On identification of elementary motion detectors

Abstract

The classical mathematical elementary motion detector (EMD) model stimulated with sinusoidal and pulsatile input signals is treated analytically. Drifting sinusoidal gratings are often used in insect vision research, enabling direct comparison with biological data. When displayed on a cathode ray tube monitor, the sinusoidal grating is modulated by the refresh rate of the monitor. Due to the resulting pulsatile nature of the visual stimuli and the corresponding biological response, a Laguerre domain identification method for estimating the dynamics of a single EMD appears to be suitable. A pool of spatially distributed EMDs is considered as the model for the measured neural output. The weights of the contributing EMDs are evaluated by a sparse optimization method to fit the experimental data.