{
  "Name": "Han2024 – SSVEP fatigue dataset with two frequency paradigms",
  "BIDSVersion": "1.9.0",
  "HEDVersion": "8.4.0",
  "DatasetType": "derivative",
  "License": "CC BY 4.0",
  "Authors": [
    "Yuheng Han",
    "Yufeng Ke",
    "Ruiyan Wang",
    "Tao Wang",
    "Dong Ming"
  ],
  "Funding": [
    "National Key Research and Development Program of China (Grant 2021YFF1200603)",
    "National Natural Science Foundation of China (Grants 62276184, 61806141)"
  ],
  "EthicsApprovals": [
    "Research Ethics Committee of Tianjin University"
  ],
  "ReferencesAndLinks": [
    "https://zenodo.org/records/10507229"
  ],
  "GeneratedBy": [
    {
      "CodeURL": "https://github.com/NeuroTechX/moabb",
      "Name": "moabb",
      "Description": "Mother of All BCI Benchmarks",
      "Version": "1.4.3"
    }
  ],
  "SourceDatasets": [
    {
      "DOI": "10.1109/TNSRE.2024.3380635",
      "URL": "https://zenodo.org/records/10507229"
    }
  ],
  "Keywords": [
    "SSVEP",
    "BCI",
    "fatigue",
    "dynamic stopping",
    "EEG"
  ],
  "PublicationYear": 2024,
  "ExperimentName": "Han2024",
  "Description": "SSVEP fatigue dataset with two frequency paradigms.\n\n\nDataset summary:\n\n#Subj      24\n#Chan      64\n#Classes      32\n#Trials / class    6-24\nTrials length     2 s\nFreq 1000 Hz\n#Sessions       2\n\n\nParticipants:\n\n- Population: healthy\n\n\nEquipment:\n\n- Amplifier: Synamps2 (Neuroscan)\n- Montage: standard_1005\n- Reference: Cz\n\n\nPreprocessing:\n\n- Data state: epoched\n\n\nData Access:\n\n- DOI: 10.1109/TNSRE.2024.3380635\n- Data URL: https://zenodo.org/records/10507229\n- Repository: Zenodo\n\n\nExperimental Protocol:\n\n- Paradigm: ssvep\n- Task type: gaze-shifting\n- Feedback: none\n- Stimulus: JFPM visual flicker\n\nDataset from [1]_.\n\nThis dataset contains 64-channel EEG recordings from 24 healthy subjects\n(12 males, 12 females, aged 18-26) performing two SSVEP-BCI tasks:\n\n- Low-frequency paradigm: 16 targets (8.0-15.5 Hz, 0.5 Hz step)\n- High-frequency paradigm: 16 targets (25.5-33.0 Hz, 0.5 Hz step)\n\nBoth paradigms used JFPM encoding with phases cycling through\n0, 0.5*pi, pi, 1.5*pi in a 4x4 matrix layout.\n\nThe experiment consisted of two phases: training (6 blocks per frequency\ncondition) and fatigue (24 blocks per condition). Each block contained\n16 trials (2 s stimulation per trial).\n\nEEG was recorded at 1000 Hz with a Synamps2 system (Neuroscan) and 64\nchannels. Each epoch spans 3000 samples (3 s at 1000 Hz).\n\n.. note::\n\nChannel selection is critical for this dataset. Using all 64 channels\nwith CCA-based methods yields near-chance accuracy because the high\nchannel-to-sample ratio causes overfitting. The paper uses 9 occipital\nchannels (PO7, PO3, POz, PO4, PO8, O1, Oz, O2, and one additional)\nand achieves >90% with TRCA. Users should pick occipital channels\nbefore classification.\n\nAdditionally, the cross-session evaluation (training on alert session\n'0', testing on fatigued session '1') is a challenging domain-shift\nproblem that standard CCA/TRCA may not handle well without\nfatigue-aware strategies.\n\nData is stored as [16, 64, 3000, N_blocks] matrices (targets, channels,\ntimepoints, blocks) in per-subject zip files on Zenodo. Each subject has\n4 separate files: low_frequency_train, low_frequency_fatigue,\nhigh_frequency_train, high_frequency_fatigue.\n\nIn MOABB, this is mapped as:\n- Session '0': Training blocks (6 blocks per condition, 12 total)\n- Session '1': Fatigue blocks (24 blocks per condition, 48 total)\n\nReferences\n----------\n.. [1] Y. Han, Y. Ke, R. Wang, T. Wang, and D. Ming, \"Enhancing\nSSVEP-BCI Performance Under Fatigue State Using Dynamic Stopping\nStrategy,\" IEEE Trans. Neural Syst. Rehab. Eng., vol. 32,\npp. 1407-1415, 2024. DOI: 10.1109/TNSRE.2024.3380635",
  "DatasetDOI": "10.82901/nemar.nm000124",
  "Version": "1.0.1"
}