Research on exercise intervention methods for the elderly based on wearable device data

Elderly individuals face challenges in exercise rehabilitation, including significant individual differences, slow heart rate response, and insufficient intervention adherence. This study investigates an intelligent exercise intervention model that integrates temporal modeling and reinforcement learning, constructed using multimodal data from wearable devices. Experimental results show that the model significantly improves prediction accuracy and stability , with an RMSE of 4.02, MAE of 3.11, and R² of 0.947. In a heart rate recovery experiment under brisk walking conditions, HRR₆₀ increased to 23.4 bpm, t₅₀ decreased to 32.1 s, and the recovery stability index decreased to 0.081. The model effectively characterizes the nonlinear features of heart rate recovery after exercise, enabling intelligent prediction of exercise load and recovery rhythm. This method can provide a new algorithmic path and theoretical support for individualized exercise intervention for the elderly.