Opto-driven multimodal DRL framework for smart healthcare in sports training

The rapid advancement of optical sensing technologies and wearable sensors has enabled the collection of multimodal data streams, including biomechanical, physiological, and environmental parameters, to optimize smart healthcare in sports training. However, existing approaches often fail to effectively integrate these heterogeneous data sources, limiting their ability to provide adaptive and personalized training strategies. To address these challenges, this study proposes a novel deep reinforcement learning (DRL) framework that dynamically processes multimodal inputs, incorporating optical data for precise motion capture, to enhance athletic performance, reduce injury risks, and maintain physiological stability. The framework utilizes a multimodal reward function to balance movement precision, injury prevention, and physiological consistency, while Proximal Policy Optimization (PPO) enables real-time adaptive training strategies. Neural network components, including convolutional and recurrent layers, are employed for spatial and temporal feature extraction, with an attention mechanism enhancing the integration of diverse data streams. Experimental results demonstrate that the proposed framework significantly outperforms traditional fixed training plans and single-modality methods, achieving faster reward convergence, a 42% reduction in injury risks, and enhanced training personalization. A case study further validates the system’s effectiveness, showcasing its ability to optimize motion trajectories and reduce deviations from biomechanical ideals by 85%. These findings highlight the transformative potential of combining optical sensing technologies with AI-driven smart healthcare systems for adaptive, safe, and personalized sports training solutions.