EPIC Models of Auditory and Visual Tasks
Abstract
As the Navy takes advantage of new technologies and seeks to do more with fewer personnel, it is essential to design systems for human operators that are effective, reliable, and efficient. This requires understanding human abilities and limitations in a rigorous and quantifiable way, so that proposed system designs or training approaches can be developed and evaluated effectively. The best way to organize this developing understanding of human abilities and limitations is with computational cognitive architectures that provide a comprehensive synthesis of psychological results and theory, and support practical predictions of human-system performance via models that simulate human performance. This proposal seeks to further extend one such cognitive architecture, EPIC, developed in previous ONR-sponsored projects, in two areas: human audition, as applied to multi-speaker speech perception and spatialized audio, and human vision, as applied to visual search tasks typical of radar console operators. More specifically, the project will focus on developing models in the following topics, using published data, and data collected by other research groups. The models are constructed as Monte-Carlo simulations, and evaluated by statistical goodness of fit to the data.1. The current model for two-talker speech tasks that works at the level of whole words in the input will be substantially refined to use fine-grain signal and timing information in the signals to explain effects in more detail, and to support generalization to multiple talkers and other classes of competing signals.2. A preliminary model for how spatial location of sound sources is used in two-talker talks will be refined, generalized to three or more talkers, and the underlying model for spatial sound localization will be applied to the use of localized signals and cues for enhancing performance in tasks such as visual search.3. The current architectural components for vision and eye movements will be upgraded to increase the accuracy and generality of models of visual search, to be both more psychologically realistic, and to start with actual rather than pre-parsed images.4. A basic model of auditory-visual integration will be proposed based on the results from modeling the use of localized sound in visual search.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 23, 2016
- Source ID
- N000141612560
Entities
People
- David Kieras
Organizations
- Board of Regents of the University of Michigan
- Office of Naval Research
- United States Navy