moabb.datasets.Huebner2017

class moabb.datasets.Huebner2017(interval=None, raw_slice_offset=None, use_blocks_as_sessions=True)[source][source]

Learning from label proportions for a visual matrix speller (ERP) dataset from Hübner et al 2017 [1].

Dataset description

The subjects were asked to spell the sentence: “Franzy jagt im komplett verwahrlosten Taxi quer durch Freiburg”. The sentence was chosen because it contains each letter used in German at least once. Each subject spelled this sentence three times. The stimulus onset asynchrony (SOA) was 250 ms (corresponding to 15 frames on the LCD screen utilized) while the stimulus duration was 100 ms (corresponding to 6 frames on the LCD screen utilized). For each character, 68 highlighting events occurred and a total of 63 characters were spelled three times. This resulted in a total of 68 ⋅ 63 ⋅ 3 = 12852 EEG epochs per subject. Spelling one character took around 25 s including 4 s for cueing the current symbol, 17 s for highlighting and 4 s to provide feedback to the user. Assuming a perfect decoding, these timing constraints would allow for a maximum spelling speed of 2.4 characters per minute. Fig 1 shows the complete experimental structure and how LLP is used to reconstruct average target and non-target ERP responses.

Subjects were placed in a chair at 80 cm distance from a 24-inch flat screen. EEG signals from 31 passive Ag/AgCl electrodes (EasyCap) were recorded, which were placed approximately equidistantly according to the extended 10–20 system, and whose impedances were kept below 20 kΩ. All channels were referenced against the nose and the ground was at FCz. The signals were registered by multichannel EEG amplifiers (BrainAmp DC, Brain Products) at a sampling rate of 1 kHz. To control for vertical ocular movements and eye blinks, we recorded with an EOG electrode placed below the right eye and referenced against the EEG channel Fp2 above the eye. In addition, pulse and breathing activity were recorded.

Parameters
  • interval (array_like) – range/interval in milliseconds in which the brain response/activity relative to an event/stimulus onset lies in. Default is set to [-.2, .7].

  • raw_slice_offset (int, None) – defines the crop offset in milliseconds before the first and after the last event (target or non-targeet) onset. Default None which crops with an offset 2,000 ms.

References

1

Hübner, D., Verhoeven, T., Schmid, K., Müller, K. R., Tangermann, M., & Kindermans, P. J. (2017) Learning from label proportions in brain-computer interfaces: Online unsupervised learning with guarantees. PLOS ONE 12(4): e0175856. https://doi.org/10.1371/journal.pone.0175856

New in version 0.4.5.

Methods

data_path(subject[, path, force_update, …])

Get path to local copy of a subject data.

download([subject_list, path, force_update, …])

Download all data from the dataset.

get_data([subjects])

Return the data correspoonding to a list of subjects.