# SSVEP-026: A Transformer-based deep neural network model for SSVEP classification

> Generated from a local paper card. Do not edit this page directly; edit the source card and rerun `vp run docs:generate`.

> Internal wiki boundary: paper-card evidence and user-provided PDF downloads are available for private reading. Extracted full-paper text is not published.

> Source: `library/paper_cards/SSVEP-026.md`

## Paper Access

* Internal PDF: <a href={"/papers/SSVEP-026.pdf"} download style={{ display: "inline-flex", alignItems: "center", justifyContent: "center", minHeight: "2.25rem", padding: "0.45rem 0.8rem", borderRadius: "6px", backgroundColor: "#047857", color: "#ffffff", fontWeight: 700, lineHeight: 1, textDecoration: "none", boxShadow: "0 1px 2px rgba(15, 23, 42, 0.22)" }}>Download Paper</a>
* DOI / official page: [needs confirmation](https://arxiv.org/abs/2210.04172)
* Open-access page: [Open access page](https://arxiv.org/abs/2210.04172)
* Deployment boundary: these PDF links are intended only for a private/protected internal wiki.

## SSVEP-026: A Transformer-based deep neural network model for SSVEP classification

## Metadata

* ID: SSVEP-026
* Title: A Transformer-based deep neural network model for SSVEP classification
* Year: 2022
* DOI / URL: https://arxiv.org/abs/2210.04172
* Local PDF: see Paper Access section above
* Text artifact: local-only path withheld from docs site
* Review status: `extracted`

## Study Type

* Track: SSVEP
* Task: inter-subject SSVEP classification with SSVEPformer
* Participants or dataset: Dataset 1 has 10 subjects and 12 classes; Dataset 2 has 35 subjects and 40 classes
* Device/electrode setup: Dataset 1 uses 8 occipital electrodes; Dataset 2 details need confirmation
* Protocol/task: public SSVEP datasets for target classification

## Methods

* Signal processing or analysis: transformer-based SSVEPformer and filter-bank SSVEPformer
* Training/calibration: inter-subject classification scenario
* Metrics: accuracy and ITR

## Key Results

* With 1 s windows, FB-SSVEPformer achieved 88.37% accuracy and 112.45 bits/min on Dataset 1.
* With 1 s windows, it achieved 83.19% accuracy and 157.65 bits/min on Dataset 2.

## Limitations

* Public fixed-target datasets, not dynamic scene overlays.
* Deep models may require careful data-regime and cross-subject validation.
* DOI/published version needs confirmation.

## Relevance To Current Review

* Represents the transformer trend in SSVEP decoding.
* Useful as a recent cross-subject/low-calibration comparator, not as direct system evidence.

## Evidence Status

| Claim | Status | Evidence Note |
| --- | --- | --- |
| Transformer-based SSVEP models target inter-subject classification and reduced calibration needs. | verified | Abstract and methods frame inter-subject SSVEP classification. |
| FB-SSVEPformer reported high 1 s accuracy/ITR on two public datasets. | verified | Local text reports the two accuracy and ITR values. |
| Transformer SSVEP will generalize to SAH-BRI-Grasp without new data. | needs confirmation | Dynamic object-box and hardware domain shift are untested. |

## Open Questions

* Should Exp1 include SSVEPformer only after FBCCA/TRCA baselines are stable?
* Can the project collect enough data for meaningful transformer fine-tuning?