- 04/05/2017 at 16:26 #6798ErikParticipant
Hello. I have two questions.
1. I need to help with algorithm for detect a fixation from X,Y position of eye movement as input. Do somebody have any algorithm in code? (C,js,…)
2. How could I know that participant was confused by his fixations (with positions of gazepoint and timestamp). What is your opinion on it? The best metric/algorithm to identify a confusion.
Thanks a lot.07/05/2017 at 22:46 #6809Grant [Tobii]Keymaster
Hi @erko94, could you kindly confirm the eye tracker model you are using please? As you may be aware, the Tobii Eye Control software provides fixation algorithms internally with fixation data exposed in the SDK as the start and end time and fixation position.
Certainly, there is a great deal of literature relating to the detection of fixations in general, with one such technique known as IV-T Filter. http://acuity-ets.com/downloads/Tobii%20I-VT%20Fixation%20Filter.pdf
If you could please clarify the software and hardware you are using, that would be helpful to provide a more complete answer for you.07/05/2017 at 23:11 #6810ErikParticipant
Thanks for your answer. So I am using EyeX (Tobii) and SW that I am using is Gazehook (http://ux.fiit.stuba.sk/GazeHook/Home/Docs) where I get X,Y coordinates of right and left eye. I need to detect fixations or the main goal is detect a confusion and then my system display a help message how the user should continues.08/05/2017 at 22:53 #6814Grant [Tobii]Keymaster
Hi @erko94, thanks for the clarification. You might already know that the long term storage of gaze data is prohibited by the SDK licence agreement as supplied with the EyeX although the gathering of fixation data might not necessarily be in violation of the agreement in you case, however have you looked at the fixation information as exposed in the Tobii EyeX SDK? Perhaps this is enough for your needs?
In any event, the calculation of fixations is something handled at the lower level of the API and not generally something we release publicly. Please check out the existing fixation stream and let us know how it suits.
- You must be logged in to reply to this topic.