A Multi-Branch Adaptive Model with Hybrid Time—Frequency Loss to the Enhanced Joint Moment Prediction of Prosthetic Control and Human Motion Applications

Applied Sciences, Год журнала: 2025, Номер 15(4), С. 1678 - 1678

Опубликована: Фев. 7, 2025

In the fields of prosthetic control and rapid response prediction for human motion, accurate joint moments is crucial understanding simulating behavior. However, traditional time series models, especially when trained using small batches limited data single-time step predictions, frequently encounter difficulties in managing long-term dependencies. This deficiency significantly compromises their ability to generalize maintain predictive accuracy over extended periods. To address these challenges, an innovative model called Multi-Branch Adaptive Encoding (MAE) has been introduced. features adaptive weight module that employs a multi-branch input strategy dynamically allocate weights different surface electromyography (sEMG) signals angles, thereby optimizing processing sample data. Additionally, feature extraction encoder, named Simplified Feature Transformer (SFT) designed. encoder substitutes attention mechanisms with Multilayer Perceptron (MLP) omits decoder component enhance model’s efficiency offer significant advantages small-batch training capabilities. A Hybrid Time–Frequency Loss (HTFLoss) also introduced complement MAE model. approach enhances handle The HTFLoss demonstrate increase Variance Accounted For (VAF) 0.08 ± 0.03, reduction Root Mean Square Error (RMSE) 1.77 0.735, improvement coefficient determination (R²) 0.09 0.05, indicating substantial superiority. These enhancements highlight extensive potential applications rehabilitation medicine, human-machine interaction. improved manage dependencies make this particularly valuable designing advanced devices can better mimic natural limb movements, improving quality life amputees.

Язык: Английский

---

Sliding-Window CNN + Channel-Time Attention Transformer Network Trained with Inertial Measurement Units and Surface Electromyography Data for the Prediction of Muscle Activation and Motion Dynamics Leveraging IMU-Only Wearables for Home-Based Shoulder Rehabilitation

Sensors, Год журнала: 2025, Номер 25(4), С. 1275 - 1275

Опубликована: Фев. 19, 2025

Inertial Measurement Units (IMUs) are widely utilized in shoulder rehabilitation due to their portability and cost-effectiveness, but reliance on spatial motion data restricts use comprehensive musculoskeletal analyses. To overcome this limitation, we propose SWCTNet (Sliding Window CNN + Channel-Time Attention Transformer Network), an advanced neural network specifically tailored for multichannel temporal tasks. integrates IMU surface electromyography (sEMG) through sliding window convolution channel-time attention mechanisms, enabling the efficient extraction of features. This model enables prediction muscle activation patterns kinematics using exclusively data. The experimental results demonstrate that achieves recognition accuracies ranging from 87.93% 91.03% public datasets impressive 98% self-collected datasets. Additionally, exhibits remarkable precision stability generative tasks: normalized DTW distance was 0.12 normal group 0.25 patient when dataset. study positions as tool extracting features data, paving way innovative applications real-time monitoring personalized at home. approach demonstrates significant potential long-term function non-clinical or home settings, advancing capabilities IMU-based wearable devices.

Язык: Английский

---

Application Progress and Future Prospects of Smart Wearable Devices in Postoperative Rehabilitation of Lower Limb Fractures: A Mini Review

Current Problems in Surgery, Год журнала: 2025, Номер unknown, С. 101755 - 101755

Опубликована: Март 1, 2025

Язык: Английский

---