Eye gaze mapping"

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(dplyr)
library(ggplot2)
library(TriDimRegression)

Bidimensional regression can be used to transform the eye gaze data into a the screen coordinate system using a calibration sequence. For this, we use known target coordinates as independent variables. Please note that the example below assumes that participants fixate faithfully for most of the time and that recording artifacts, such as blinks, were already removed. This example will use the \code{EyegazeData} example dataset.

Plotting raw data

ggplot(data= EyegazeData, aes(x= x, y= y, color= target, fill= target)) +
  geom_point(data= EyegazeData %>% group_by(target, target_x, target_y) %>% summarise(.groups="drop"),
             aes(x= target_x, y= target_y), shape= 21, size= 10, fill= 'white') + 
  geom_point(alpha= 0.5, shape= 21, show.legend=FALSE) + 
  ggtitle('Raw eye gaze')

Using lm2 to transform the eye gaze

lm2aff <- fit_transformation(target_x + target_y ~ x + y,  EyegazeData, transformation = 'affine')
adjusted_gaze <- predict(lm2aff, summary=TRUE, probs=NULL)
colnames(adjusted_gaze) <- c('adjX', 'adjY')
adjusted_gaze <- cbind(EyegazeData, adjusted_gaze)


ggplot(data= adjusted_gaze, aes(x= adjX, y= adjY, color= target, fill= target)) +
  geom_point(data= adjusted_gaze %>% group_by(target, target_x, target_y) %>% summarise(.groups="drop"),
             aes(x= target_x, y= target_y), shape= 21, size= 10, fill= 'white') + 
  geom_point(alpha= 0.5, shape = 21, show.legend = FALSE) + 
  xlab('x')+
  ylab('y')+
  ggtitle('Adjusted eye gaze')


Try the TriDimRegression package in your browser

Any scripts or data that you put into this service are public.

TriDimRegression documentation built on Sept. 13, 2023, 5:07 p.m.