Gain, Intercept and Pixels

This forum deals with BCI2000 configuration issues.
Locked
Elizabeth
Posts: 19
Joined: 20 Jan 2005, 16:29

Gain, Intercept and Pixels

Post by Elizabeth » 15 Feb 2005, 22:44

Hello,

I am trying to fully understand how the cursor moves in the X direction in a 1D BCI task towards targets located on the right side of the screen.

First, when setting up such a task, I am wondering what your default values are for Xgain, Xmean and Xpixelspersecond? With the default settings, approximately how fast does the cursor move across the screen? I have read in some BCI2000 studies that it does so in 2-3 seconds. Is this standard?

I have X adaptation turned off, Xgain = 60, Xintercept = -5 and Xpixelspersecond = 250. If I increase or decrease Xpixels, it seems to have no effect on the cursor movement. What exactly does this variable do? The manual says it controls the rate of horizontal cursor movement, but I don't see it doing that.

If I make Xintercept positive, the cursor will not move. As I make it increasingly negative it will move faster. Increasing Xgain also makes the cursor move faster.

In the recent PNAS paper, there is an equation given for how cursor movement is calculated in the horizontal direction:
Mh=ah(wrhRh + wlhLh +bh)

I assume ah corresponds to Xmean and bh corresponds to Xgain. Since there are no weights associated with horizontal cursor movement, the equation becomes:
Mh=ah * bh

It states in the article that a positive value will make the cursor move right and a negative value left. Why then would Mh = -5 * 60 make the cursor move right as it does with our current settings?

I appreciate your help!

Thanks,
Elizabeth

gschalk
Posts: 615
Joined: 28 Jan 2003, 12:37

Re: Gain, Intercept and Pixels

Post by gschalk » 17 Feb 2005, 08:56

Elisabeth,

These are very good questions. Here's what I think.
First, when setting up such a task, I am wondering what your default values are for Xgain, Xmean and Xpixelspersecond? With the default settings, approximately how fast does the cursor move across the screen? I have read in some BCI2000 studies that it does so in 2-3 seconds. Is this standard?
There really are no default values. However, you are right in that usually, cursor movement lasts about 2-3 seconds. Thus, you can use any combination of XGain and XMean to produce a trial that lasts as long as you want. Initially, when the subject is not well trained, we might have longer trials (but not longer than about 5 seconds) and then speed things up as the subject gets better.

I have X adaptation turned off, Xgain = 60, Xintercept = -5 and Xpixelspersecond = 250. If I increase or decrease Xpixels, it seems to have no effect on the cursor movement. What exactly does this variable do? The manual says it controls the rate of horizontal cursor movement, but I don't see it doing that.
If you have XAdaptation turned on, XPixelsPerSecond has no effect. XMean/Gain always holds the currently used values for the mean and the gain, and XPixelsPerSecond is a target value for the gain if adaptation is turned on. In this case, the computer will change XGain so as to produce a cursor 'speed' of XPixelsPerSecond (which are actually in some arbitrary units).
If I make Xintercept positive, the cursor will not move. As I make it increasingly negative it will move faster. Increasing Xgain also makes the cursor move faster.
This is because cursor movement is calculated online as delta x=(classified value from MLR - XMean)*XGain. Since position is increasing from left to right, you need negative values in either xmean or xgain to have the coordinate increase.
In the recent PNAS paper, there is an equation given for how cursor movement is calculated in the horizontal direction:
Mh=ah(wrhRh + wlhLh +bh)

I assume ah corresponds to Xmean and bh corresponds to Xgain. Since there are no weights associated with horizontal cursor movement, the equation becomes:
Mh=ah * bh

It states in the article that a positive value will make the cursor move right and a negative value left. Why then would Mh = -5 * 60 make the cursor move right as it does with our current settings?
This is correct except that ah is xgain and bh is xmean. Also, we described it as +bh instead of -bh because it makes the equation slightly easier to interpret. Thus, you are absolutely correct with your interpretations.

I hope this helps.

The Gerv

Locked

Who is online

Users browsing this forum: No registered users and 2 guests