Hey Jihad,

W.r.t. the scene viewing vs. visual search classification, there was a recent paper SubsMatch 2.0: Scanpath comparison and classification based on subsequence frequencies by Kübler et al. (Behavior Research Methods, 2016).

W.r.t. programming: Matlab is a very good but rather expensive tool for analysing data. There is a free (but less comfortable, syntaxwise) alternative called R if your institution does not have access to ML.

For our setup we use:

• Matlab (R2013 or later for visually more pleasing, but much slower figures)
• Psychtoobox (free; for stimulus presentation)

http://psychtoolbox.org

• Signal Processing Toolbox (for data processing, e.g. filtering)

http://www.mathworks.com/products/signal

• Statistics and Machine Learning Toolbox (for statistics; more recent version will also provide you with more test options)

http://www.mathworks.com/products/statistics

For the visual degrees, use simple geometry. If d is the viewing distance in [cm] and s is the size in [cm] you want to convert to angle alpha in [deg] at the given viewing distance, then

  alpha = 360deg/pi*tan(s/(2d))    s = 2d*atan((pi*alpha)/360deg)

This is also the reason why in many setups d = 57cm in which case s = 1 cm corresponds to 1deg viewing angle.

In general I can recommend the book Eye tracking - A comprehensive guide to methods and measures by Holmqvist et al. or, for the more perfectionist type, Vision Research: a Practical Guide to Laboratory Methods by Carpenter & Robson (although this is slightly outdated, it is a VERY complete approach to building an eye tracking lab).

Greetings, David

http://psychtoolbox.org/

www.mathworks.com/products/signal

www.mathworks.com/products/statistics

Article Eye Tracking: A Comprehensive Guide To Methods And Measures

Book Vision Research: a Practical Guide to Laboratory Methods

Article SubsMatch 2.0: Scanpath comparison and classification based ...

Hello David!

Thank you as usual for your detailed answer.

Now I have the book you recommended earlier, and one thing that confuses me till now is how a calculated scanpath length in pixel is converted into degrees of visual angle? The book doesn't mention any mathematical equation.

What is the standard way of converting sth in pixel to degrees and vice versa? I know how to calculate how many pixel a 1 degree of visual angle is equal to given by another researcher, which is

pix2deg = screen resolution (hor/vert) / (2*arctan (screen physical (width or height) / 2* viewing distance))).

Is this applicable in any case when dealing with visual stimuli?

Thank you in advance for your help

Dear Jihad,

converting some size from pixels (s_px) to centimeters (s_cm) can either be done by using the screen resolution r_px and the physical screen size r_cm (measured or taken from the monitors manual):

   s_cm = s_px * r_cm/r_px

or by using the screen's pixel pitch p in [centimeters per pixel] (given in the manual):

   s_cm = s_px * p

After acquiring any size in centimeters (so also your traveling distance for the scan path) use the geometric relationship given above to convert to degree.

Cheers, David

Thank you. I will try this in my calculations. 

By the way, may I know if you are working on Mac or Windows? I read Psychtoolbox works well only with Windows 7, and better to use it on Mac. Also with your eye tracker, did you use a specific programming language such as python to synchronize between eye movement recording and viewing visual stimulus? 

We use Windows10 (which is bad). For best performance use a Linux system.But you also require good hardware, e.g. a dedicated graphics card (there is some article on the PTB page only on that matter).

All programming is done in Matlab.

Greetings, David

 Hi Jihad,

I've used Psychtoolbox in Matlab on a Mac to run my eye tracking experiment with Eyelink II for a visual search study. The eye tracker was controlled by my experiment scripts in Matlab as well. For details see the 2D experiment in my study in the publication attached. I've also heard people using PsychoPy (http://www.psychopy.org/), an open-source python package for vision experiments.  

Article Memory and visual search in naturalistic 2D and 3D environments

Hello Chia

Thank you for your answer.

May I know if your program is publicly available? Also, any advice regarding the choice of images for the 2D experiment. Measuring the eye movement during visual search is a first step in my prototype. 

Additionally, did your images span your whole monitor?

Did you use the above calculation to determine how many degrees of visual angle does your visual stimuli covers of the visual field?

Any advice is highly appreciated 

Jihad

Hi Jihad,

It's not yet. If you use the same type of eye tracker it'd be more useful to see my codes, since the code is very specific for the Eyelink I used, at least the part that were written for eye tracking integration. I can share it with you if it'd be helpful. 

I generated my images from the virtual reality environment I used for the 3D experiment, so I didn't use any image database. Perhaps you can look up some natural scene studies and see if any releases their image database. Aude Oliva has some natural images released on her site: http://cvcl.mit.edu/database.htm, and there's one from UPenn: http://tofu.psych.upenn.edu/~upennidb/. Stefan Winkler has a list too: http://stefan.winkler.net/resources.html.

Yes the images span the whole monitor. You can control that through Psychtoolbox. Whether they should span the whole width/height depends on your experimental needs. In my case I'd like to create a search condition parallel to the search paradigm in 3D virtual reality, so I had the images span the whole width.

The visual angle can be calculated by arctangent(width or height of monitor/(2*distance)). 

Recently the lab of Prof. Dra. Pia Knoeferle released a research paper where several guidelines are described for carrying out an eye-tracking experiment, especially in psycholinguistics, but you can perfectly use most of them for any other area. Currently, this document is used in my lab as a fundamental reading before running any eye-tracking experiment. Article Eye Tracking During Visually Situated Language Comprehension...

Jaime Andres Riascos Salas Thanks!