Dataset issue

[Darkhan92]

Hello, dear professor,

I am analyzing the data set produced during callibration and are not being able to understand it is structure of Y. For example why Y after offline analysis it has shape (43,731,25). So there is 25 vector contains zeros and ones. So I cannot understand why it is so?

[Darkhan92]

There clear structure regarding to X dataset where it has shape 43 trials and 731 data points and 8 channels. Where actually we have only 6 input data.

[jadref]

Hi,

This is mentioned in passing here https://mindaffect-bci.readthedocs.io/en/latest/offline_analysis.html.

But briefly we have:

• X is 3-dimensional with shapes: (Trials, Samples, Channels)
• Y is 3-dimensions with shape: (Trials, Samples, Outputs+1)

So what is an output: Well, for example if you have a 3x3 selection matrix on the screen, then there are 9 possible outputs, One for each 'flickering' box on the screen. In this array if the indicated output is 'on' (bright flash) then it has a 1 or a 0 if it is off (dark).

As mentioned above Y has #outputs+1 rows. To make analysis easier, the first row of Y is a 'virtual-output' which always contains the on/off sequence of the cued target output. So if you are analyzing calibration data you can just use the first row to know at what samples the output the user was looking at was on or off (bright/dark resp.)

Note: X may indeed have more 'channels' than expected, as many amplifiers add additional channels to the datastream for things like the reference voltage, or a sample counter.

[Darkhan92]

Thank you for your prompt reply dear professor Jason! But still when I look the row it is only zeros not informative. Maybe you want to say look for only first column. But there, then how we can understand which letter (or box) is flickering? Also, we cannot understand why there at one time instance withing 731 there 26 data information which corresponds to the boxes of letters. But there is withing this 26 data several ones in row? Because during callibration only one box or letter is flickering. Thank you!

[Darkhan92]

Also, I am talking about SSVEP, just to clarify

[Darkhan92]

So how here is encoded the flickering with different frequencies for 25 letters. Is that in one trial first column can represent flickering frequency of one specific letter? So, during calibration we need to look for only for one column, during free typing it looks for other 25 columns that corresponds to free typing? But then for next letter for next trial how it decides to choose another letter box?