Thesis (MScEng)--Stellenbosch University, 2012. / ENGLISH ABSTRACT: Several accents of English are spoken in South Africa. Automatic speech recognition (ASR) systems should therefore be able to process the di erent accents of South African English (SAE). In South Africa, however, system development is hampered by the limited availability of speech resources. In this thesis we consider di erent acoustic modelling approaches and system con gurations in order to determine which strategies take best advantage of a limited corpus of the ve accents of SAE for the purpose of ASR. Three acoustic modelling approaches are considered: (i) accent-speci c modelling, in which accents are modelled separately; (ii) accent-independent modelling, in which acoustic training data is pooled across accents; and (iii) multi-accent modelling, which allows selective data sharing between accents. For the latter approach, selective sharing is enabled by extending the decision-tree state clustering process normally used to construct tied-state hidden Markov models (HMMs) by allowing accent-based questions. In a rst set of experiments, we investigate phone and word recognition performance achieved by the three modelling approaches in a con guration where the accent of each test utterance is assumed to be known. Each utterance is therefore presented only to the matching model set. We show that, in terms of best recognition performance, the decision of whether to separate or to pool training data depends on the particular accents in question. Multi-accent acoustic modelling, however, allows this decision to be made automatically in a data-driven manner. When modelling the ve accents of SAE, multi-accent models yield a statistically signi cant improvement of 1.25% absolute in word recognition accuracy over accent-speci c and accentindependent models. In a second set of experiments, we consider the practical scenario where the accent of each test utterance is assumed to be unknown. Each utterance is presented simultaneously to a bank of recognisers, one for each accent, running in parallel. In this setup, accent identi cation is performed implicitly during the speech recognition process. A system employing multi-accent acoustic models in this parallel con guration is shown to achieve slightly improved performance relative to the con guration in which the accents are known. This demonstrates that accent identi cation errors made during the parallel recognition process do not a ect recognition performance. Furthermore, the parallel approach is also shown to outperform an accent-independent system obtained by pooling acoustic and language model training data. In a nal set of experiments, we consider the unsupervised reclassi cation of training set accent labels. Accent labels are assigned by human annotators based on a speaker's mother-tongue or ethnicity. These might not be optimal for modelling purposes. By classifying the accent of each utterance in the training set by using rst-pass acoustic models and then retraining the models, reclassi ed acoustic models are obtained. We show that the proposed relabelling procedure does not lead to any improvements and that training on the originally labelled data remains the best approach. / AFRIKAANSE OPSOMMING: Verskeie aksente van Engels word in Suid Afrika gepraat. Outomatiese spraakherkenningstelsels moet dus in staat wees om verskillende aksente van Suid Afrikaanse Engels (SAE) te kan hanteer. In Suid Afrika word die ontwikkeling van spraakherkenningstegnologie egter deur die beperkte beskikbaarheid van geannoteerde spraakdata belemmer. In hierdie tesis ondersoek ons verskillende akoestiese modelleringstegnieke en stelselkon gurasies ten einde te bepaal watter strategie e die beste gebruik maak van 'n databasis van die vyf aksente van SAE. Drie akoestiese modelleringstegnieke word ondersoek: (i) aksent-spesi eke modellering, waarin elke aksent apart gemodelleer word; (ii) aksent-onafhanklike modellering, waarin die akoestiese afrigdata van verskillende aksente saamgegooi word; en (iii) multi-aksent modellering, waarin data selektief tussen aksente gedeel word. Vir laasgenoemde word selektiewe deling moontlik gemaak deur die besluitnemingsboom-toestandbondeling-algoritme, wat gebruik word in die afrig van gebinde-toestand verskuilde Markov-modelle, uit te brei deur aksent-gebaseerde vrae toe te laat. In 'n eerste stel eksperimente word die foon- en woordherkenningsakkuraathede van die drie modelleringstegnieke vergelyk in 'n kon gurasie waarin daar aanvaar word dat die aksent van elke toetsspraakdeel bekend is. In hierdie kon gurasie word elke spraakdeel slegs gebied aan die modelstel wat ooreenstem met die aksent van die spraakdeel. In terme van herkenningsakkuraathede, wys ons dat die keuse tussen aksent-spesi eke en aksent-onafhanklike modellering afhanklik is van die spesi eke aksente wat ondersoek word. Multi-aksent akoestiese modellering stel ons egter in staat om hierdie besluit outomaties op 'n data-gedrewe wyse te neem. Vir die modellering van die vyf aksente van SAE lewer multi-aksent modelle 'n statisties beduidende verbetering van 1.25% absoluut in woordherkenningsakkuraatheid op in vergelyking met aksent-spesi eke en aksent-onafhanklike modelle. In 'n tweede stel eksperimente word die praktiese scenario ondersoek waar daar aanvaar word dat die aksent van elke toetsspraakdeel onbekend is. Elke spraakdeel word gelyktydig gebied aan 'n stel herkenners, een vir elke aksent, wat in parallel hardloop. In hierdie opstelling word aksentidenti kasie implisiet uitgevoer. Ons vind dat 'n stelsel wat multi-aksent akoestiese modelle in parallel inspan, e ense verbeterde werkverrigting toon in vergelyking met die opstelling waar die aksent bekend is. Dit dui daarop dat aksentidenti seringsfoute wat gemaak word gedurende herkenning, nie werkverrigting be nvloed nie. Verder wys ons dat die parallelle benadering ook beter werkverrigting toon as 'n aksent-onafhanklike stelsel wat verkry word deur akoestiese en taalmodelleringsafrigdata saam te gooi. In 'n nale stel eksperimente ondersoek ons die ongekontroleerde herklassi kasie van aksenttoekennings van die spraakdele in ons afrigstel. Aksente word gemerk deur menslike transkribeerders op grond van 'n spreker se moedertaal en ras. Hierdie toekennings is nie noodwendig optimaal vir modelleringsdoeleindes nie. Deur die aksent van elke spraakdeel in die afrigstel te klassi seer deur van aanvanklike akoestiese modelle gebruik te maak en dan weer modelle af te rig, word hergeklassi seerde akoestiese modelle verkry. Ons wys dat die voorgestelde herklassi seringsalgoritme nie tot enige verbeterings lei nie en dat dit die beste is om modelle op die oorspronklike data af te rig.

Thesis (MScEng)--Stellenbosch University, 2012 / ENGLISH ABSTRACT: This thesis presents the design and implementation of an autonomous navigational system and the automation of a practical demonstrator vehicle. It validates the proposed navigation architecture using simple functional navigational modules on the said vehicle. The proposed navigation architecture is a hierarchical structure, with a mission planner at the top, followed by the route planner, the path planner and a vehicle controller with the vehicle hardware at the base. A vehicle state estimator and mapping module runs in parallel to provide feedback data. The controls of an all terrain vehicle are electrically actuated and equipped with feedback sensors to form a complete drive-by-wire solution. A steering controller and velocity control state machine are designed and implemented on an existing on-board controller that includes a six degrees-of-freedom kinematic state estimator. A lidar scanner detects obstacles. The lidar data is mapped in real time to a local three-dimensional occupancy grid using a Bayesian update process. Each lidar beam is projected within the occupancy grid and the occupancy state of a ected cells is updated. A lidar simulation environment is created to test the mapping module before practical implementation. For planning purposes, the three-dimensional occupancy grid is converted to a two-dimensional drivability map. The path planner is an adapted rapidly exploring random tree (RRT) planner, that assumes Dubins car kinematics for the vehicle. The path planner optimises a cost function based on path length and a risk factor that is derived from the drivability map. A simple mission planner that accepts user-de ned waypoints as objectives is implemented. Practical tests veri ed the potential of the navigational structure implemented in this thesis. / AFRIKAANSE OPSOMMING: In hierdie tesis word die ontwerp en implementering van 'n outonome navigasiestelsel weergegee, asook die outomatisering van 'n praktiese demonstrasievoertuig. Dit regverdig die voorgestelde navigasie-argitektuur op die bogenoemde voertuig deur gebruik te maak van eenvoudige, funksionele navigasie-modules. Die voorgestelde navigasie-argitektuur is 'n hi erargiese struktuur, met die missie-beplanner aan die bo-punt, gevolg deur die roetebeplanner, die padbeplanner en voertuigbeheerder, met die voertuighardeware as basisvlak. 'n Voertuigtoestandsafskatter en karteringsmodule loop in parallel om terugvoer te voorsien. Die kontroles van 'n vierwiel-motor ets is elektries geaktueer en met terugvoersensors toegerus om volledig rekenaarbeheerd te wees. 'n Stuur-beheerder en 'n snelheid-toestandmasjien is ontwerp en ge mplementeer op 'n bestaande aanboordverwerker wat 'n kinematiese toestandsafskatter in ses grade van vryheid insluit. 'n Lidar-skandeerder registreer hindernisse. Die lidar-data word in re ele tyd na 'n lokale drie-dimensionele besettingsrooster geprojekteer deur middel van 'n Bayesiese opdateringsproses. Elke lidar-straal word in die besettingsrooster geprojekteer en die besettingstoestand van betrokke selle word opdateer. 'n Lidar-simulasie-omgewing is geskep om die karteringsmodule te toets voor dit ge mplementeer word. Die drie-dimensionele besettingsrooster word na 'n twee-dimensionele rybaarheidskaart verwerk vir beplanningsdoeleindes. Die padbeplanner is 'n aangepaste spoedig-ontdekkende-lukrake-boom en neem Dubinskar kinematika vir die voertuig aan. Die padbeplanner optimeer 'n koste-funksie, gebaseer op padlengte en 'n risiko-faktor, wat vanaf die rybaarheidskaart verkry word. 'n Eenvoudige missie-beplanner, wat via-punte as doelstellings neem, is ge mplementeer. Praktiese toetsritte veri eer die potensiaal van die navigasiestruktuur, soos hier beskryf.

Thesis (MScEng (Electrical and Electronic Engineering))--University of Stellenbosch, 2005. / This thesis describes techniques proposed for analysing images of borehole cores. We address two problems: firstly, the automated stratigraphic classification of cores based on texture and secondly, the location of thin chromitite layers hidden in pyroxenite cores. Texture features of different rock types are extracted using wavelets, the theory of which provides an intuitive and powerful tool for this purpose. A Bayesian classifier is trained and used to discriminate between different samples. Thin, planar chromitite layers are located using a shortest path algorithm. In order to estimate the physical orientation of any layer found, a sinusoidal curve is fitted. The proposed algorithms were implemented and tested on samples taken from photographed cores. A high success rate was obtained in rock classification, and thin planar layers were located and characterised.

Thesis (MScEng (Electrical and Electronic Engineering))--University of Stellenbosch, 2007. / This thesis is concerned with the fabrication of high temperature bi-epitaxial grain boundary Josephson junctions for use in superconducting microelectronic circuits. It aims to provide a proof-of-concept manufacturing process which can serve as a basis for future research at the University of Stellenbosch. The work in this thesis integrates ...

Thesis (MScIng (Electrical and Electronic Engineering))--University of Stellenbosch, 2006. / Despite enormous advances in digital communication, the voice is still the primary tool with which people exchange ideas. However, uncompressed digital speech tends to require prohibitively high data rates (upward of 64kbps), making it impractical for many applications. Speech coding is the process of reducing the data rate of digital voice to manageable levels. Parametric speech coders or vocoders utilise a-priori information about the mechanism by which speech is produced in order to achieve extremely efficient compression of speech signals (as low as 1 kbps). The greater part of this thesis comprises an investigation into parametric speech coding. This consisted of a review of the mathematical and heuristic tools used in parametric speech coding, as well as the implementation of an accepted standard algorithm for parametric voice coding. In order to examine avenues of improvement for the existing vocoders, we examined some of the mathematical structure underlying parametric speech coding. Following on from this, we developed a novel approach to parametric speech coding which obtained promising results under both objective and subjective evaluation. An additional contribution by this thesis was the comparative subjective evaluation of the effect of parametric speech coding on English and Xhosa speech. We investigated the performance of two different encoding algorithms on the two languages.

Thesis (MScEng (Electrical and Electronic Engineering)--Stellenbosch University, 2008. / This thesis presents the design of a control system for the autonomous precision landing of an unmanned aerial vehicle aided by an infra red camera for precision position measurements. An optimal kinematics state estimator was designed using two Kalman filters. A Monocular vision algorithm that uses markers on the runway was developed to supply accurate position measurements on the final approach of the landing. Inner-loop controllers as proposed by [14] and implemented in [5] are used to reduce the aircraft dynamics to a point mass with steerable acceleration vector. Outer-loop controllers as proposed by [13] were modified and expanded to guide the aircraft on the circuit and final approach. The hardware in the loop simulator that was designed in [6] was expanded for optical measurements and used to verify the system. An infra red camera node was designed and built to supply the optical measurements. The system was installed on a model aircraft and partially tested with practical test flights.

Thesis (MScEng (Electrical and Electronic Engineering))--Stellenbosch University, 2008. / There is currently a disparity in the amount of research done in underwater communication when compared to terrestrial communication. Therefore, it was the goal of this work to try and make an initial step towards bridging that gap. To start with, an introductory analysis was made of the ocean as a communications medium, focusing on any areas where the ocean characteristics could negatively affect communication. Furthermore, an overview was conducted of current communication schemes, to determine where ocean communication would differ from terrestrial communication, with the idea of determining the limiting parameters of such communication, specifically in terms of protocol design for swarms and sensor networks. Using this research, a n-ary tree-based routing algorithm was designed and incorporated into an overall protocol in line with current ISO convention. The strategy was simulated using the Erlang platform and it was found that underwater communication can be achieved with favourable results. It was however also found that using Erlang as a communications tool is currently not the best option and has various shortcomings, although with further work it could be more usable. The implemented strategy appears eminently feasible and should provide a basis for further research and practical implementation

Thesis (MScEng (Electrical and Electronic Engineering))--University of Stellenbosch, 2007. / SumbandilaSat is the second South African satellite, and is scheduled to be launched in April/May 2007. A software defined radio (SDR) automatic identification system (AIS) receiver is proposed as a possible experimental payload for this satellite. The AIS receiver can be used to track and store movement of ships at sea, and then forward this information to the ground station upon request. This thesis demonstrates the design of a SDR AIS receiver for Sumbandila satellite. The design of a GMSK/FMmodem as used in AIS is presented. Models are developed and simulated in Matlab. Digital signal processing algorithms developed for the AIS receiver are highlighted. Algorithms are developed to decode and translate the AIS encapsulated binary messages. The models are transferred to C++ and the AIS receiver is implemented on the SDR architecture. Finally the real time performance of the AIS receiver is presented along with some test results and performance analysis.

Thesis (MScEng (Electrical and Electronic Engineering))--University of Stellenbosch, 2009. / Nanoelectromechanical systems (NEMS) are the integration of nanotechnology and microelectromechanical systems (MEMS). The idea of this thesis is to give a basic preparation to approach NEMS, a subcategory of nanotechnology. This is performed by gaining knowledge of the two separate fields NEMS consist of, namely MEMS and nanotechnology. This document will discuss the basic theory and fabrication tools for microelectromechanical systems, as well as its limits as MEMS feature sizes decrease, reaching toward nanotechnology. It starts with an introduction that includes the history, giving a basic background. The essence is captured with theory of the equipment, the main material, some concepts of design and the creation of structures. To stimulate further interest, this thesis also illustrates practical design and fabrication methods. With the knowledge and skills that were discussed, one should be ready to get started with NEMS.

Thesis (MScEng (Electrical and Electronic Engineering))--Stellenbosch University, 2008. / This thesis relates to the analysis and design of a SQUID microscope. Superconductor theory is discussed in depth to provide a thorough understanding of Josephson junctions and of dc SQUID magnetometers. The behaviour and suitability of different types of single-layer dc SQUIDs are looked at. The quality of the superconducting material patterned onto a substrate and the Josephson junction design used affect the behaviour of a practical dc SQUID. Noise and cooling play an integral part in the design and operation of a dc SQUID. The source of noise is looked at in an effort to minimize its effect. Cryocooling is essential to real world operation so different cooling strategies and their consequences are analyzed. This thesis focuses on modeling the behaviour of the dc SQUID to creating a practical system for use inside a SQUID microscope. Operating the dc SQUID with the appropriate electronics will linearize the device, reduce the effect of noise, and create a device with wide bandwidth. Each step in creating a practical system is discussed in detail. Simulations are used to create models predicting the behaviour of the dc SQUID and the electronics. They are then used to design and create practical electronic systems. Measurements are performed on Josephson junctions and dc SQUID magnetometers using the designed electronics. The Josephson junctions behave as predicted and were successfully tested. The dc SQUIDs did not behave as predicted and were not successfully tested. The SQUIDs were damaged, either by a malfunction in the cryocooler or through age related deterioration. A full test of the flux-locked loop was not possible and the dc SQUID was not linearized.