Sign up Calendar Latest Topics
 
 
 


Reply
  Author   Comment  
Geenaianni

Junior Member
Registered:
Posts: 27
Reply with quote  #1 
Hi, 
I am running ML2, downloaded 01/08/18, windows 10, matlab 2017b. My rig contains a PCIe6343, and a BNC2090A. 
 
I am attempting to route analog eye-signals into ML. My eyetracker outputs +/- 5V, via two BNC coax cables. My wiring is as follows:
 
Eye X (via bnc coax 1) --> AI0 on BNC2090A
Eye Y (via bnc coax 2) --> AI1 on BNC2090A
 
I have configured the BNC2090A to "nonreferenced single-ended mode". SE/DIFF switches for AI0 and AI1 are set to "SE", and  RSE/NRSE Switch set to "NRSE", to ground to AI SENSE. 
 
Finally, I have set the "AI configuration" option in Monkeylogic main panel to "non-referenced single end". 
 
When I attempt to test this input by running "I/O Test", I am able to see a small red dot on the eye position plot, but it does not move as a test subject moves their eye. I am sure that the bnc coax 1/2 carry voltage changes, as they are able to be read by a separate computer just fine. 
 
I am wondering if this has to do with gain -- could it be that the gain must be adjusted, via one of the the eye calibration methods, in order for ML to transform the voltage changes to appropriate distances in space? If this is the case -- or if not -- please advise. 
 
Thanks in advance.
best,
G
0
Jaewon

Administrator
Registered:
Posts: 708
Reply with quote  #2 
Hi Geena,

It appears that your hardware setting is fine.  "I/O Test" displays signals for the full input range (-10 V to +10 V) and does not adjust gain, so, if you want to test eye signals with gain, do it with the "Origin & Gain" calibration tool please.

You can change the shape of the eye tracer in the main menu. A line is the default shape but it is difficult to see, if there is no movement.
0
Geenaianni

Junior Member
Registered:
Posts: 27
Reply with quote  #3 
Hi again, 

we are experiencing some strange issues using ML in conjunction with ISCAN ETL-200. This eyetracker is supposed to output voltages of +/-5V. my current configuration is as in the previous post. 
When I use the origin&gain method to calibrate my eye signals, it requires gainX ~= +7 for the subject's eye to reach the target directly to the right of center. With this value for gainX however, the subject's eye could not reach the target directly left of center. Eventually, I realized that to reach the target directly to left of center, it required gainX ~= - 10. As it is my understanding that negative gain inverts the polarity of the signal, I suspected my eyetracker was outputting only positive voltages (resulting in only the right-half of the horizontal screen being represented). When I used NIMAX to visualize the voltages sensed by my DAQ card, this turned out to be correct. The DAQ card (and ML, as a result) is receiving voltage changes of ~+0.5 - + 1.5V, with no negative voltages. 

I am pursuing this issue with the eyetracking people separately, but for now, two questions for ML  --- 
1) if the eye signal is only represented by positive voltage changes, is it possible to set the "origin" or zero-point at the left-most fixation point in O&G, rather than at the center? According to my understanding, this would make the left-most part of the screen represented as very small, positive voltages and the right-most screen as larger, but still positive voltages.

2) Am I correct in thinking that it is impossible to calibrate in O&G when gainX is positive for right-sided targets and negative for left-sided targets? 

Thanks much. 
Best,
Geena


0
Jaewon

Administrator
Registered:
Posts: 708
Reply with quote  #4 
The eyetracker doesn't know where the center is, so whether its output is positively skewed or negatively skewed doesn't mean anything. Whatever output the eyetracker sends out, it is the calibration that determines the final representation. The origin & gain converts input voltages into visual angles like the following.

visual_angle = (raw - origin) * gain

Since you said the range of your eyetracker's output is 0.5 to 1.5V, the origin should be around 1V. If the gain is 1, the range of visual_angle is:

visual_angle = ([0.5 1.5] - 1) * 1 = [-0.5 0.5]

If the gain is -1, visual_angle = ([0.5 1.5] -1) * -1 = [0.5 -0.5], so the direction of the visual_angle is reversed.

Your problem is that the saccades of one direction are overshoot and the others are undershoot. There can be two reasons. First, your origin may not be at the center of the input range. If the origin is 1.2 in the above example, the range of visual_angle becomes ([0.5 1.5] - 1.2) * 1 = [-0.7 0.3], so the leftward direction is over-represented. The other reason is that the visual field itself may not be symmetric. Maybe your camera is filming from the far left and therefore is not capturing rightward movements well (or the other way around). In this case, you should try adjusting the position of the camera. The origin & gain method assumes that visual space is evenly represented in the input. If it is not the case, you should use 2-D spatial transformation, which allows you to correct more severe distortions (but requires more fixation points to be calibrated).
0
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.