This study aims to develop a “sensing-actuation integrated” intelligent pressurized thigh band to assist the quadriceps, indirectly alleviate knee joint load, and achieve high-precision recognition of movement modes. The system comprises a portable integrated controller and a textile-integrated flexible pneumatic actuator. Experiments were conducted to evaluate the effects of different air bladder pressure conditions on metabolic rate and muscle activity. Simultaneously, pneumatic data corresponding to six common activities were collected, and a lightweight deep learning model was developed to enable high-precision motion classification. Finally, the model was deployed to an embedded platform to demonstrate its application potential. Results indicate that appropriate air bladder pressure significantly reduces quadriceps muscle activation and average metabolic cost. Furthermore, the deep learning model achieved 99.17% accuracy in recognizing the six activities and was successfully deployed to the embedded platform. This study validates the effectiveness of the intelligent pressurized thigh band in improving locomotor performance under static pressures and demonstrates the potential of air bladder pressure variations as a proxy indicator for movement intent for future closed-loop control.

Continuous monitoring of lower-limb kinematics in natural environments is essential for gait analysis and rehabilitation but remains challenging due to the limitations of optical systems and the inaccuracy of sparse inertial sensor methods. To address this, we propose a high-precision, minimalist wearable system utilizing only three inertial measurement units placed on the pelvis and shanks. In the data preprocessing stage, engineering modifications are made based on the traditional gradient descent algorithm to implement adaptive channel adjustment on the acceleration and magnetic data of a single IMU, aiming to alleviate the impact of motion acceleration and external magnetic interference on the temporal feature manifold. Subsequently, a pure Transformer neural network is utilized to capture long-range temporal dependencies, reconstructing full lower-limb kinematics without relying on rigid biomechanical assumptions. The model was optimized and deployed on an STM32N647 microcontroller to achieve real-time edge inference with a low latency of approximately 17 ms. Experimental results demonstrate that the proposed method achieves a mean absolute error of 2.41° for level walking, significantly outperforming traditional constrained Kalman filter approaches. Furthermore, it maintains high tracking robustness during complex nonlinear movements such as squatting and lunging. In conclusion, this edge-computing-enabled framework provides an accurate, comfortable, and real-time solution for unconstrained human motion capture in daily scenarios.

Wearable electronic systems require compact and sustainable power sources together with reliable motion-sensing functions. This study presents a gear-driven plantar energy harvester that integrates biomechanical energy conversion with self-sensing locomotion recognition. The device converts low-frequency vertical foot loading into rotary motion through a wedge–lever transmission and amplifies the rotational speed using a multistage gear train with a total transmission ratio of 12. A one-way bearing enables directional power transmission during loading and prevents reverse rotation during recovery. The generated voltage serves both as the electrical output and as the sensing signal for locomotion recognition. Human-subject experiments were conducted under six locomotion modes: walking at 1, 2 and 3 m/s; running; ascending; and descending. Voltage signals were sampled at 2000 Hz and segmented into overlapping sequences. A CNN–LSTM model was used to extract local waveform features and temporal dependencies from the nonstationary signals. The model achieved an overall recognition accuracy of 98.8%, with most errors occurring between ascending and descending. The results demonstrate that a single plantar device can simultaneously harvest biomechanical energy and provide motion-related information, offering a compact solution for integrated energy harvesting and self-sensing in wearable systems.