motion patterns. Compared with conventional ML approaches, the ICE model improves recognition accuracy by 20.7% and reduces processing time by 54.12%, demonstrating its suitability for real-time wearable applications. Bench tests demonstrate that the compact OW-EH device generates an average power output of 206.5 mW, corresponding to normalized power densities of 3.40 mW Hz⁻² cm⁻³ and 0.96 mW Hz⁻² g⁻¹. This output power level satisfies the operational requirements of typical portable electronics and supports self-powered navigation and triaxial acceleration monitoring during daily activities at speeds ranging from 4 to 8 km h⁻¹. Notably, the system achieves a personnel identification accuracy of 95% and a motion detection accuracy of 90.39% using the ICE framework. Supported by ML, the OW-EH offers a sustainable and privacy-preserving solution for personalized biometric sensing, laying the solid foundation for digital twin systems to enhance intelligent living.</p>