Keynotes | K&Ts | GACs | Talks | Posters | Search

Dynamical Neural Model of Establishing Perception-Action Coupling through Hebbian Learning

Xinrui Jiang¹, Martin A. Giese¹; ¹Eberhard-Karls-Universität Tübingen

Presenter: Xinrui Jiang

Visual perception and action execution interact continuously, yet the mechanisms linking neural representations for vision and motion remain poorly understood. We present a dynamic neural model for perception--action coupling, consisting of interconnected visual and motor neural fields together with a visual front end that maps hand-movement images onto the perceptual representation. The two fields represent perceived and executed actions as distributed activity patterns, and reciprocal visuomotor coupling is established through Hebbian learning from the co-activation of sensory and motor patterns. Simulations show that the learned model captures key experimental findings on perception-action coupling within a unified framework. In particular, it shows delay-dependent modulation of visual responses during concurrent action execution, as well as the effect of visuomotor mismatch on stability of the visual-motor interaction. These results suggest that perception-action interaction effects can emerge from learned coupling between visual and motor representations in a framework that allows mathematical analysis.

Topic Area: Methods, Tools, Theory & Neural Coding