EVA-Net: Subject-Independent EEG Motor Decoding with Video-Derived Motor Priors

2026-06-01

AI Summary

This work addresses the challenges of poor cross-subject generalization, high calibration dependency, and entanglement between motor semantics and subject-specific noise in EEG-based movement decoding. To this end, the authors propose EVA-Net, a two-stage framework that introduces dynamic action videos as semantic priors for the first time. In a shared embedding space, EEG and video features are aligned, and semantic knowledge is transferred to an EEG-only classifier via video category prototypes and knowledge distillation. By integrating cross-modal contrastive learning, supervised contrastive loss, and prototype learning, EVA-Net substantially improves cross-subject performance without increasing inference overhead. Experiments on two public datasets demonstrate its effectiveness, achieving an 8.66% accuracy gain over leave-one-subject-out (LOSO) baselines on the EEGMMI dataset and significantly outperforming methods relying on static textual priors.

Abstract

Practical non-invasive Brain-Computer Interface (BCI) systems require EEG decoders with strong cross-subject generalization and minimal calibration. However, inter-subject variability and signal non-stationarity often entangle motor semantics with subject-specific noise, limiting subject-independent decoding. Recent multimodal approaches use text as a semantic anchor, yet text provides sparse and static supervision for inherently dynamic motor processes. To address this issue, we propose EVA-Net, a two-stage framework that uses action videos as semantic priors for subject-independent EEG motor decoding. In the first stage, EEG and video features are aligned in a shared space using cross-modal and supervised contrastive objectives to reduce subject-specific variation. In the second stage, video category prototypes and knowledge distillation transfer video-derived priors to an EEG-only classifier without adding inference overhead. Experiments on two public datasets show that EVA-Net achieves strong subject-independent decoding performance, including an 8.66% LOSO accuracy gain on EEGMMI. Ablation results further suggest that video provides a more effective semantic anchor than the text baseline considered in this work.

Problem

• subject-independent EEG decoding
• cross-subject generalization
• motor imagery
• brain-computer interface
• semantic priors

Innovation

• subject-independent EEG decoding
• video-derived motor priors
• cross-modal contrastive learning
• knowledge distillation
• brain-computer interface