Accuracy in Tobii Eye X

[Rasa Bhattarai]

Hello Tobii team,

I am using Tobii Eye X for my game. I recently came across an Research article which was about calibrating tobii Eye X and it mentioned:
“Near the center of the monitor, accuracy and precision can be considered to be < 0.4◦, and < 0.2◦ respectively. At more than 5 degrees away from the center of the monitor, accuracy and precision worsen to < 0.6◦, and< 0.25◦ respectively.”

Would you explain what does it mean ? I did not understand in terms of degree ? What does it mean by degrees ?

Also, some papers mention that error is greater than 0.2 degrees ? What does it mean in terms of degrees ? How do I calculate it ? Is there any formula ?

Thank you. Any help would be greatly appreciated.

[Grant [Tobii]]

Hi @rasa, the publication you referenced was not produced by Tobii, so I can only speculate but it seems that when they mention:

At more than 5 degrees away from the center of the monitor, accuracy and precision worsen to < 0.6◦, and< 0.25◦ respectively.

I would interpret that to mean that the angle between the screen and the imaginary line leaving the screen to the front of the users’ head.

For example, if you were looking straight ahead with your eyes in the middle of the screen, this angle would be around 0, but if you were to look at the screen non directly (imagine facing directly the far-left side of the monitor whilst looking toward the centre) then this angle would increase…hopefully I have made it clear.

In terms of what the ‘error angle’ signifies, I would take this to refer to the maximum angle of difference between the intended object of fixation (say an ‘x’ on screen) and where the gaze data is actually reported.

Whilst not specific to the EyeX, we have produced a number of documents that explain these concepts in greater details.

You can start here @ https://www.tobiipro.com/learn-and-support/learn/eye-tracking-essentials/what-affects-the-accuracy-and-precision-of-an-eye-tracker/ for an introduction and therafter can browse the whitepaper @ https://www.tobiipro.com/siteassets/tobii-pro/learn-and-support/use/what-affects-the-performance-of-an-eye-tracker/tobii-test-specifications-accuracy-and-precision-test-method.pdf/?v=2.1.1

Please let us know if we can be of any further assistance.

[Rasa Bhattarai]

Hello sir,

Thank you for providing with such detailed information.
I do have few concern though:

1. On the second paragraph of Conclusion of the Paper it mentions that “To ease the process of data gathering and parsing, an open source software tool (Accuracy Test Tool) that collects the gaze data and performs the metrics calculations has also
been introduced.” I could not find this open source software. Is it available only for Tobii pro eye trackers because the one I have is Tobii Eye X ?

2. Also, does the above mentioned details regarding accuracy, precision etc. from the research article also valid to Tobii Eye X ?

3. Do we have to create the test experiment designs by ourselves or we can use some software to measure the accuracy and precision of Tobii Eye X ?

I now have a clear understanding of accuracy and precision in this regard but I do not know how to apply this to Tobii Eye X.
Thank you.

[Grant [Tobii]]

Hi @rasa,

Is it available only for Tobii pro eye trackers because the one I have is Tobii Eye X ?

That’s correct. It may be possible to use the software in conjunction with the EyeX but this requires the purchase of a special unlock licence (the same licence you would need to use the EyeX with the Tobii Pro SDK)

oes the above mentioned details regarding accuracy, precision etc. from the research article also valid to Tobii Eye X ?

No, the whitepaper quoted has only been verified to be applicable to the EyeX. No internal document has been produced comparable with the EyeX in mind as this is intended as a consumer/gaming product and not for scientific or analytical use.

Do we have to create the test experiment designs by ourselves or we can use some software to measure the accuracy and precision of Tobii Eye X ?

Yes, in the same way that the author of the research article (Gibaldi et al) did perform, but I would say that you can cite their results directly for your own use. However, please bear in mind that using the EyeX for analytical or research purposes is expressly forbidden in the licence agreement without the purchase of a special licence to do so.

Further details @ https://analyticaluse.tobii.com/?utm_source=developer.tobii.com

[Rasa Bhattarai]

Hello,
Thank you fro your reply. I want to create an experiment by myself just as the author of the above paper. Initially, I have created a stimulus in unity and using Tobii eye tracker to calculate the gaze points. I have read so many papers and it mentions about the importance of visual angle in such calculation. I really need help on this. My concerns are :-
1) In my experiment, an object(circle) is shown at center of screen. I can calculate the screen coordinates at this point using unity and it shows (600,280). Now after 2 seconds the position of the object changes to (800, 280) then to (600, 400) etc. My question is: Based on these screen coordinates position, how do I calculate visual angle ?
2) I have read the formula for visual angle. From the formula, i can calculate the distance between screen and observer but how do I know about the size of the object on the screen ? It is measured in pixels but the formula requires same unit of measurement for size of object and distance between visual display and observer. How do I convert the screen coordinates point to physical unit ?
Any help would be highly appreciated. Thank you.

[Rasa Bhattarai]

For detail explanation of my above question:
On the last page of the tobii Article, it mentions ” The nine stimuli points are placed within a 20° visual angle. See detailed information in the “Accuracy and Precision Test Method
for Remote Eye Trackers” specification at Tobii.com”.

My question is how can we calculate 20 degree visual angle while placing the stimuli points ? Would you please elaborate this ?

Thank you

[Grant [Tobii]]

Hi @rasa, I would say the easiest thing for you to do (and the quickest) is to first make an assumption on the distance between the user and the screen..lets say 30 centimetres. (this is nominal, but you could do some testing over a period of time to determine an average)

So, if we place the first (lower left) stimulus point at (100,100) then the next stimulus point along the x axis would be 20 degrees to the right..

To get this position, this can be calculated with simple geometry: Distance to next point along x-axis = tan(20 degrees) * 30 cm = 10.91 cm

We now convert this to pixels (https://www.pixelto.net/cm-to-px-converter) assuming a DPI of 96dpi: 10.91 cm = 412 Pixels

Therefore next stimulus points would be (100+412,100) = (512,100)

You then apply this similar technique to the rest of the points. Hopefully, this procedure is clear enough.
Please let us know if we can of any further assistance.

[Rasa Bhattarai]

Hi @grant,

Thank you very much for helping out.
I need your assistance in one more thing: I have gaze points but I need to find the angle. For example:- The distance between observer and screen is 30cm. In the screen an object moves in an elliptical motion. So how do i calculate the visual angle made in every 0.5 seconds ? Can the tangent formula be used ? Because in the elliptical motion the object moves at so many random angles, so how to assume perpendicular and base value to calculate the tangent ?

Just this query. Thank you for your assistance.

[Grant [Tobii]]

Hi @rasa, what I would recommend is that at each 0.5 second interval you calculate the distance (call it stimulus point distance) between the start location and end location using Pythagoras theorem and with this “stimulus point distance” you can determine the approximate angle travelled relative to the user.

For example: let’s say at time=0 : stimulus location = (100,100) pixels
at time=0.5 : stimulus location = (200,250) pixels

=> distance travelled = sqrt ((200-100)^2 + (250-100)^2) = 180.27 Pixels = 4.86 cm (refer to previous post link to calculate)

=> angle travelled between stimulus points = tan-1 (4.86 cm / 30 cm (user distance) ) = 9.2 Degrees

This simplified way of doing things does make some rather large assumptions such that you are using the average distance between the stimulus screen plane and users eye and not the actual distance between the stimulus point and the user eyes and indeed that the user distance is unchanged between stimulus points.

That being said, if you follow this simplified workflow you should be able to get the numbers you need. Let me know how you get on.

[Rasa Bhattarai]

Hi @grant,

Thank you for very clear explanation. I have one query: Is the above method for calculation of the visual angle same as calculating saccades amplitude (Angular distance traveled by the eyes) ?

Thank you

[Grant [Tobii]]

Hi @rasa, no the method I described would not be suitable to calculate the saccadic amplitude as this metric is determined between *fixations* of eye movements, not raw gaze locations which are inherently noisy.

You would need to integrate a fixation filter yourself from the literature (we do not provide one) in the code and thereafter determine the distance travelled between the two fixation points over a time period.

A number of third parties have created Matlab fixation filters which perhaps you can translate for your purposes in Unity:
https://se.mathworks.com/matlabcentral/fileexchange/56236-djangraw-gazevistoolbox
& one in R:
https://github.com/davebraze/FDBeye/wiki/Researcher-Contributed-Eye-Tracking-Tools

So, either you create the Fixation Filter yourself or work with the Tobii Pro Lab Software (https://www.tobiipro.com/product-listing/tobii-pro-lab/) which will determine this data for you.

In any event, you need to be aware that any analysis of data undertaken using the EyeX and associated SDK’s (Unity, Core SDK, etc) in expressly prohibited without the purchase of a special analytical licence. More information regarding this can be found @ https://analyticaluse.tobii.com/

[Author]

Posts