Investigate classification tasks/datasets

[ErikBjare]

We'll investigate previous approaches to classify EEG data that is similar to our task at hand.

See also #17.

Classifying tasks

Likely the most similar type classification.

Synchronized Brainwave Dataset (2015)

Dataset on Kaggle: https://www.kaggle.com/berkeley-biosense/synchronized-brainwave-dataset

Stimuli:

• Video 1: https://www.youtube.com/watch?v=zkGoPdpRvaU
• Video 2: https://www.youtube.com/watch?v=sxqlOoBBjvc

They both follow the same process:

• Blinking
• Relax (closed eyes, focus on breathing)
• Arithmetic
• Relax with music
• Video clip
• Come up with examples from category
• Count squares of chosen color

A popular subset of the stimuli is relaxation vs math:

• With a SVM: https://www.kaggle.com/elsehow/classifying-relaxation-versus-doing-math
• With a NN: https://www.kaggle.com/wpncrh/classifying-tasks-using-eeg-data-w-tensorflow-nn

Reading prose vs code (Fucci et al)

No publicly available dataset. Ask Fucci?

Classifying sleep stages

Shares some similarities (long recordings, "organic" data).

See the excellent YASA: https://github.com/raphaelvallat/yasa

Classifying emotion

Might be somewhat similar. Often uses 1min clips of happy/sad movie scenes as stimuli.

Sometimes split into arousal/valence.

Classifying mental states (focus etc)

Classifying things like focus is sometimes considered a simpler task where acceptable classification can be achieved with a simple power band ratio.

EEG data for Mental Attention State Detection (focused, unfocused, drowsy)

Dataset on Kaggle (MATLAB files): https://www.kaggle.com/inancigdem/eeg-data-for-mental-attention-state-detection