Memristor-Based Computing Architecture with Advanced Signal Processing Capabilities

Olumodeji, Olufemi Akindele (2017) Memristor-Based Computing Architecture with Advanced Signal Processing Capabilities. PhD thesis, University of Trento.

[img]PDF - Doctoral Thesis
Restricted to Repository staff only until 9999.

24Mb
[img]PDF - Disclaimer
Restricted to Repository staff only until 9999.

196Kb

Abstract

Memristor-based computing architecture with advanced signal processing capabilities investigates analogue applications of memristors, particularly in image processing, combining them with conventional electronic circuitry. The concept of memristor (short for memory resistor) was first theorised in 1971 by Prof. Chua while reasoning on the theoretical grounds of of the symmetry of equations governing the fundamental passive circuit theory. This thesis can be split into two parts as follows: i) Memristor Device, Modelling, Characterisation and Programming and ii) Memristor Circuit Applications. In the first part, an overview of the theory of memristors which gives an introductory background is discussed alongside the device modelling to fit experimental data. Electrical characterisation was carried out on fabricated memristors to validate the fundamental fingerprint of these devices and finally, the different programming techniques, particularly the pulse-based technique which was widely used in this work, was exhaustively treated. In terms of applications, starting from a novel memristor-based light to resistance encoder, a more complex architecture based on adaptive background subtraction for scene interpretation used for the analyses of motion and its association to a particular object in the scene is treated in this work. Throughout this thesis, the intention of an overview on the application of memristors in image processing algorithm is emphasised and the last chapter discusses a neural network architecture based on memristors. The intended neural network architecture was trained to perform colour classification targeting applications based on gesture detection. In essence, Memristor-based computing architecture with advanced signal processing capabilities gives an insight into the advantages to come having an hybrid system of standard CMOS image processing techniques with memristive devices particularly in computation-intensive applications requiring high speed and massive parallel signal processing. Typically the realisation of such powerful and dense networks in an integrated circuit with acceptable size which is not resource hungry by using the commonly encountered elements and conventional CMOS technology is becoming increasingly difficult to achieve. Computing architectures based on memristors present the advantages that could help overcome the limitations of an overall implementation with conventional electronic elements.

Item Type:Doctoral Thesis (PhD)
Doctoral School:Materials, Mechatronics and Systems Engineering
PhD Cycle:29
Subjects:Area 09 - Ingegneria industriale e dell'informazione > ING-IND/31 ELETTROTECNICA
Funders:Autonomous Province of Trento, Fondazione Bruno Kessler (FBK)
Repository Staff approval on:18 Dec 2017 13:22

Repository Staff Only: item control page