Neuromorphic object tracking architecture, based on compound eyes, and implementation on FPGA

Satrajit Chakraborty, P. Priyanka, Sarthak Gupta, Saeed Afshar, Tara Hamilton, Chetan Singh Thakur

Research output: Chapter in Book / Conference PaperConference Paperpeer-review

Abstract

Recent findings in neuroscience, show that rapid changes in flight direction of a housefly/blowfly (mainly to track objects) are attributable to neural circuits distributed behind its photo-receptors. While tracking objects, using its compound eye structure, a fly is able to detect changes in the motion of the object quickly and changes its own motion accordingly. The working of these neural circuits may be modelled as a set of leaky integrate and fire neurons connected in a special manner to form a competitive feedback control. Based on this knowledge, we present a neuromorphic competitive control circuit utilizing an inference neuron model to control N actuators and analyze their outputs for tracking an object. This model was simulated in software first and then implemented on a Xilinx Artix-7 XC7A35T- ICPG236C FPGA board using Verilog. The results show an observable decoherence phenomenon between the neurons and support the working principle of the model.
Original languageEnglish
Title of host publicationProceedings of the 61st IEEE International Midwest Symposium on Circuits and Systems (MWSCAS), 5-8 Aug. 2018, Windsor, Ontario, Canada
PublisherIEEE
Pages668-671
Number of pages4
ISBN (Print)9781538673928
DOIs
Publication statusPublished - 2018
EventMidwest Symposium on Circuits and Systems -
Duration: 5 Aug 2018 → …

Publication series

Name
ISSN (Print)1548-3746

Conference

ConferenceMidwest Symposium on Circuits and Systems
Period5/08/18 → …

Keywords

  • computer vision
  • field programmable gate arrays
  • housefly
  • neuromorphics
  • object tracking

Fingerprint

Dive into the research topics of 'Neuromorphic object tracking architecture, based on compound eyes, and implementation on FPGA'. Together they form a unique fingerprint.

Cite this