Human vs Instrument Noise:
Fixating on a stationary point

Analysis of movements:

time series

(differences across consecutive data points)

probability distributions

(pdf)

(Click images below for magnified versions)

power spectra
(FFT)

 

Descriptive statistics

&

relative variability

***********
Overall, real eye produces (over 20%) greater variability relative to artificial eye
************

Real human eye

(as seen through eye-tracker IR camera)

Fixational eye-mvmts

time series for human eye fixated on stationary point

Large range of variability*

Human eye pdf-- notice long tail

 Human eye produces brown noise in power spectra

Note in xls files:
RE=Real human eye
AE= Artificial eye...

 

Long tail(s)
Brown noise

Artificial eye

(2D reflective eye image)

Eyetracker vibrations produce data variability;
time series of stationary reflectant image of artifiicial eye

 artificial eye-- gaussian pdf artificial eye produces mostly white noise noise in this power spectra
RE results (xls)


AE results (xls)

Smaller range of variability*
Shorter tails/ Gaussian distribution

(Mostly) white noise


Instrument -vs- human eye mvmt home

In this preliminary work, distinct patterns of noise emerge across human & artificial eye. There is more white (gaussian) noise in the artificial eye condition (except in the lowest frequency region (which I believe is an artifact of limited data set). I'm collecting more data on human and different artificial eyes to confirm these different patterns of behavior. (stay tuned)

*Note: different scales on the time-series were done for illustrative purposes. There is an actual difference in variability which can be seen in the Descriptive Stats files on the right. 		Next step:
Comparing overt eye-movements in human vs robot

http://aks.rutgers.edu/EyeTrack/Noise2.html
D.J. Aks 6/19/06