# Grasp-and-Lift EEG Detection — Data

来源：https://www.kaggle.com/competitions/grasp-and-lift-eeg-detection/data

采集时间：2026-10-24T15:45:23Z

This data contains EEG recordings of subjects performing grasp-and-lift (GAL) trials. The following video shows an example of a trial:
A detailed account of the data can be found in
Luciw MD, Jarocka E, Edin BB (2014) Multi-channel EEG recordings during 3,936 grasp and lift trials with varying weight and friction.
Scientific Data
1:140047.
[www.nature.com/articles/sdata201447](http://www.nature.com/articles/sdata201447)
There are 12 subjects in total, 10 series of trials for each subject, and approximately 30 trials within each series. The number of trials varies for each series. The training set contains the first 8 series for each subject. The test set contains the 9th and 10th series.
For each GAL, you are tasked to detect 6 events:
HandStart
FirstDigitTouch
BothStartLoadPhase
LiftOff
Replace
BothReleased
These events always occur in the same order. In the training set, there are two files for each subject + series combination:
the *_data.csv files contain the raw 32 channels EEG data
(sampling rate 500Hz)
the *_events.csv files contains the ground truth frame-wise labels for all
events
The events files for the test set are not provided and must be predicted. Each timeframe is given a unique id column according to the subject, series, and frame to which it belongs. The six label columns are either zero or one, depending on whether the corresponding event has occurred
within ±150ms (±75frames). A perfect submission will predict a probability of one for this entire window.
Important Note
When predicting, you may not use data from the future!
In other words, if you are predicting labels for id subj1_series9_11, you may not incorporate any frame after 11 of series 9 from subject 1. In the real application for of a detection algorithm like this, future data doesn't exist.
You must be careful not to include data leakage.
For example, you may not center the signals for subj1_series9_11 based on the mean of all frames for subj1_series9. Instead, you must use the mean based on frame 0 to 10. The host will be checking models to ensure they do not violate this rule. We recommend structuring your code so that it is abundantly clear that you are not accessing future data (or features based on future data) when predicting.
If you wish to, you may train on other subjects and series outside of the series for which you are predicting. For example, you can use all of subj2_series6 when predicting subj1_series5.
The columns in the data files are labeled according to their associated electrode channels. You may make use of the spatial relationship between the electrode locations, as shown in this diagram:
![](assets/b1ad7cfc75d100716e7674c7.jpg)