Joint Attention Mechanism Learning to Facilitate Opto-physiological Monitoring during Physical Activity

On November 12, 2025, researchers unveiled the attention-mechanism-based generative adversarial network (AM-GAN), designed to improve opto-physiological monitoring by mitigating motion artefacts that compromise the quality of photoplethysmography (PPG) signals during physical activity. The study involved 43 participants engaged in various activities ranging from low to high intensity (6-12 km/h). The AM-GAN demonstrated its effectiveness by achieving a mean absolute error (MAE) of 1.81 beats/min for heart rate on the IEEE-SPC dataset and less than 1.37 beats/min on an in-house dataset. Additionally, it recorded a respiratory rate MAE of 2.49 breaths/min and a SpO2 accuracy of 1.65% across different oxygen levels. This innovation is crucial for enhancing the reliability of non-invasive health monitoring technologies, particularly in dynamic environments where traditional methods struggle to maintain accuracy.