# SSVEP Literature

> Generated from repository source files. Do not edit this page directly; edit the source file listed below and rerun `vp run docs:generate`.

> Source: `literature/01-ssvep-history-and-paradigms.md`

## SSVEP History And Paradigms

Status: `inferred`

This page summarizes the local SSVEP evidence after the 2026-07-12 deep-dive intake. It should be read as a research map, not as a claim that SAH-BRI-Grasp dynamic object-box SSVEP has already been validated.

## Review Questions

* What neural and visual mechanisms explain SSVEP reliability?
* How did SSVEP/VEP BCIs evolve from early device control to high-speed spellers?
* What is the method lineage from FFT/PSD to CCA, FBCCA, TRCA, and recent deep models?
* What do user-coverage and benchmark datasets say about practical SSVEP?
* What does AR/VR SSVEP evidence contribute, and what does it still not prove for YOLO-generated object boxes?

## Local Evidence Map

| Evidence Need | Candidate IDs | Status |
| --- | --- | --- |
| Mechanism and visual response baseline | SSVEP-001; SSVEP-007 | extracted |
| Early SSVER/VEP/SSVEP BCI systems | SSVEP-011; SSVEP-012; SSVEP-013; SSVEP-014 | extracted |
| CCA lineage | SSVEP-016; SSVEP-019 | extracted |
| High-speed fixed speller baselines | SSVEP-003; SSVEP-004; SSVEP-005 | extracted |
| User coverage / population usability | SSVEP-020 | extracted |
| Public benchmark datasets | SSVEP-005; SSVEP-021; SSVEP-029 | extracted |
| AR/VR and spatial interaction bridge | SSVEP-022; SSVEP-023 | extracted |
| Deep, transfer, augmentation, attention trends | SSVEP-010; SSVEP-026; SSVEP-027; SSVEP-028; SSVEP-030; SSVEP-031 | extracted |
| Detector-bound dynamic object-box SSVEP | project experiment needed | needs confirmation |

## Lineage

### 1. Visual Response And Early BCI Control

`SSVEP-001` and `SSVEP-007` provide the visual-neuroscience baseline: SSVEP depends on frequency-tagged visual stimulation and frequency-dependent cortical response. `SSVEP-011` then anchors the early BCI transition: steady-state visual-evoked responses were used as control signals before modern speller benchmarks.

`SSVEP-012` and `SSVEP-013` show the early practical BCI question: SSVEP can achieve useful transfer rates for some users, but user variability and system preparation matter. `SSVEP-014` frames VEP/SSVEP practical design as a system problem involving target count, electrode count, display constraints, cost, and usability.

### 2. CCA To High-Speed Spellers

`SSVEP-016` is the local CCA anchor: it uses canonical correlation analysis to match multichannel EEG against reference signals and improves over FFT-style spectral recognition in the local paper text.

`SSVEP-019` complicates the CCA story: standard CCA is not the only CCA-family method, and individual calibration data can improve detection. This matters because SAH-BRI-Grasp must choose between no-calibration convenience and calibrated performance.

`SSVEP-003`, `SSVEP-004`, and `SSVEP-005` then represent the high-speed fixed-speller stage: FBCCA, TRCA/eTRCA, and large 40-target benchmark evaluation. These papers are strong baselines for Exp1, but their targets are fixed display cells rather than scene objects.

### 3. Usability And Datasets

`SSVEP-020` is important because it asks how many users can actually use SSVEP. In the local card, the 53-subject four-LED study reported high average accuracy and high user coverage, but this is simpler than a dynamic object-selection interface.

`SSVEP-005`, `SSVEP-021`, and `SSVEP-029` support the dataset trend: public 40-target benchmark data, BETA, and Dual-Alpha make SSVEP algorithm comparison more systematic. They also help with channel/time-window studies before project-specific collection.

### 4. AR/VR And Real-World Overlay Direction

`SSVEP-022` and `SSVEP-023` are the closest local SSVEP sources to spatial interface evidence. They show that SSVEP can be studied in VR/AR or HoloLens-like settings, but they do not evaluate YOLO-generated physical-object boxes, detector jitter, robot safety gates, or grasp execution.

This distinction is central: AR/VR SSVEP reduces the gap from fixed screens to spatial interfaces, but it does not close the SAH-BRI-Grasp evidence gap.

### 5. Recent Decoder Trends

Recent papers add methods that may matter later:

* `SSVEP-010`: compact CNN / asynchronous SSVEP comparator.
* `SSVEP-026`: transformer-based inter-subject SSVEP classification.
* `SSVEP-027`: GAN-based short-window SSVEP data extension.
* `SSVEP-028`: one-shot cross-subject SSVEP transfer with TRCA/TDCA components.
* `SSVEP-030`: data augmentation and language-model priors for spellers.
* `SSVEP-031`: attention-based robust decoding with multiple signal representations.

These papers support trend discussion. They should not replace classical FBCCA/TRCA baselines in the first project experiment unless there is enough local data to justify training or adaptation.

## Implications For SAH-BRI-Grasp

* Use `SSVEP-003`, `SSVEP-004`, `SSVEP-005`, and `SSVEP-021` as the main fixed-target SSVEP benchmark layer.
* Use `SSVEP-011` to `SSVEP-016` for historical and method lineage.
* Use `SSVEP-020` for user-coverage context, not as a guarantee for dynamic object-box performance.
* Use `SSVEP-022` and `SSVEP-023` as AR/VR bridge evidence.
* Keep deep SSVEP models as later comparators after classical baselines and local data collection.

## Remaining Gaps

| Gap | Current Handling |
| --- | --- |
| YOLO-generated dynamic object boxes may flicker, shift, overlap, or disappear. | Exp1 must test target freezing, tracking, and visual comfort. |
| Fixed speller ITR may not transfer to object-level selection. | Report task success, selection latency, false selection, and subjective comfort, not only ITR. |
| AR/VR SSVEP evidence is not the same as robot grasp evidence. | Treat AR/VR papers as interface comparators only. |
| Deep SSVEP methods may require calibration and data scale. | Start with FBCCA/TRCA; add deep models only after local baseline evidence exists. |