Sensory Cue Combination

Robert Jacobs
University of Rochester

This talk will overview recent research on how people integrate information from multiple sensory signals. We start by deriving a linear cue combination rule which is optimal (given certain assumptions) in a maximum likelihood sense. We then report the results of a set of experiments (due to Battaglia, Jacobs, and Aslin, 2003) studying whether this optimal model provides a good account of human subjects' responses on a spatial localization task based on visual and auditory signals. Lastly, we examine a nonlinear approach to sensory cue combination based on the idea of "competitive priors" (Yuille and Bulthoff, 1996), and examine experimental data on a visual slant judgment task (due to Knill, 2007) evaluating this approach.

Presentation (PowerPoint File)

Back to Graduate Summer School: Probabilistic Models of Cognition: The Mathematics of Mind