Data set V ‹mental imagery, multi-class›

Data set provided by IDIAP Research Institute (Silvia Chiappa, José del R. Millán)

Correspondence to José del R. Millán ⟨jose.millan@idiap.ch⟩

Experiment

This dataset contains data from 3 normal subjects during 4 non-feedback sessions. The subjects sat in a normal chair, relaxed arms resting on their legs. There are 3 tasks:
  1. Imagination of repetitive self-paced left hand movements, (left, class 2),
  2. Imagination of repetitive self-paced right hand movements, (right, class 3),
  3. Generation of words beginning with the same random letter, (word, class 7).
All 4 sessions of a given subject were acquired on the same day, each lasting 4 minutes with 5-10 minutes breaks in between them. The subject performed a given task for about 15 seconds and then switched randomly to another task at the operator's request. EEG data is not splitted in trials since the subjects are continuously performing any of the mental tasks. The algorithm should provide an output every 0.5 seconds using the last second of data (see clarification in the paragraph 'Requirements and Evaluation'.) Data are provided in two ways:
  1. Raw EEG signals. Sampling rate was 512 Hz.
  2. Precomputed features. The raw EEG potentials were first spatially filtered by means of a surface Laplacian. Then, every 62.5 ms --i.e., 16 times per second-- the power spectral density (PSD) in the band 8-30 Hz was estimated over the last second of data with a frequency resolution of 2 Hz for the 8 centro-parietal channels C3, Cz, C4, CP1, CP2, P3, Pz, and P4. As a result, an EEG sample is a 96-dimensional vector (8 channels times 12 frequency components).

Format of the Data

For each subject there are 3 training files and 1 testing file (the last recording session). Training files are labelled while testing files are not. Data are provided in ASCII format.

Requirements and Evaluation

Please provide your estimated class labels (2, 3, or 7) for every input vector of the 3 test files (one per subject). The labels must be estimated in the following way: Also give a description of the used algorithm. The performance measure is the classification accuracy (correct classification divided by the total number of samples) averaged over the 3 subjects.
There will be a special prize to the best algorithm working with the precomputed samples (in the case it does not achieve the best absolute result).

Technical Information

EEG signals were recorded with a Biosemi system using a cap with 32 integrated electrodes located at standard positions of the International 10-20 system. The sampling rate was 512 Hz. Signals were acquired at full DC. No artifact rejection or correction was employed.

Reference


[ BCI Competition III ]