A Neuromorphic Model For A Robust, Adaptive Photoreceptor Reduces Variability In Correlation Based Motion Detectors

David C O'Carroll, Paul D Barnett, Eng Leng Mah, Karin Nordstrom, Russell S A Brinkworth

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

We present here a parametric model for motion detection based on a correlational elementary motion detector (EMD) and inspired by our analysis of responses of neurons in the motion detection pathway of flying insects. This model incorporates a biomimetic photoreceptor stage that fully accounts for the nonlinear adaptive normalization of contrast observed in fly photoreceptors. We find that inclusion of this front-end leads to a substantial improvement in performance compared with a basic EMD model. Our model lends itself to elaboration into analog electronic hardware, including neuromorphic analog VLSI. We have developed an initial implementation of the photoreceptor model and a single EMD using discrete electronic components. We have tested both the hardware and digital simulations of elaborated EMD arrays using high dynamic
range (HDR) panoramic scenes derived from nature. Our data confirm that this photoreceptor model is robust enough to have a variety of applications and should be used as a front end wherever wide-field velocity information is of value (e.g. in optical flow analysis).
Original languageEnglish
Title of host publicationBICS 2006 - Brain Inspired Cognitive Systems
Number of pages4
Publication statusPublished - 2006
Externally publishedYes

Fingerprint Dive into the research topics of 'A Neuromorphic Model For A Robust, Adaptive Photoreceptor Reduces Variability In Correlation Based Motion Detectors'. Together they form a unique fingerprint.

Cite this