Enhancing Heart Rate Prediction with Mamba and Simple-Attention Modules for Early Warning in Exercise

According to incomplete statistics, over 500,000 people in China die from sudden cardiac death annually, with one of the most common triggers being sudden cardiac arrest during intense physical activity. To prevent exercise-induced sudden death, heart rate prediction based on exercise data has emerged as a critical approach. This paper proposes an HRP model based on Mamba and Simple-Attention, which can be used for real-time heart rate prediction during exercise and issue early warnings for abnormal conditions. The HRP model is trained on the Endomondo dataset, capturing contextual information from both current and historical user activity. The Mamba module enhances long-sequence data processing through efficient parallel computation and hardware optimization, while the Simple-Attention module effectively captures key features in the short term and reduces the complexity of traditional attention mechanisms. Experimental results show that the HRP model significantly reduces scores on RMSE and MAE metrics compared to traditional methods, with minimum values of 1.23 and 0.78, respectively. This indicates that the HRP model offers a novel technical solution for real-time heart rate monitoring and exercise safety warnings, with important practical applications.