Datasets:

webxos
/

BCI-FPS

language:
  - en
tags:
  - reinforcement-learning
  - neural-decoding
  - motor-imagery
  - human-computer-interaction
  - neuralink
task_categories:
  - reinforcement-learning
task_ids:
  - natural-language-inference
  - entity-linking-classification
  - acceptability-classification
  - intent-classification
size_categories:
  - 10K<n<100K

Dataset Card for BCI-FPS MOTOR_IMAGERY Dataset

Dataset Description

UNDER DEVELOPMENT

Use Cases: BCI Intent Data Study and Testing (conceptual early design)

Training machine learning models for neural signal decoding without needing large real neural datasets, addressing data scarcity and privacy issues.

Augmenting real-world BCI data with synthetic samples to improve model robustness and diversity, as in GAN-based approaches.

Testing and calibrating BCI systems for motor imagery tasks like prosthetic control before human trials.

Simulating neural responses in assistive technologies for disabled individuals, enabling faster iteration in labs like Neuralink.

Developing predictive models for intent recognition in human-AI interactions and rehabilitative BCIs.

Enhancing clinical research datasets for disease risk assessment and patient outcome prediction in neuroengineering.

Validating algorithms in frontier labs (e.g., Neuralink, Paradromics) for high-data-rate implants by generating idealized signals.

This dataset contains high-bandwidth neural training data collected from BCI-FPS, a specialized training platform for brain-computer interface research.

Dataset Summary

Training Mode: MOTOR IMAGERY
Session ID: bci_fps_motor_imagery_1767171179245
Duration: 52 seconds
Sampling Rate: 1000 Hz
Neural Channels: 32
Data Points: 11,314

Supported Tasks

Motor Imagery Training for prosthetic control
Neural Decoding: Training models to decode user intent from neural signals
BCI Calibration: Providing ground truth data for BCI system calibration
Disability Research: Supporting development of assistive technologies

Languages

English (interface and documentation)

Dataset Structure

Data Instances

{
  "timestamp": 1767171127035,
  "session_time": 2,
  "channels": {
    "channel_0": 0.7145493839481488,
    "channel_1": 0.6894168445867142,
    "channel_2": 0.08142761930267149,
    "channel_3": -0.4847495027079371,
    "channel_4": -0.7151022782142631,
    "channel_5": -0.30725177077599913,
    "channel_6": 0.41521139153211245,
    "channel_7": 0.8975965762479154,
    "channel_8": 0.40940126876082966,
    "channel_9": -0.4091680578228324,
    "channel_10": -0.8292701881852992,
    "channel_11": -0.5904045145284711,
    "channel_12": 0.12196528544955941,
    "channel_13": 0.7040845591149026,
    "channel_14": 0.5296790688037042,
    "channel_15": 0.018181536760527098,
    "channel_16": -0.6973668262179662,
    "channel_17": -0.7437997196398959,
    "channel_18": -0.10714886215673841,
    "channel_19": 0.6246891444747351,
    "channel_20": 0.8560240877317689,
    "channel_21": 0.155749695078711,
    "channel_22": -0.4754514086663171,
    "channel_23": -0.7632646743624881,
    "channel_24": -0.42658424045199833,
    "channel_25": 0.47380668620054267,
    "channel_26": 0.7558851981047924,
    "channel_27": 0.5145527444334146,
    "channel_28": -0.22899647502709344,
    "channel_29": -0.8498710316208474,
    "channel_30": -0.5816021940073672,
    "channel_31": 0.2096020563849897
  },
  "intent_context": {
    "mouse_movement": [
      0,
      0
    ],
    "keyboard_state": {
      "mouse": false
    },
    "camera_rotation": [
      0,
      0,
      0
    ],
    "active_targets": 0
  }
}

Data Fields

See metadata.json for complete schema documentation.

Dataset Creation

Source Data

Platform: Web-based BCI-FPS Training Environment
Sampling Rate: 1000 Hz
Collection Method: Real-time telemetry during BCI training tasks
Neural Simulation: Synthetic neural data representing ideal BCI signals

Annotations

Annotation process: Automatic intent labeling during gameplay
Annotation types: Motor imagery, visual stimuli, handwriting intent
Who annotated: System automatically labels based on game state

Personal and Sensitive Information

No personal information is collected. All data is synthetic/anonymous.

Considerations for Using the Data

Social Impact

This dataset enables research in:

Neuralink-style brain-computer interfaces
Assistive technologies for disabled individuals
Human-AI interaction systems
Neural decoding algorithms

Discussion of Biases

Synthetic neural data may not perfectly represent biological signals. Results should be validated with real neural recordings.

Other Known Limitations

Simulated neural signals
Idealized game environment
Limited to specific training tasks

Additional Information

Dataset Curators

BCI-FPS Research Team

Licensing Information

MIT License

Citation Information

@misc{bci_fps_motor_imagery_2024,
  title={BCI-FPS motor_imagery Training Dataset},
  author={Neuralink Research},
  year={2024},
  note={High-frequency intent decoding data for brain-computer interface development}
}