From a computational perspective much can be learned from studying the brain. For auditory processing three biological attributes are presented as being responsible for good hearing performance in challenging environments: Firstly, the scale of biological cell resource allocated to the sensory pathway and the cortical networks that processes auditory information. Secondly, the format that information is encoded in the brain of precisely timed spiking action potentials. Finally, the adaptation mechanisms generated by the descending feedback projections between regions of the brain involved in hearing. To further understand these attributes using simulation a digital model of the complete auditory pathway must be built; the scale of such a model requires that it is mapped onto a large parallel computer. The work presented in this thesis contributes towards this goal by developing a system that simulates the conversion of sound into spiking neural action potentials inside the ear and the subsequent processing of some auditory brain regions. This system is intentionally distributed across massively parallel neuromorphic SpiNNaker hardware to avoid large scale simulation performance penalties of conventional computer platforms when increasing the number of biological cells modelled. Performance analysis as scale varies highlights issues within the current methods used for simulating spiking neural networks on the SpiNNaker platform. The system presented in this thesis has the potential for expansion to simulate a complete model of the auditory pathway across a large SpiNNaker machine.
Date of Award | 1 Aug 2020 |
---|
Original language | English |
---|
Awarding Institution | - The University of Manchester
|
---|
Supervisor | James Garside (Supervisor) & Dirk Koch (Supervisor) |
---|
- large scale
- auditory pathway
- cochlear modelling
- SpiNNaker
- neuromorphic hardware
- spiking neural networks
- parallel computing
Spikes from sound: A model of the human auditory periphery on SpiNNaker
James, R. (Author). 1 Aug 2020
Student thesis: Phd