Perception & Attention Lab
Faculty:
Dr. Evan M. Palmer
432 Jabara Hall
Graduate Students:
Christopher M. Brown
447 Jabara Hall
cmbrown1@wichita.edu
Air Traffic Visualization
Research Topics:
We study the computational demands and temporal dynamics of perception and attention. Our work covers both basic and applied areas and seeks to understand how human visual performance can be understood and improved. We employ a psychophysics approach to the study of vision and test human observers in computer-based behavioral experiments. Computational modeling is employed to evaluate and extend our findings. Below, we describe several of our ongoing research projects.
Many air traffic management displays use aircraft icons that are all the same size and shape. The three-dimensional position of the aircraft are depicted through two different information channels. Latitude and longitude are depicted graphically by the position of the aircraft icon on the radar scope, and altitude is depicted alphanumerically via the icon’s associated data tag. We believe that the alphanumeric data tags may be creating a bottleneck in the air traffic manager’s ability to visualize the three-dimensional airspace.
To alleviate this bottleneck, we add the perceptual depth cues of relative size and aerial perspective to aircraft icons in a manner that is correlated with altitude. Assuming an overhead viewing perspective, aircraft with higher altitudes are larger and/or darker and aircraft with lower altitudes are smaller and/or lighter. We have established that adding these cues greatly improved observers’ ability to detect aircraft conflicts. In terms of capacity, observers who searched through displays with both size and contrast cues are able to process about five more aircraft at a given level of performance than those who search through displays with no added cues.
One advantage of our altitude encoding scheme is that it does not rely on the use of multiple colors, which are often reserved for special purposes (such as warnings or weather depictions). An additional benefit of our grayscale implementation of altitude encoding is that it reduces visual clutter that can result from overuse of salient visual cues.
Visual Search
Classic theories of attention and visual search model the process as proceeding through a series of ordered stages, akin to information passing through a bank of filters. Recent data and theories suggest that guidance to a target’s location may be more iterative, with information moving up and down the hierarchy such that attentional selection evolves over time. To examine this possibility, we have been running a series of experiments in which bits of target information (e.g., color, orientation) are revealed over time as an observer searches a display. The data indicate that there are both immediate and evolving benefits to receiving target information before the onset of the search display. These results suggest that observers are indeed able to set up “filters†for particular kinds of searches, but that there is also an iterative “feedback†system that plays out over several hundred ms and helps to speed search.
Shape Perception
This project aims to describe the visual processes that allow us to perceive dynamically occluded objects—moving objects whose projection to the eyes is interrupted by other intervening objects or surfaces. An example would be looking through a thick hedge and seeing a car drive down the street. Gaps between the leaves admit only small moving fragments of the car, yet we perceive a whole connected object.
We have proposed the model of Spatiotemporal Relatability to explain how the visual system processes dynamically occluded objects. First, when parts of an object become occluded they nonetheless continue to be part of the observers’ visual experience. Second, the occluded yet visible fragments appear to travel along behind the occluder in the same direction as previously observed. Finally, currently visible and occluded but persisting fragments are perceptually combined via contour interpolation to form whole, complete shapes.
Perception of dynamically occluded shapes is quite robust under normal, ecological circumstances, but sometimes a powerful illusion can occur. If one piece of an object becomes occluded and is not seen again, it appears to slow down. This can lead to robust illusions of misalignment that we have been investigating for some time now.
Computational Modeling
One major modeling project is the mathematical analysis and modeling of response time (RT) distributions from various visual search tasks. We have developed a novel non-parametric statistical method for pooling data across participants to create group RT distributions. This method allows one to warp distributions from different search tasks into the same normalized space, combine them across subjects, and then compare them on equal footing. One surprising finding from this analysis is that variations in set size make no difference in the shape of normalized distributions for various tasks, contrary to predictions of many models of visual search. We are currently examining this finding to determine its implications for models of visual attention.
Another project examines the possibility of uniting signal-detection and classic two-stage models of visual search into a common approach. In the literature, these two accounts of visual search have each been quite successful in their own domains, but have been hard to reconcile with each other since each conceptualizes the processing of information in different ways. Through a series of experiments, we have been able to establish that different kinds of visual searches require different kinds of decision rules in a signal-detection framework. This means that the way observers collect information and perceive search displays changes as function of the type of search. This finding provides a bridge between these two styles of visual search models since classic two-stage models can predict what kinds of decision rules should be used within a signal-detection framework.
Collaborators
Visual Attention Laboratory:
http://search.bwh.harvard.edu
Human Perception Laboratory:
http://kellmanlab.psych.ucla.edu