PanLUNA: An Efficient and Robust Query-Unified Multimodal Model for Edge Biosignal Intelligence

Abstract

PanLUNA is a compact 5.4M-parameter pan-modal foundation model that jointly processes EEG, ECG, and PPG within a single shared encoder. Extending LUNA’s channel-unification module, it treats multimodal channels as entries in a unified query set augmented with sensor-type embeddings, enabling efficient cross-modal early fusion while remaining robust to missing modalities at inference time. It matches or exceeds models up to 57× larger — 81.21% balanced accuracy on TUAB and state-of-the-art 0.7416 balanced accuracy on HMC multimodal sleep staging — and INT8 deployment on the GAP9 microcontroller achieves 325.6 ms latency and 18.8 mJ per 12-lead ECG inference.