SIGMA-PPG: Statistical-prior Informed Generative Masking Architecture for PPG Foundation Model

Current foundation model for photoplethysmography (PPG) signals is challenged by the intrinsic redundancy and noise of the signal. Standard masked modeling often yields trivial solutions while contrastive methods lack morphological precision. To address these limitations, we propose a Statistical-prior Informed Generative Masking Architecture (SIGMA-PPG), a generative foundation model featuring a prior-guided adversarial masking mechanism, where a reinforcement learning-driven teacher leverages statistical priors to create challenging learning paths that prevent overfitting to noise. We also incorporate a semantic consistency constraint via vector quantization to ensure that physiologically identical waveforms—even those altered by recording artifacts or minor perturbations—map to shared indices. This enhances codebook semantic density and eliminates redundant feature structures. Pre-trained on over 120,000 hours of data, SIGMA-PPG achieves superior average performance compared to five state-of-the-art baselines across 12 diverse downstream tasks. The code and model weights are available at https://github.com/ZonghengGuo/SigmaPPG.

Smartwatches and medical devices use light sensors (PPG) to monitor heart health, but the signals they collect are often noisy and highly repetitive. Building advanced AI "foundation models" to understand these human signals is difficult because the AI can easily get confused by the noise or take shortcuts instead of learning meaningful health patterns. To solve this, we created SIGMA-PPG, an AI model trained using a smart, challenging learning process. We introduced an AI "teacher" that strategically hides parts of the heartbeat signal, forcing the main model to truly understand the underlying physiological patterns rather than just memorizing useless noise. We also taught the model to recognize that slightly messy or shifted heartbeat recordings still represent the same health state. By learning from over 120,000 hours of real-world data collected from patients in intensive care units (ICUs), our model analyzes heart signals much more accurately than existing tools, paving the way for more reliable and robust health monitoring systems.