| From Data to Decision The Role of Machine Learning in Assessing and Training Motor Skills |
| Paper ID : 1239-SPORTCONGRESS |
| Authors |
|
Meysam Rezaei *1, Maedeh Ahmadpoor2 1Assistant Professor,Deparment of physical education and sport sciences,MashhadBranch,Islamic Azad 2PhD Student, Department of Sports Behavioral and Cognitive Sciences, Faculty of Sports and Health Sciences, University of Tehran, Tehran, Iran |
| Abstract |
| Abstract Introduction: This review aimed to synthesize current evidence on how machine learning (ML) contributes to the assessment and training of motor skills across sports, rehabilitation, and human–computer interaction domains. Methods: This study employed a narrative descriptive review methodology to examine peer-reviewed articles published between 2020 and 2025 from databases including Scopus, PubMed, Web of Science, and IEEE Xplore. The inclusion criteria comprised studies applying ML algorithms—such as support vector machines, convolutional neural networks, recurrent neural networks, and reinforcement learning—to motor skill analysis, performance classification, and adaptive training. The review systematically categorized selected studies according to ML techniques, data modalities (e.g., video, EMG, IMU, EEG, kinematics), and application domains (e.g., clinical rehabilitation, sports, education). Descriptive synthesis was used to integrate conceptual, technical, and practical insights, emphasizing adaptive learning frameworks and multimodal sensor fusion. Results: The analysis revealed a clear trend toward hybrid and multimodal ML architectures that combine spatial–temporal modeling with real-time feedback mechanisms. Deep learning approaches, particularly CNN–LSTM hybrids, outperformed classical models in recognizing movement sequences and assessing motor quality. Reinforcement learning was found to be the dominant paradigm for adaptive and personalized feedback, especially in robotic and virtual reality (VR)-based rehabilitation systems. Predictive analytics models accurately estimated motor recovery trajectories, fatigue thresholds, and performance trends. Cross-disciplinary innovation—integrating biomechanics, neuroscience, and AI—enabled the development of intelligent feedback systems that translate raw sensor data into actionable insights for motor optimization. Conclusion: ML has revolutionized motor skill assessment and training by establishing adaptive, predictive, and personalized learning ecosystems. Its integration with robotics, VR, and brain–computer interfaces marks a paradigm shift toward continuous, data-informed motor improvement. The findings underscore the necessity of ethical, explainable, and equitable ML applications to ensure transparency and accessibility in future human–machine collaboration. |
| Keywords |
| Machine learning; motor skill learning; rehabilitation; biofeedback; deep learning; reinforcement learning; virtual reality; adaptive systems; human–robot interaction |
| Status: Abstract Accepted (Oral Presentation) |
