Manipulating ov files or streams

[kadirfu]

Hi,

I am trying to see the relation between the dataset size and corresponding classification performance. I am using bci-motor-imagery dataset.

I was wondering how I can fetch only some part of the dataset (viz. stream). One solution might be slicing the signal into smaller pieces and then merging them if necessary. Another solution would be counting the number of stimuli presented to the participant and cutting the experiment when n>=m (n is the counter value and m is the predefined set size). I think for this I need a Switch Director kind of box which will count number of stimuli presented and cut the connection after that point. By changing the counter parameter, I would have classification scenarios with different sized datasets.

As I have just started using OpenVibe, I am quite a newbie. I would appreciate if you can suggest me a better shortcut, an easier way to manipulate data stream, or a guide to create aforementioned counter-box.

Thanks in advance,
Kadir

[jtlindgren]

Hello Kadir,

unfortunately there is no straightforward way to slice and dice the dataset in this manner. However, you could follow the route you suggest. In that case whats probably the easiest is to make a .lua script for the Lua Stimulator box and make that count the incoming stimulations and then output the 'train' stimulation to the classifier trainer after some count is reached and possibly use stream switch to discard the rest of the data.

The script

share/openvibe/scenarios/bci-examples/ssvep-mind-shooter/scripts/performance-test-classifier.lua

would be a good starting point, it seems to have the main concepts in it, you just need to change what it does.


Happy hacking,
Jussi