CardioState-JEPA (shared cardiac encoder)

A single shared encoder for ECG, PPG, and PCG, trained with a delay-aware cross-modal joint-embedding predictive architecture. This repo hosts the frozen encoder used for the downstream results; it maps a waveform to a pooled cardiac code.

Usage

from transformers import AutoModel
import torch

model = AutoModel.from_pretrained("hamzashafiq/CardioState-Jepa", trust_remote_code=True).eval()

ppg = torch.randn(2, 1, 1250)                 # [batch, 1 ch, 10 s @ 125 Hz]
with torch.no_grad():
    out = model(ppg, modality="ppg", fs=125.0)
print(out.pooler_output.shape)                # (2, 768) cardiac code

modality is one of "ecg" (12 leads @ 500 Hz), "ppg" (1 ch @ 125 Hz), or "pcg" (1 ch @ 4000 Hz); pass the matching fs.

Paper

CardioState-JEPA: Delay-Aware Cross-Modal Learning of a Shared Cardiac Representation

Hamza Shafiq, Hung Manh Pham, Bin Zhu, Pan Zhou, Jun Hu, and Aaqib Saeed.

arXiv:2608.12944, 2026.

• Paper: https://arxiv.org/abs/2608.12944
• Hugging Face Papers: https://huggingface.co/papers/2608.12944

Citation

If you use CardioState-JEPA in your research, please cite:

@misc{shafiq2026cardiostatejepa,
  title         = {CardioState-JEPA: Delay-Aware Cross-Modal Learning of a Shared Cardiac Representation},
  author        = {Shafiq, Hamza and Pham, Hung Manh and Zhu, Bin and Zhou, Pan and Hu, Jun and Saeed, Aaqib},
  year          = {2026},
  eprint        = {2608.12944},
  archivePrefix = {arXiv},
  primaryClass  = {cs.LG}
}