Back to search

University of Southampton

Neural signal processing with metal-oxide RRAM devices

Abstract

dc:description.abstract

This thesis delves in the issue of processing huge volume of neural data recorded using in-vitro monitoring technologies. The real-time processing of neuronal signals imposes excessive strain on bandwidth, energy and computation capacity prohibiting scaling of present neural interfaces. This work offers a unique solution to the challenge of encoding electrophysiological neural bio-signal information in a power efficient way thus capable of impacting the field of neuroprosthetic applications or for instance, the emerging area of bio-electronic medicines.<br/><br/>The thesis mainly focussed on exploiting the intrinsic properties of nanoscale metal-oxidemetal devices commonly known as `memristors' for demonstrating spike detection and spike sorting at the proof-of-concept level. Memristive devices were fed with extracellular neuronal activity and the thresholded integrating property of both non-volatile and volatile devices was used to differentiate between high-amplitude spiking events and low-amplitude<br/>background noise that forms the majority of the neuronal signals samples. The spiking events of supra-threshold strength are detected as memory state transitions thus compressing information on neuronal spikes in real-time. These experiments show a substantial improvement in the bandwidth required per sensing site (200:1), while concurrently offering more energy efficient paradigm, estimated at 100nW per channel as compared to the present state-of-the-<br/>art spike detection techniques. For all the experiments, for quantification of the obtained results, the spike detection performance was benchmarked against state-of-the-art template matching.<br/><br/>Importantly, the experimental work carried for demonstrating this application also involved developing electrical characterisation modules for carrying out en-masse testing of the fabricated devices and ensure process development. Furthermore, the same concept was used to demonstrate a much more computationally intensive task i.e. `spike-sorting'. It is a procedure of identifying the activity of individual neurons from the data collected through<br/>electrophysiological experiments. The experiments performed show how the intrinsic analogue programmability of memristive devices can be used to perform the task of spike sorting.This idea can thus potentially open new avenues for performing spike-detection and sorting both on-chip using miniaturised chips at minimal power costs, demonstrating the technology's<br/>potential to build scalable, yet energy efficient on-node processors for brain-chip interfaces.

Degree

thesis:*
Name dc:type.qualificationname
Ph.D.
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Southampton
Year dc:date.issued
2018

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Gupta, Isha
Advisor dc:contributor.advisor
  • De Groot, Cornelis
ORCID person:identifier.orcid

Rights

Language dc:language
English, eng

Chain of custody

source
Harvested from
University of Southampton
Base URL
eprints.soton.ac.uk/cgi/oai2
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Gupta, Isha. Neural signal processing with metal-oxide RRAM devices. doctoral thesis, University of Southampton, 2018.