Gaze Control during Scene Viewing: Behavioral and Computational Approaches

当我们观看视觉世界时，我们的眼睛以每秒三次的速度从一个位置跳到另一个位置，这种运动被称为扫视。有用的视觉信息只有在注视时才能获得，也就是目光停留在物体或场景特征上的短暂时间。通过一个场景实时引导扫视和注视的认知和神经过程被称为“凝视控制”。这个项目的重点是揭示人类的凝视控制如何在活跃的现实世界场景感知中运作。这个项目从理解眼球运动的时间将为了解控制凝视的潜在认知和神经系统提供关键的见解开始，研究人类的凝视控制。该研究将创新的眼球追踪方法与模拟眼球运动控制的工作计算模型相结合。在本研究计划中，经验和计算线程是互补和协同的。一方面，我们可以通过确定模型是否能产生与人类相似的眼球运动来测试我们对凝视控制的理解。另一方面，该模型的见解可以作为一种工具来增强我们对凝视控制的理论认识，这些见解可以通过新的实验进一步验证。这个项目的结果将加强对人类如何感知和理解视觉世界的基本科学理解。该项目还对创造新的显示技术和可以通过眼球运动控制的机器界面具有广泛的影响。研究结果对于设计能够主动跟踪和关注环境中相关信息的新型人工视觉系统具有重要意义。

When we view the visual world, our eyes flit from one location to another about three times per second, in movements called saccades. Useful visual information is acquired only during fixations, brief periods of time when gaze rests on an object or scene feature. The cognitive and neural processes that direct saccades and fixations through a scene in real time fall under the term 'gaze control'. This project focuses on unraveling how human gaze control operates during active real-world scene perception. This project approaches human gaze control by starting with the insight that understanding eye movement timing will provide key insight into the underlying cognitive and neural systems that control gaze. The research combines innovative eye-tracking methods with a working computational model that simulates eye movement control. In this research program, the empirical and computational threads are complementary and synergistic. On the one hand, we can test our understanding of gaze control by determining whether the model can produce eye movements that look like those produced by people. On the other hand, insights from the model can be used as a tool to enhance our theoretical understanding of gaze control, and these insights can be further tested with new experiments. The results from this project will enhance basic scientific understanding of how humans perceive and understand the visual world. The project also has wide-ranging implications for the creation of new display technologies and machine interfaces that can be controlled by eye movements. And the results are relevant for the design of new artificial vision systems that actively track and focus on relevant information in the environment.