Now Shipping! – The EyeLink 3; combined head and eye tracking at up to 1000 Hz.

FAQ: How is EyeLink data filtered?
#1
The EyeLink uses an online heuristic filter to decrease noise in the data output. The core principles behind the filtering algorithm are described in detail in the attached paper:

Quote:Stampe, D.M. (1993) Heuristic filtering and reliable calibration methods for video-based pupil-tracking systems. Behavior Research Methods, Instruments, & Computers, 25(2), 137-142.

.pdf   Stampe1993_Article_HeuristicFilteringAndReliableC.pdf (Size: 779.47 KB / Downloads: 55)


The filter has two levels. The Standard level is a single-stage filter that introduces a delay of one sample, and the Extra level is a two-stage filter that introduces a two-sample delay. The filter can be applied independently to both the "File Data" (the data written to the EDF file during recording) and the "Link / Analog Data" (the data transmitted over the Link to the display PC or output on the optional analog output card). By default, the File Data filter is set to Extra. This minimizes the amount of noise in the recorded data, and as this data will be analyzed after the recording there is no cost to the extra delay. For the Link / Analog Data, the Standard filter is enabled by default so that only a one-sample delay is introduced when accessing data over the Ethernet link or the analog output. This minimizes delay while also reducing unwanted noise.

Note that the filter levels can be set prior to data collection by going to the Set Options screen on the host and changing the File Sample Filter and Link / Analog Filter settings. It can also be programmatically set from the experimental script by issuing a heuristic_filter command. The command takes two arguments: one for the filter that applies to the EDF file and the other for the Link / Analog data.

Code:
heuristic_filter <file filter level> <link filter level> 

The filter levels can be set with a value of 0 (off), 1 (single-stage), or 2 (two-stage) to denote the level of filtering to be applied to each data stream.

Note that the syntax of how to send this command programmatically will vary depending on which stimulus presentation software you are using for your task.