Practical Solutions for Non-Invasive Health Monitoring
Overcoming Challenges in Physiological Signal Measurement
Accurately measuring heart rate (HR) and heart rate variability (HRV) from facial videos is challenging due to factors like lighting variations and facial movements. PhysMamba offers a solution by efficiently extracting precise physiological signals for real-time health monitoring.
Innovative Framework for Physiological Measurement
PhysMamba utilizes the Temporal Difference Mamba (TD-Mamba) block and a dual-stream SlowFast architecture to capture local and long-range dependencies from facial videos. This approach enhances accuracy in remote photoplethysmography (rPPG) signal estimation, outperforming traditional CNN and Transformer models.
Significant Performance Improvements
PhysMamba achieved remarkable results on benchmark datasets, with low Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). It excelled in capturing subtle physiological changes across different scenarios, making it ideal for real-time heart rate monitoring from facial videos.
