Abnormal saccade and pursuit detection

[ruw001]

Hi,

Thanks for developing this great tool. It saved me a ton of time. I used a tobii 120hz eye tracker, and conducted a series of reading tasks where we track people's eyes when they read static passage. Most of the eye-movement events look reasonable, but for some participants, there are many pursuits which I think are falsely classified. I looked at the data, it seems some small saccades are not detected in the first phase (see below), so they are included inside fixation/pursuit. Is there any way to work around that? The only parameter I changed is savgol-length. I change it to 0.025 so it fits my sampling rate.

Another miss classification is the saccade onset (below). It looks like the adjacent small saccades are not detected successfully, thus the major saccade duration is very long. What should i do?

Thanks, Ru

[adswa]

I don't have experience with reading data myself. I guess the saccades are smaller than during free viewing paradigms? I wonder if the parameters --min-saccade_duration and --min-pursuit_duration could be of help to allow shorter saccades and disallow short pursuits, and whether the default value of 2Hz for --max-initial-saccade-freq holds for reading tasks

[ruw001]

Hi Adina,

Thanks for your reply. I tweaked different parameters regarding saccade detection, and found that the saccade detection looks great when I set max_initial_saccade_freq to 4, with other parameters unchanged.

For some participants with low calibration accuracy, their data might be noisier, so super long saccade duration still happens. I can work around by further setting noise_factor to 10.

[adswa]

I'm glad you found a working solution. I hope I read correctly that I can close this issue? Please feel free to reopen if something is yet unaddressed.