A new AI model called Phy-BP is proposed for non-invasive blood pressure monitoring using triaxial bodyseismography. The model includes an adaptive quality-control algorithm to select high-quality BSG segments and a physical model that describes wave propagation in the body-bed system. This helps align multi-axis features during training, improving robustness against distortions. Experiments on a hospital dataset show that Phy-BP can dynamically filter out low-quality measurements and provide faithful BP monitoring. --- Why it matters: This matters to researchers because it addresses a common issue with contactless blood pressure monitoring: the need for accurate alignment of multi-axis features during training, which is crucial when training samples are limited. This model's ability to improve robustness against distortions can lead to more reliable and widely applicable results. Source: https://arxiv.org/abs/2608.23562