Motion perception
The ability of sensing and reacting to motion signals is widely observed in animals and insects. Evidence show that at least two distinct motion mechanisms [1] can be identified in the human visual system. The output of the mechanisms can tell the brain the direction of the motion they sense, based on which reactions may be conducted (e.g. to run away from the preditor). One of the mechanisms (for convenience, lets call it the First-order mechanism) is primarily responsible for detecting fast and faint movement, and dominates our peripheral viewing (sights falling out of the eye's focus). Interestingly, the first-order mechanism also seems to exist even in drosophila [2]. My research is to understand the computations underlying this particular first-order mechanism that has obviously played a vital role in the survivals of many animal species. The study of the first-order mechanism in drosophila is fruitful and its computation has been well characterized by a "Reichart detector" [2]. It is sufficient to think of the "Reichart detector" as a mechanism that "suggests" for directions perpendidular to spatial patterns. For instance, if a straight line slides across in any direction, the "Reichart" detector will point to the direction that is always perpendicular to the line orientation. For long time the corresponding first-order mechanism in humans is believed to behave the same way. However, in a series of behavioral experiments and modeling studies, I show that the first-order mechanism in humans have an important integration step following the initial "Rechart" detector. Many "Rechart"-detectors are connected spatially and temporally to give rise to more complex motion percept in the next level of processing stream. In five demos, I will show that the first-order mechanism does not only generate motion percept that is perpendicular to local spatial patterns, it also integrate the initially generated local motions into a elaborated, new motion percept. First of all, Demo 1 and 2 serve as an example that there exist two distinct motion mechanisms that generate different motion percepts. When only one of the mechanism dominates, the corresponding percept dominates. In demo 1 and 2, the stimulus compositions are the same except for speed and contrast. But they look dtrastically different because in each condition they trigger mostly only one of the motion mechanisms. |
Demo 1: Fast and faint motion almost triggers the First-order mechanism only whose output "suggests" for diagonal motion.
Demo 2: The stimulus is exactly same as Demo 1 except that its contrast is enhanced and played at a much lower rate. This slower and more salient version of Demo 1 triggers a different mechanism whose output "suggests" for more vertical motion. View this in periphery will restore the motion to be diagonal. From [3]
|