Geometric Features for Brain-Computer Interface Classification: Benchmark Results on PhysioNet Motor Imagery

OmniSciences Technical Report TR-2026-003 We present benchmark results for proprietary geometric feature extraction applied to EEG-based brain-computer interface (BCI) motor imagery classification. The method operates on spatial covariance matrices on the SPD manifold. Key results: PhysioNet (109 subjects, 64 channels): +0.78pp over pyRiemann baseline (p=0.0048, Wilcoxon signed-rank test) Strongest advantage at low channel counts (n=3-8), relevant for consumer BCI Cross-subject transfer: +8.5pp over standard MDM Patent-pending (U.S. Provisional Application No. 64/004,047).