Mapping eye-movements to time series


Our dominant eye-movements are along horizontal (x) &  vertical (y) dimensions which map easily onto computer screen coordinates.   As our eyes search a scene for a target we record corresponding x and y positions of successive eye-movements & produce (x &) y time-series plots like the two shown below. The time series resembles Brownian motion as does the x & y (raw) eye fixation series in some of the search demo conditions.  


 

eye


<--------horizontal (x) movement---------->

 vertical eye mvmt axis



X and Y movements are represented by the red dots below,
the green dot illustrated how these time series are produced. Each are 1-D cases of brownian motion


Horizontal (X) eye movements as a function of time (t)

 Vertical (Y) eye-movements as a function of time (t)
horiz eye mvmts     time

  

 


horizEyeMvmt


horizEyeMvmt
 

refreshRefresh your browser to repeat the animation & its generation of the brownian time series.refresh
Rotating the graph 90deg  gives us the standard horizontally oriented series
(enlarged swf version)		 These brownian motion animations are modification of stock market trends used in the Yale fractal class .

An important characteristic of brown noise: increments are random!
Note: Brownian noise includes brown (1/f^2) & pink (1/f) forms of noise.

rg

Preliminary demo results show evidence for brownian  eye-movement . 
  • Microsaccades during -- steady fixations--- show clear signs of brown noise.
  • Microsaccades & saccades during -- search-- show instances of brown and pink noise. Future work is aimed at understanding which conditions induce these different forms of colored noise.
D.J. Aks 1/15/06