2 s stays separate from 5 s
A longer analysis window can change decoding performance. The five-second classical results are an operating point, not a comparator for the two-second frozen encoders.
Motion robustness · scalp and ear EEG
On the move
Performance declines as participants move, but the size and meaning of that decline depend on the task, time window, electrode layout and metric. The two mobile derivatives come from one original study, so they are related evidence rather than independent replication cohorts.
Matched operating point
Two seconds, eight selected scalp channels
The direct model comparison fixes the window at [0, 2 s), selects the same eight posterior scalp channels and evaluates 23 people on three SSVEP targets. Uniform chance is 33.3%.
| Method | Standing | Slow walk · 0.8 m/s | Fast walk · 1.6 m/s |
|---|---|---|---|
| Author-style CCAanalytic reference | 82.3%74.6%–89.4%82.3% balanced accuracy | 70.7%62.3%–78.6%70.7% balanced accuracy | 68.6%60.1%–76.7%68.6% balanced accuracy |
| Spectral ridgetrained on the other 22 standing participants | 78.7%70.7%–86.1%78.7% balanced accuracy | 69.9%61.2%–78.1%69.9% balanced accuracy | 68.6%60.4%–76.3%68.6% balanced accuracy |
| LaBraMfrozen encoder + ridge, trained on other participants | 59.5%53.6%–65.5%59.5% balanced accuracy | 54.7%49.5%–60.1%54.7% balanced accuracy | 47.8%43.3%–52.3%47.8% balanced accuracy |
| CBraModfrozen encoder + ridge, trained on other participants | 59.9%54.3%–65.4%59.9% balanced accuracy | 46.2%41.1%–51.4%46.2% balanced accuracy | 46.0%42.2%–49.6%46.0% balanced accuracy |
Bars span 0–100%; chance is 33.3%.
Matched here means matched within this table. The foundation-model rows are fixed frozen-encoder adapters, not best fine-tuned results. LaBraM uses verified standard channel IDs; CBraMod uses an exploratory positional channel order. Record-level pretraining exposure is not certified.
A separate operating point
Five-second scalp and ear comparisons
Longer windows and different electrode layouts change the problem. These classical five-second results do not share a ranking with the two-second foundation-model table above.
| Method / modality | Standing | Slow walk · 0.8 m/s | Fast walk · 1.6 m/s |
|---|---|---|---|
| Author-style CCA8 selected scalp channels · 5 s | 88.7%80.0%–95.9% | 82.9%74.3%–90.4% | 80.9%72.0%–88.8% |
| Spectral ridge8 selected scalp channels · 5 s | 85.0%77.0%–92.0% | 79.2%70.1%–87.5% | 77.2%67.7%–85.7% |
| Author-style CCA14 ear contacts · 5 s | 53.1%47.4%–59.6% | 43.4%39.6%–47.8% | 39.1%36.7%–41.8% |
| Spectral ridge14 ear contacts · 5 s | 52.0%46.9%–57.8% | 45.4%40.7%–51.2% | 40.0%36.4%–43.7% |
ERP session transfer
AUC falls toward chance while moving
A separate temporal-feature L2 logistic regression is fit for each participant and modality on the first standing session, then used unchanged later. The endpoint is ROC AUC, not balanced accuracy.
−0.276 AUCScalp ROC AUC change: standing to running, same 17 people
Use the paired common cohort for the running contrast
Running contains 17 participants while standing and walking contain 24. Among the same 17 people, the standing-to-running change is −0.276 AUC for scalp and −0.225 AUC for ear. Their descriptive 95% intervals are −0.322 to −0.227 AUC and −0.277 to −0.175 AUC.
| Modality | Standing | Slow walk · 0.8 m/s | Fast walk · 1.6 m/s | Running · 2.0 m/s |
|---|---|---|---|---|
| Scalp · 32 channelsstanding-session fit · unchanged weights | 0.8590.820–0.897n=24 | 0.7400.703–0.777n=24 | 0.6350.602–0.673n=24 | 0.5860.561–0.613n=17 |
| Ear · 14 contactsstanding-session fit · unchanged weights | 0.7340.694–0.774n=24 | 0.6280.595–0.661n=24 | 0.5630.540–0.587n=24 | 0.5170.491–0.541n=17 |
Methods & limits
Four distinctions that keep the result honest
Motion is the common theme, but the protocols answer different questions. No score is promoted into a cross-task leaderboard.
Scalp 8 is a selection
The mobile SSVEP scalp view selects eight posterior channels from 32 recorded. It is not the native eight-channel wearable device used on the dry-versus-wet page.
Balanced accuracy differs from AUC
The SSVEP endpoint is participant-mean balanced accuracy for a three-class task. The ERP endpoint is participant-mean ROC AUC. Their numeric values do not share a scale of practical meaning.
The derivatives share a study
NEMAR nm000125 and nm000201 are SSVEP and ERP derivatives of the same original mobile BCI study, with different assessed sample counts.
Lee2021 – SSVEP paradigm of the Mobile BCI dataset
Lee, Y., Shin, G., Lee, M., & Lee, S. (2026). Lee2021 – SSVEP paradigm of the Mobile BCI dataset (Version v1.0.2) [Data set]. NEMAR. https://doi.org/10.82901/nemar.nm000125
Lee, Y.-E., Shin, G.-H., Lee, M., & Lee, S.-W. (2021). Mobile BCI dataset of scalp- and ear-EEGs with ERP and SSVEP paradigms while standing, walking, and running. Scientific Data, 8, 315. https://doi.org/10.1038/s41597-021-01094-4
NEMAR/MOABB-generated BIDS derivative · CC BY 4.0.
ERP paradigm of the Mobile BCI dataset
Lee, Y., Shin, G., Lee, M., & Lee, S. (2026). ERP paradigm of the Mobile BCI dataset (Version v1.0.2) [Data set]. NEMAR. https://doi.org/10.82901/nemar.nm000201
Lee, Y.-E., Shin, G.-H., Lee, M., & Lee, S.-W. (2021). Mobile BCI dataset of scalp- and ear-EEGs with ERP and SSVEP paradigms while standing, walking, and running. Scientific Data, 8, 315. https://doi.org/10.1038/s41597-021-01094-4
NEMAR/MOABB-generated BIDS derivative · CC BY 4.0.
Data source: reviewed aggregate JSON · schema bci-report-public-deployment-topics-v1 · generated 2026-09-20.