The Prevalence of Correspondence Computations in Visual Processing

Visual Cognition;Correspondence Computation;Attention;Working Memory;Motion Perception;Psychology

Ma, ZhengMcCloskey, Michael E. ;

Johns Hopkins University

关键词: Visual Cognition;    Correspondence Computation;    Attention;    Working Memory;    Motion Perception;    Psychology;   

Others  :  https://jscholarship.library.jhu.edu/bitstream/handle/1774.2/39709/MA-DISSERTATION-2016.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: JOHNS HOPKINS DSpace Repository

【 摘 要 】

Correspondence computation is the general process during which received perceptual signals are assigned to the same or different sources. It is a pervasive process involved in many visual cognitive tasks. People need to make this computation both for simultaneously received signals and for temporally separated signals. Due to the noisy nature of visual input signals, it is an error-prone process. Therefore, it serves as the cause of many limits in human performance. However, its role in human cognitive abilities has often been ignored or underestimated. Besides, how correspondence computations are done, and how it is related to other computations in the visual system, haven’t been explored a lot. In this dissertation, I combined evidence from human behavioral experiment, computational modeling, and testing of brain-damaged patient to investigate these questions.In Chapter 2, I showed that human participants couldn’t accurately report the number of presented objects in a typical visual working memory task. I then showed that a clustering algorithm with noise in the correspondence process could simulate human performance very well. Therefore, parts of the limits in human memory ability could be explained by imperfect correspondence computations.In Chapter 3, I explored how different correspondence computation algorithms are combined to solve the problem of motion direction judgment. I found that a lower- level luminance transient detection system and a higher-level position comparison system work complementary to each other. The relative contribution of each system depends on the signal strength of that system.In Chapter 4, I showed that limit in object tracking ability is a result of noisy visual inputs, suboptimal eye-movement strategy, and probabilistic correspondence computations. No external resource-like limits are needed to understand human’s limited capacity in tracking multiple objects at the same time.In sum, these results suggest that correspondence computations play important roles in visual cognition. It is pervasive, and a failure in this process could lead to further failures in related tasks. Human visual cognition should be understood in terms of the involved computations and the possible errors could arise during these computations.