A Cyber-Physical Framework for Smart Physical Education Based on Wearable IoT Devices

The recent progress in wearable IoT and sensor technologies has made it possible to continuously monitor physiological signals, opening doors for continuous fitness assessment and performance analysis within intelligent sports and health systems. Precise classification of fitness levels from such high-dimensional and time-series data is still a significant challenge because of data variability, noise, and poor parameter tuning in deep learning models. Trustworthy and automated categorization is essential for facilitating personalized training, avoiding injuries, as well as data-driven decision-making in contemporary cyber-physical learning environments. To tackle these challenges, this research outlines an RSO-BiLSTM framework, which incorporates Rat Swarm Optimization (RSO) for effective hyperparameter optimization with a Bidirectional Long Short-Term Memory (BiLSTM) network to extract intricate temporal dependencies from physiological signals. Based on the Collective Sports DB dataset, the model was tested on both binary (Unfit vs Fit) and multi-class (Low, Medium, High level of performance) classification problems. Experimental results indicated excellent predictive performance with 96.00% and 97.00% accuracies and macro F1-scores approximately equal to 0.90–0.91, affirming the robustness, scalability, and good generalization of the model. In summary, the suggested framework presents an efficient and interpretable method for real-time fitness assessment and performance classification in IoT-based smart sports systems.