On July 1, 2026, a team of researchers led by Cheng He introduced I²RiMA, a novel approach for detecting mental stress through EEG signals. This method addresses significant challenges in cross-subject EEG stress detection, which has been hindered by subject-dependent patterns and frequency-specific characteristics.
The I²RiMA framework, or Intra-Inter Riemannian Manifold Attention Network, constructs spatial covariance matrices at each frequency point, effectively preserving channel-wise geometry along with frequency-specific cues. The innovative design also aggregates frequency clusters, reducing redundancy by forming compact, data-driven clusters aligned with EEG rhythms.
Advancements in EEG Stress Detection Techniques
Traditional Riemannian methods primarily model spatial covariance in the time domain, often overlooking critical neural oscillations necessary for decoding cognitive states. I²RiMA introduces a sophisticated intra-inter slice attention module, which integrates local slice-level spectral dynamics with global temporal context across EEG sequences.
Experiments conducted on three distinct datasets demonstrate that I²RiMA significantly outperforms five state-of-the-art baselines. The method achieved a remarkable 82.78% balanced accuracy while maintaining efficiency with just 1.60M parameters and 31.95M FLOPs.
