moabb.datasets.MAMEM1#

class moabb.datasets.MAMEM1[source]#

SSVEP MAMEM 1 dataset.

PapersWithCode leaderboard: https://paperswithcode.com/dataset/mamem1-moabb

Dataset summary

#Subj

#Chan

#Classes

#Trials / class

Trials length

Sampling rate

#Sessions

10

256

5

12-15

3s

250Hz

1

Dataset from [1].

EEG signals with 256 channels captured from 11 subjects executing a SSVEP-based experimental protocol. Five different frequencies (6.66, 7.50, 8.57, 10.00 and 12.00 Hz) have been used for the visual stimulation,and the EGI 300 Geodesic EEG System, using a stimulation, HydroCel Geodesic Sensor Net (HCGSN) and a sampling rate of 250 Hz has been used for capturing the signals.

Check the technical report [2] for more detail. From [1], subjects were exposed to non-overlapping flickering lights from five magenta boxes with frequencies [6.66Hz, 7.5Hz, 8.57Hz 10Hz and 12Hz]. 256 channel EEG recordings were captured.

Each session of the experimental procedure consisted of the following:

  1. 100 seconds of rest.

  2. An adaptation period in which the subject is exposed to eight 5 second windows of flickering from a magenta box. Each flickering window is of a single isolated frequency, randomly chosen from the above set, specified in the FREQUENCIES1.txt file under ‘adaptation’. The individual flickering windows are separated by 5 seconds of rest.

  3. 30 seconds of rest.

  4. For each of the frequencies from the above set in ascending order, also specified in FREQUENCIES1.txt under ‘main trials’:

    1. Three 5 second windows of flickering at the chosen frequency,

      separated by 5 seconds of rest.

    2. 30 seconds of rest.

This gives a total of 15 flickering windows, or 23 including the adaptation period.

The order of chosen frequencies is the same for each session, although there are small-moderate variations in the actual frequencies of each individual window. The .freq annotations list the different frequencies at a higher level of precision.

Note: Each ‘session’ in experiment 1 includes an adaptation period, unlike experiment 2 and 3 where each subject undergoes only one adaptation period before their first ‘session’.

From [3]:

Eligible signals: The EEG signal is sensitive to external factors that have to do with the environment or the configuration of the acquisition setup The research stuff was responsible for the elimination of trials that were considered faulty. As a result the following sessions were noted and excluded from further analysis: 1. S003, during session 4 the stimulation program crashed 2. S004, during session 2 the stimulation program crashed, and 3. S008, during session 4 the Stim Tracker was detuned. Furthermore, we must also note that subject S001 participated in 3 sessions and subjects S003 and S004 participated in 4 sessions, compared to all other subjects that participated in 5 sessions (NB: in fact, there is only 3 sessions for subjects 1, 3 and 8, and 4 sessions for subject 4 available to download). As a result, the utilized dataset consists of 1104 trials of 5 seconds each.

Flickering frequencies: Usually the refresh rate for an LCD Screen is 60 Hz creating a restriction to the number of frequencies that can be selected. Specifically, only the frequencies that when divided with the refresh rate of the screen result in an integer quotient could be selected. As a result, the frequendies that could be obtained were the following: 30.00. 20.00, 15.00, 1200, 10.00, 857. 7.50 and 6.66 Hz. In addition, it is also important to avoid using frequencies that are multiples of another frequency, for example making the choice to use 10.00Hz prohibits the use of 20.00 and 30.00 Mhz. With the previously described limitations in mind, the selected frequencies for the experiment were: 12.00, 10.00, 8.57, 7.50 and 6.66 Hz.

Stimuli Layout: In an effort to keep the experimental process as simple as possible, we used only one flickering box instead of more common choices, such as 4 or 5 boxes flickering simultaneously The fact that the subject could focus on one stimulus without having the distraction of other flickering sources allowed us to minimize the noise of our signals and verify the appropriateness of our acquisition setup Nevertheless, having concluded the optimal configuration for analyzing the EEG signals, the experiment will be repeated with more concurrent visual stimulus.

Trial duration: The duration of each trial was set to 5 seconds, as this time was considered adequate to allow the occipital part of the bran to mimic the stimulation frequency and still be small enough for making a selection in the context

References

1(1,2)

MAMEM Steady State Visually Evoked Potential EEG Database https://archive.physionet.org/physiobank/database/mssvepdb/

2

V.P. Oikonomou et al, 2016, Comparative evaluation of state-of-the-art algorithms for SSVEP-based BCIs. arXiv. <http://arxiv.org/abs/1602.00904>-

3

S. Nikolopoulos, 2016, DataAcquisitionDetails.pdf https://figshare.com/articles/dataset/MAMEM_EEG_SSVEP_Dataset_I_256_channels_11_subjects_5_frequencies_/2068677?file=3793738 # noqa: E501