Date of Award

Spring 5-2020

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computational and Data Sciences

First Advisor

LouAnne Boyd

Second Advisor

Deanna Hughes

Third Advisor

Vincent Berardi

Abstract

"Social communication is the use of language in social contexts. It encompasses social interaction, social cognition, pragmatics, and language processing” [3]. One presumed prerequisite of social communication is visual attention–the focus of this work. “Visual attention is a process that directs a tiny fraction of the information arriving at primary visual cortex to high-level centers involved in visual working memory and pattern recognition” [7]. This process involves the integration of two streams: the global and local streams; the global stream rapidly processes the scene, and the local stream processes details. This integration is important to social communication in that attending to both the global and local features of a scene are necessary to grasp the overall meaning. For people with autism spectrum disorder (ASD), the integration of these two streams can be disrupted by the tendency to privilege details (local processing) over seeing the big picture (global processing) [66]. Consequently, people with ASD may have challenges integrating visual attention, which may disrupt their social communication. This doctoral work explores the hypothesis that visual attention can be redirected to the features of an image that contain holistic information about a scene, which when highlighted might enable people with ASD to see the forest as well as the trees (i.e., seeing a scene as a whole rather than parts). The focuses are on 1) designing a global filter that can shift visual attention from local details to global features, and 2) evaluating the performance of a global filter by leveraging eye-tracking technology. This doctoral work manipulates visual stimuli in an effort to shift the visual attention of people with ASD.

This doctoral work includes two development life cycles (i.e., design, develop, evaluate): 1) low-fidelity filter, and 2) high-fidelity filter. The low-fidelity filter life cycle includes the design of four low-fidelity filters for an initial experiment which was tested with an adult participant with ASD. The performance of each filter was evaluated by using verbal responses and eye-tracking data in terms of visual analysis, fixation analysis, and saccade analysis. The results from this cycle informed the decision for designing a high-fidelity filter in the next development life cycle. In this second cycle, ten children with ASD participated in the experiment. The performance of the high-fidelity filter was evaluated by using both verbal responses and eye-tracking data in terms of eye gaze behaviors. Results indicate that baseline conditions slightly outperform global filters in terms of verbal response and the eye gaze behaviors.

To unpack the results in more details beyond group comparisons, three analyses (e.g., luminance, chroma, and spatial frequency) of image characteristics are performed to ascertain relevant aspects that contribute to the filter performance. The results indicate that there are no significant correlations between the image characteristics and the filter performance. However, among the three characteristics, spatial frequency is depicted as the most correlated factor with the filter performance. Additional analyses using neural networks, specifically Multi-Layer Perceptron (MLP) and Convolutional Neural Network (CNN), are also explored. The result shows that CNN is more predictive of the relationship between an image and visual attention than MLP. This is a proof of concept that neural networks can be employed to identify images for future experiments, by avoiding any variance or bias in terms of unbalanced characteristics of images across the experimental image pool.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.