When the Brain Meets the Eye: Tracking Object Motion

摘要: To accurately track a moving object of interest with appropriate smooth eye movements, the brain needs to reconstruct single velocity vector describing global motion this object. Because aperture problem (see Chap. 1), visual system must integrate piecewise local information from either elongated edges and contours or particular features such as corners texture elements. Here, we show that investigating movements unveil several dynamical properties integration stage. Signals are weighted according their uncertainties. The is highly - being always launched first in simplest, linear (vector sum) prediction. Tracking trajectories then progressively adjusted match trajectory after 200 ms pursuit. Such strategy immune higher factors prediction about incoming 2D target trajectory. On contrary, mixing retinal extra-retinal signals become important later during pursuit accommodate partial total occlusion for instance. We propose framework computing representing through two recurrent loops (V1-MT MST-FEF, respectively), area MST playing role gear. architecture would constraints motor behavior: quick reaction new event utilization out transient changes image those occurring when an moves crowded environment.