Neuromorphic engineering is the modelling of neurobiological systems in silicon and / or through computer simulation. It draws inspiration from neurobiology in an attempt to gain insight into biological processes, or to develop novel circuits and solutions to real engineering problems. This thesis describes the development of a hardware implementation of the integrate-and-fire neural network stage of the sound segmentation algorithm. It discusses issues involved in the development of novel aVLSI techniques capable of modelling biologically realistic processes. Comparisons are made between hardware and software implementation of the integrate-and-fire neural networks for real time data. Various approaches have been investigated. These include: fixed / variable interconnection, fixed / variable weight strengths, programmable biologically realistic time constants. Cascading techniques have been used to investigate the feasibility of inter-chip communication. Arguments are proposed for design modifications. Results show that aVLSI circuits can be produced which will realistically model biological processes and that there is a high degree of similarity between hardware and software implementation of the integrate-and-fire neural network. Suggestions are made for further investigation and work which include circuit modifications to increase flexibility and performance.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:651544 |
| Date | January 1999 |
| Creators | Glover, Mark A. |
| Publisher | University of Edinburgh |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://hdl.handle.net/1842/13922 |
Page generated in 0.0017 seconds