ANALYSIS
To use eye tracking in interactive systems to its fullest potential, we need to do pre-processing beyond the abilities of the current parsers. Even when included parsers mark blinks, certain artifacts remain in the data. Removing windows/trials/instances where blinks are present will significantly decrease available data and introduce a delay in interactive systems. As such, we recommend the following processing steps for eye-tracking data.
Our analysis are applied on the following datasets
| Eye Tracker | Screen | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Author(s) | Year | Venue | Company | Device | Freq. [Hz] | Inch | Aspect Ratio | Distance [cm] | |
| S01 | Foster et al. | 2017 | Psychol Sci. | EyeLink | 1000 Plus | 1000 | 17 | 16:9 | 100 |
| S02 | Marzecová et al. | 2017 | Biolo. Psychol | EyeLink | 1000 | 500 | 19 | 4:3 | 57 |
| S03 | Krstić et al. | 2018 | EJPE | SMI | RED-m | 60 | 15.6 | 16:9 | 60 |
| S04 | Marzecová et al. | 2018 | Scientific Rep. | EyeLink | 1000 | 500 | 19 | 4:3 | 57 |
| S05 | Schuetz et al. | 2019 | ACM CHI | EyeLink | 1000 Plus | 1000 | 113 | 8:5 | 180 |
| S06 | Annerer-Walcher et al. | 2021 | Cogn. Sci. | SMI | RED250 | 250 | 24 | 16:9 | 70 |
| S07 | Felßberg and Dombrowe | 2022 | Vision Res. | EyeLink | 1000 Plus | 1000 | 27 | 16:9 | 85 |
| S08 | Hollenstein et al. | 2022 | LREC | EyeLink | 1000 Plus | 1000 | 27 | 16:9 | 85 |
| VR01 | Stein et al. | 2022 | IEEE VR | Tobii | Pro | 90 | 3.5 | 9:10 | - |
| VR02 | Steil et al. | 2019 | ACM ETRA | Pupil Labs | Add-on | 30 | 5.7 | 8:9 | - |
| M01 | Steil et al. | 2018 | ACM ETRA | Pupil Labs | Pro | 30 | - | - | - |
Our recommentations apply independently of sampling frequency and type of eye tracker.