This thesis explores the real-time implementation of a biologically inspired pitch
detection system in digital electronics. Pitch detection is well understood and has been
shown to occur in the initial stages of the auditory brainstem. By building such a
system in digital hardware we can prove the feasibility of implementing neuromorphic
systems using digital technology.
This research not only aims to prove that such an implementation is possible but to
investigate ways of achieving efficient and effective designs. We aim to achieve this
complexity reduction while maintaining the fine granularity of the signal processing
inherent in neural systems. By producing an efficient design we present the possibility
of implementing the system within the available resources, thus producing a
demonstrable system. This thesis presents a review of computational models of all the
components within the pitch detection system. The review also identifies key issues
relating to the efficient implementation and development of the pitch detection
system. Four investigations are presented to address these issues for optimal
neuromorphic designs of neuromorphic systems.
The first investigation aims to produce the first-ever digital hardware implementation
of the inner hair cell. The second investigation develops simplified models of the
auditory nerve and the coincidence cell. The third investigation aims to reduce the
most complex stage of the system, the stellate chopper cell array. Finally, we
investigate implementing a large portion of the pitch detection system in hardware.
The results contained in this thesis enable us to understand the feasibility of
implementing such systems in real-time digital hardware. This knowledge may help
researchers to make design decisions within the field of digital neuromorphic systems.
A Doctoral Thesis. Submitted in partial fulfilment of the requirements for the award of Doctor of Philosophy of Loughborough University.