This paper presents the design, development, and validation of an innovative soft robotics-based elbow rehabilitation device. Unlike conventional rigid passive rehabilitators, the proposed system prioritizes safety in human–robot interaction while introducing a novel approach to enhancing elbow mobility. Functional tests were systematically conducted on pouch motors to assess their suitability for integration into the rehabilitator. The device achieved an elbow range of motion of 0°–90° and generated an optimal force of ~98 N at 2 bar of pressure. The experimental results revealed a distinct nonlinear performance behavior, which was thoroughly analyzed to refine system operation and modeling. The study establishes a comprehensive framework encompassing actuator manufacturing, experimental testing, dynamic system identification, and open-loop control validation. Collectively, these stages culminate in the validation of the proposed dynamic model. Overall, this work represents a meaningful advancement in rehabilitation technology, providing a safer and more adaptable alternative for elbow rehabilitation, with strong potential for application across diverse healthcare settings.
Background This study compared lower‐limb kinematics and electromyographic characteristics between outside hitters (OH) and middle blockers (MB) during volleyball spike take‐off, aiming to delineate position‐specific joint‐angle profiles and relative muscle activation contributions. Methods Twelve national first‐class male volleyball athletes were recruited (OH, n = 6; MB, n = 6), and three‐dimensional kinematics (240 Hz) and surface electromyography (EMG; 2000 Hz) were synchronously collected. The analysis window was defined from right‐foot contact in the final approach step to bilateral toe‐off (T1–T4), with the braking phase set as T1–T3 and the propulsion phase as T3–T4; T3 was treated as the transition boundary between the two phases rather than as an independent phase. All participants were right‐handed hitters; therefore, the right side was defined as the dominant hitting side and the left side as the non‐dominant/non‐hitting side. Sagittal‐plane hip, knee, and ankle joint angles and phase‐specific relative iEMG contribution percentages of seven muscles were calculated and compared between groups. Results OH exhibited a smaller right/dominant hitting‐side hip angle than MB at T1, indicating a more flexed sagittal‐plane hip posture ( p < 0.01), and a smaller left/non‐dominant hitting‐side hip angle during T2–T3 ( p < 0.05). The right/dominant hitting‐side knee angle was consistently smaller in OH than in MB throughout T1–T3, indicating greater knee flexion ( p < 0.01). For the ankle joint, OH showed a greater left/non‐dominant hitting‐side sagittal‐plane ankle included angle during T1–T3 ( p < 0.01), whereas MB showed a greater right/dominant hitting‐side sagittal‐plane ankle included angle at T2 and T3 ( p < 0.01). During braking, OH showed a right/dominant hitting‐side iEMG contribution of 68.90% and a left/non‐dominant hitting‐side contribution of 31.09%, whereas MB showed 65.20% and 34.79%, respectively. During propulsion, OH primarily relied on the left/non‐dominant hitting side (59.40% vs. 40.60%), whereas MB demonstrated a more balanced bilateral activation pattern (49.63% vs. 50.37%). Conclusion OH and MB exhibited distinct lower‐limb kinematic–EMG coupling patterns during spike take‐off. OH showed greater sagittal‐plane hip and knee flexion in the early take‐off phase and a force‐production shift from right/dominant hitting‐side braking to left/non‐dominant hitting‐side propulsion, whereas MB demonstrated relatively larger joint angles and a more balanced bilateral activation pattern. These findings support position‐specific take‐off technique training and neuromuscular control strategies.
Background:This study aimed to examine the effects of a 12-week cadence control training program on patellofemoral joint mechanics and lower-limb joint work distribution in runners, thereby providing a theoretical basis for optimizing running strategies and reducing running-related injury (RRI) risk. Methods:A total of 44 recreational runners were randomly assigned to a control (n = 22) and an experimental group (EG; n = 22). The former performed conventional running training, while the latter underwent cadence control training for 12 weeks. Three-dimensional (3D) kinematic and kinetic data were collected before and after the intervention using the Simi Motion capture system and Kistler force platforms, respectively. Patellofemoral joint mechanical parameters and joint work contribution ratios were subsequently calculated. Results:Posttraining, the EG exhibited a significant increase in cadence and positive ankle joint work contribution and a significant reduction in patellofemoral joint stress (PFJS), compared with pretraining values. Conclusions:Twelve weeks of cadence control training significantly modified the patellofemoral joint mechanics and joint work distribution in runners. An increased cadence reduced patellofemoral joint loading and induced a distal shift in the relative contribution of lower-limb joint work from proximal to distal segments, which may represent a biomechanical mechanism underlying the alleviation of knee joint load. Therefore, cadence-increasing running strategies are recommended, particularly for runners experiencing patellofemoral pain (PFP), to mitigate the risk of knee joint injury.
This study is the first to extensively analyze the application of artificial intelligence (AI) to human activity recognition (HAR) in the field of sports and exercise through bibliometrics and visualization methods. We used CiteSpace, VOSviewer, and Bibliometrix R‐package software to perform a bibliometrics and visual analysis of 432 articles on the application of AI in HAR in sports and exercise from 2001 to 2024. During the period 2001–2024, the number of publications on AI in HAR in sports and exercise showed a steady and then rapid growth, and the number of articles published after 2018 rose year by year. The top 10 journals have 157 publications, accounting for 36.3%, of which SENSORS ranks first in terms of publications and citations. Jalal and Khan tied for first place in terms of number of publications. China ranked first in the number of publications, accounting for 44.7% of the total. Beijing Sport University and Princess Nora Bint Abdulrahman University are tied for the top spot in terms of the number of articles published in this field. High‐frequency keywords include “deep learning,“ “human activity recognition,” “action recognition,” “machine learning,” “wearable sensors,” and so on. Technical performance analysis of highly cited literature reveals that deep learning models achieve over 90% recognition accuracy on sports‐specific datasets, significantly outperforming traditional methods. Future research should focus on further improving the accuracy and real‐time performance of HAR technology, especially the ability to be applied in complex scenarios. Technical adaptation solutions for scenarios including individual routine sports, team confrontations, and equipment‐involved sports must be refined to overcome practical bottlenecks such as multimodal data synchronization and small‐sample data scarcity. At the same time, interdisciplinary cooperation should be strengthened to combine AI technology with sports science, psychology, physiology, and other fields to explore the application of HAR in emerging fields, while promoting the synergistic optimization between algorithm lightweight and wearable hardware performance. In addition, attention should be paid to AI ethics and data privacy protection to ensure the safety and legality of the technology application, promoting the sustainable development and widespread application of HAR technology in the sports field.
