Abstract
Human smooth pursuit tracking eye movements were analyzed in the frequency domain. The stimulus used was a sum of sinusoids, in lieu of white Gaussian noise. The feasibility of the nonlinear analysis by a sum of sinusoids consisting of only 6 frequencies was evaluated through computer simulation with a known nonlinear system. A new method which makes use of the linear property of the discrete Fourier transform and uses the sum of sinusoids as stimulus with which to determine interaction frequencies is presented for the estimation of different order nonlinearities and primitive frequencies of a nonlinear system. The data of smooth tracking electrooculogram (EOG) responses to the sum of sinusoids from eight subjects were applied and analyzed in the frequency domain. The average power of these linear components which dominate the responses takes 63.66% in the smooth pursuit tracking responses for the eight subjects. With the notion of primitive frequencies, the outputs at the primitive frequencies of order n (n=4 at most) were detected. The proportions of both the magnitudes and powers of different order primitive frequencies were calculated and analyzed. It is significant that magnitude proportions for the eight subjects are very close to each other. The average magnitude proportions are 18.75% for order 2, 30.68% for order 3, and 50.49% for order 4; however, there are great differences among the power proportions for the eight subjects. In addition, strength indices indicate that the 2nd order primitive frequencies have more average magnitude contribution to the nonlinearity of the human smooth pursuit system than the 3rd and 4th order primitive frequencies in EOG responses to the stimulus pattern 0.
Sa, Zhijun (1995). Human smooth pursuit tracking eye movement analysis in the frequency domain. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -1995 -THESIS -S2.