Gaze Analysis System Design

This document elaborates on the system design for analyzing raw eye movement data using Python, focusing on the EyeMovement class.

Input Data Format

The system processes .csv files in the following format:

YYYYMMDDHHMM.csv

Containing JSON-formatted attributes:

{
  "time": "timestamp",
  "left_gaze_point_on_display_area": "(x, y)",
  "left_gaze_point_validity": "0 or 1",
  "right_gaze_point_on_display_area": "(x, y)",
  "right_gaze_point_validity": "0 or 1",
  "left_pupil_diameter": "value",
  "left_pupil_validity": "0 or 1",
  "right_pupil_diameter": "value",
  "right_pupil_validity": "0 or 1"
}

System Configuration

Monitor Specifications

Resolution: 1920x1080 pixels
Size: 609.2mm x 349.4mm

Distance to Monitor

65 cm from the participant

Data Processing

Preprocessing

Eye Selection: Choose the eye with more valid data.
Data Interpolation: Linear interpolation for missing data (validity 0), up to a gap of 4 points.
Blink Identification: Detected when 5 or more consecutive frames are missing for both eyes.

Analysis

Saccade Detection: Calculating movement speed between frames, comparing to a threshold (20 degrees per second).
Fixation Identification: Instances lasting at least 6 consecutive frames are tagged as fixations. The center point of fixations (over 100ms) is calculated.

Error Handling

Instances not meeting the criteria for fixation or saccade are classified as error states.

Visualization Techniques

Heatmap: Uses a Gaussian matrix for gaze points, color-coded with the jet colormap.
Scanpath: Shows sequential gaze movement.
Pie Chart: Displays percentages of fixations, saccades, blinks, and errors.

System Design Features

Threading: Enables multitasking, such as simultaneous data collection, stimuli display, and user input handling.
Modular Classes: Ensures compatibility with different eye trackers and integration into Psychopy Builder.

Comparison with Tobii Pro Lab

Different criteria for fixation identification and visualization techniques.

Future Work

Explore alternative methods for saccade and fixation identification (I-DT method).
Automate extraction of eye movement features.

Visit our GitHub repository for more information and updates.

shiroyang/EyeTracker