Hi Jaewon, ML community,
We are using an ISCAN eye tracker that sends in eye position x,y data via analog BNC cables. We are then feeding this signal pair into an NI PCI-6221 card. We have checked with the Matlab->Apps->Analog input recorder that the incoming signal that is read into Matlab from the NI card api's, is faithful to the analog voltage levels that are coming in from the BNC cable.
The incoming analog signal has the range [-5, 5] volts. Since the signal is a Pupil-CR difference signal, the variations are small and a -20x gain on both X and Y within the ML "Origin & Gain" menu, followed by centering in the same menu match the eye position to screen dva location mapping.
What we see is a steady drift of the signal as both the analog X and Y values as shown by the red dot in experimentor visualisation screen keeps drifting down and to the left. We have to periodically interrupt the training and re-center the signal to continue training. We also find that after centering a few times, we are not able to see the red dot anymore. When this happens we find that the data.AnalogData.eye has identical readings in both x and y fields and the same number repeats across rows as well.
So far we have checked in various ways that the drift is not there in the eye-tracker output and is coming in the NI output-> ML visualisation path. It will be great if you can help us resolve this issue.
best,
Harish