MRieHy first computes Riemannian means of covariance matrices from cross-day training data to align multi-day distributions, then builds one hypergraph over covariance matrices using Riemannian distance and a second over deep features using cosine similarity, fusing them with adaptively learned weights and joint optimization with label projection matrices. During online testing it maintains a first-in-first-out buffer of recent samples, applies Riemannian alignment, and decodes with the learned hypergraph. The researchers report notable performance gains over state-of-the-art baselines on a private four-class ECoG dataset and two public four-class EEG datasets.
MRieHy Framework for Online MI-BCI Adaptation
Summary
Researchers propose MRieHy, a multi-feature Riemannian hypergraph framework for online test-time adaptation of motor imagery BCI decoding. It aligns multi-day distributions via Riemannian means of covariance matrices, builds one hypergraph with Riemannian distance and a second with cosine similarity, fuses them with adaptively learned weights, and decodes buffered online samples after Riemannian alignment. On a private four-class ECoG dataset and two public four-class EEG datasets, MRieHy shows notable gains over state-of-the-art baselines, targeting the cross-day transferability and online operation that clinical MI-BCI still lacks.
Why it matters
Cross-day drift is a persistent barrier to clinical MI-BCI; MRieHy blends Riemannian geometry with hypergraph structure for a testable fix, and its public-dataset validation makes the claim easy to check.
Compiled by BCIwiki from public sources
Sources · 1
arxiv.org 2026-08-17
Chinese Academy of Medical Sciences & Peking Union Medical CollegePeking UniversityDecoding AlgorithmsMotor DecodingChina