A Spatiotemporal Probe of the Human Visual System by Application of Nonlinear Systems Identification Theory

Citation

Chen, Michael Jiu-Wei (1982) A Spatiotemporal Probe of the Human Visual System by Application of Nonlinear Systems Identification Theory. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/f613-bd08. https://resolver.caltech.edu/CaltechETD:etd-09132006-154529

Abstract

This thesis describes an attempt to apply signal processing and systems theory to the task of analyzing and interpreting evoked potential data and locating evoked potential sources by physical principles. Random impulse trains were used as inputs to characterize the human visual system. The method is analogous to the Wiener method for a continuous Gaussian white noise input. The restricted-diagonal Volterra series for discrete inputs is used by making certain restrictions on the integrals in a Volterra series. A modification of Lee and Schetzen's method was used in the estimation of the kernels.

Forty-channel first-order kernels were computed for briefly appearing checkerboard patterns placed in left or right visual fields. The measured potential distribution showed a radical dependence on stimulus locus. Equivalent dipoles generally give excellent fits to the measured data, and the mapping between the visual field and these equivalent sources is similar to the commonly accepted mapping between the visual field and the visual cortex. Also, the results resemble those using conventional signal averaging.

First order kernels show better signal-to-noise ratio when compared to conventional signal averaging for the same experiment duration. Multichannel first-order kernels show that sources from early components are deep in the head as expected and in a believable region.

Results for the second-order kernels reveal occlusive interactions in the visual system and are interpreted relative to the first-order kernel. These inhibitions display different lengths of memories which suggest that they might arise from different neural origins.

Thesis Files