This study proposes a magnetoelastic strain sensor for highly deformed posture perception. The sensor can maintain stable sensing signal output under significant strain (200%) and at varying stretch rates (ranging from 1000 to 8000 mm min⁻¹). The sensor design is simple and can be applied to soft robots and wearable devices. The characteristics of the magnetoelastic sensor can be adjusted by varying the weight ratio of the magnetic powders, which were evaluated experimentally to determine their influence on the mechanical properties and sensing characteristics of the sensors. The experimental results indicate that while the stiffness, sensing signal response, and signal‐to‐noise ratio of the sensor increase with the weight ratio of the magnetic powder, there is a concomitant decrease in mechanical stability. Therefore, the weight ratio of the magnetic powder may require optimization for different scenarios. Notably, the sensing signal remained consistent across different stretching rates. Practical applications of the magnetoelastic sensors were demonstrated through integration with a soft actuator, achieving an root‐mean‐square error (RMSE) of 2.27° over a 90° bending range for bending angle estimation, and onto a knee brace, achieving an RMSE of 3.25° over a 50° range for knee flexion/extension angle measurements at normal walking speeds.

Respiratory muscle training can improve respiratory function by strengthening muscle mass, which is of great help to populations with respiratory system diseases and athletes. Existing respiratory muscle training methods rely on resistance that hinders breathing, and the resistance cannot be adjusted automatically. However, the detection of the user's current muscle fatigue state and precise adjustment of resistance during respiratory muscle training are crucial to training efficiency. Here, we have developed a hybrid sensor that combines a triboelectric nanogenerator and a piezoelectric nanogenerator. This hybrid sensor can simultaneously collect both high‐frequency and low‐frequency signals generated by the Karman vortex street effect with low hysteresis. When the airway height is 30 mm, the sensor size is 52 μm × 40 mm × 17 mm, the output performance of the sensor is optimal, and the minimum response amplitude for the sensor is approximately 3 mm. Under normal breathing conditions, the output peak voltage is 7 V, the current is 100 μA, the charge transfer amount generated by one movement is 55 nC, the response time is 0.16 s, and the sensitivity is 0.07 V/m·s⁻¹. With the help of the principal component analysis algorithm, features related to the fatigue state of muscles were extracted from the collected signals, and the accuracy rate can reach 94.4%. Subsequently, the stepper motor will rotate to adjust the resistance appropriately. We fused the hybrid sensor, machine learning, control circuits, and stepper motors and fabricated a resistance self‐adaptation program. Our findings inspire researchers in the field of rehabilitation and sports training to evaluate training status and improve training efficiency. image

Stroke is a leading cause of long-term disability worldwide, necessitating efficient motor function assessment to guide personalized rehabilitation. The Fugl–Meyer Assessment for the Upper Extremity (FMA-UE) is a clinical gold-standard tool, but it is time consuming and requires trained clinicians, which limits its frequency of use and accessibility. While wearable sensors and deep learning offer promising avenues for remote assessment, accurately estimating detailed sub-scores of specific motor functions remains a significant challenge. This work introduces a deep learning framework for automated estimation of FMA-UE total and subdivision scores. Data was collected from 15 participants using four inertial measurement units (IMUs) positioned on the arm and trunk. Each participant performed seven specialized functional motions designed for comprehensive joint synergy involvement within ten minutes. A therapist-rated FMA-UE provided true scores. The proposed model leverages the integration of an LSTM-based autoencoder and mixup augmentation to enhance generalization and robustness. Evaluated through a leave-one-subject-out cross-validation (LOSOCV), the estimator demonstrated strong performance, achieving R² values exceeding 0.82. Pearson’s correlation coefficient r was more than 0.90, and the normalized root-mean-square errors (NRMSE) were below 0.14 for all subparts (A–D). Crucially, the total FMA-UE score was estimated with an NRMSE of 0.0678. These results show that a concise, sensor-based assessment can reliably predict detailed motor function scores.

The pediatric period is a crucial window for motor function learning and growth. Individuals with central nervous system injuries like cerebral palsy commonly display severe crouch gait in the lower limbs. Hyperflexion of the knee joints promotes the forward trunk and increases reliance on the handle frame of a walker for support. In this study, we developed a quantitative partial body weight training system integrated with a soft pneumatic exosuit to assist the knee extension during the stance phase of the gait cycle. In the preliminary results for five pediatric cerebral palsy subjects, compared to the baseline condition, excessive knee flexion ameliorated with the assistance of the soft pneumatic exosuit. The peak knee extension and range of motion increased by 19.72° (±3.47°) and 15.46° (±5.06°), respectively. With exosuit assistance, the subjects demonstrated improved gait retraining compared to baseline. They were able to bear significantly more body weight on their affected limb, as evidenced by a 33.3% increase in the fraction of body weight measured by the force plate. Additionally, they relied less on the handrail for support during walking. With more extended knee joints to bear the load over gravity, the pediatric subjects transferred the reliance from external support and upper limbs back to the lower limbs as a more independent status during the loading response to terminal stance.