Mu Cursor Task Classifier

Forum for discussion on different user applications
Locked
dkiviat
Posts: 7
Joined: 20 Jun 2011, 11:06

Mu Cursor Task Classifier

Post by dkiviat » 26 Jul 2011, 19:17

I'm a relatively new BCI user, and I'm trying to implement the Mu cursor task as described in the online tutorial. I'm a little fuzzy on some of the aspects of the classifier matrix:

Input Channel: According to the tutorial we simply right C3_out, C4_out etc, but assuming we are using a slightly different channel, does the name matter so long as it matches one of the outputs of our spatial filter?

Output channel: The tutorial sets this value to 2, but Im still not clear what this value denotes. If I use multiple features, does everything get outputted to the same channel, and what are other channels used for?

Weight: The tutorial sets this value to -1, so does this mean I instruct the subject to attend to the proper motor imagery to increase downward movement of the cursor (it doesn't say whether - denotes down or up, though that would be the logical assumption), and attend less to move up? Alternatively can I use two features and give them weights of opposite sign, so that the subject attends to different limbs to move the cursor in different directions? My ultimate goal is 2d movement, so which of these strategies would be more appropriate? Thanks!

jawilson
Posts: 109
Joined: 28 Feb 2005, 16:31

Re: Mu Cursor Task Classifier

Post by jawilson » 02 Aug 2011, 14:59

dkiviat wrote:I'm a relatively new BCI user, and I'm trying to implement the Mu cursor task as described in the online tutorial. I'm a little fuzzy on some of the aspects of the classifier matrix:

Input Channel: According to the tutorial we simply right C3_out, C4_out etc, but assuming we are using a slightly different channel, does the name matter so long as it matches one of the outputs of our spatial filter?
That is correct. The output labels are assigned to the channels at the output of the spatial filter. Alternatively, you can also just use the channel number instead.
Output channel: The tutorial sets this value to 2, but Im still not clear what this value denotes. If I use multiple features, does everything get outputted to the same channel, and what are other channels used for?
This is the X-Y-Z channel, where X is 1, Y is 2, and Z is 3. So, the features for all channels labelled "2" are summed and set to control the Y-direction of the cursor. Similarly for "1" and X-direction.
Weight: The tutorial sets this value to -1, so does this mean I instruct the subject to attend to the proper motor imagery to increase downward movement of the cursor (it doesn't say whether - denotes down or up, though that would be the logical assumption), and attend less to move up? Alternatively can I use two features and give them weights of opposite sign, so that the subject attends to different limbs to move the cursor in different directions? My ultimate goal is 2d movement, so which of these strategies would be more appropriate? Thanks!
That is correct, with a weight of -1, it will move in the opposite direction. Remember, motor imagery produces a DECREASE in the mu/beta rhythms, so if we compare to some baseline value, the feature value will be negative during imagery. To make this work with the expected cursor movement direction (i.e., "right hand" = "move cursor right", or "move feet" ="move cursor down"), we need the negative weight.
I BELIEVE up is now the positive-Y direction (based on http://bci2000.org/wiki/index.php/User_ ... esentation), but that was not always the case. You used to have to do like 3 mental tranformations to get the correct Y-movement.

For left/right movement, yes, you will give different weights for C3 and C4 to move the cursor in opposite directions.
Adam

Locked

Who is online

Users browsing this forum: Google [Bot] and 26 guests