Fixation Extraction

The fixation_extraction submodule provides methods to infer fixations from raw gazes.

While squashing gazes, we would like to extract fixations and keep the path trajectory. Thus, the general approach is as follows: mark each gaze with \(0\) (not a part of fixation) or \(1\) (a part of fixation), and then squash consecutive ones into single fixation. Here we denote \(n\) fixations as triplets \(\{(x_i, y_i, t_i)\}_{i=1}^{n}\) - \(x\) coordinate, \(y\) coordinate, timestamp.

There are three main algorithms to accomplish that:

• IDT
• IVT
• IHMM

Reference

Algorithms are taken from Salvucci & Goldberg (2000). Identifying saccades and fixations in eye-tracking protocols.