In this paper, we propose to generalize the saccade target method and state that perceptual stability in general arises by learning the effects one's actions have on sensor responses. The apparent visual stability of color percept across saccadic eye movements can be explained by positing that perception involves observing how sensory input changes in response to motor activities. The changes related to self-motion can be learned, and once learned, used to form stable percepts. The variation of sensor data in response to a motor act is therefore a requirement for stable perception rather than something that has to be compensated for in order to perceive a stable world. In this paper, we have provided a simple implementation of this sensory-motor contingency view of perceptual stability. We showed how a straightforward application of the temporal difference enhancement learning technique yielding color percepts that are stable across saccadic eye movements, even though the raw sensor input may change radically.
Haihui Wang, Haihui Wang,
"A model of color vision with a robot system", Proc. SPIE 6033, ICO20: Illumination, Radiation, and Color Technologies, 60330O (11 January 2006); doi: 10.1117/12.668080; https://doi.org/10.1117/12.668080