As the quest for higher data rates continues, future generations of wireless communications are expected to concur even higher frequency bands, particularly at terahertz (THz) frequencies. Even though the vast bandwidths at the THz band promise terabit-per-second (Tbps) data rates, current baseband technologies do not support such high rates. In particular, the complexities of Tbps channel code decoding and ultra-massive multiple-input multiple-output data detection are prohibitive. This work addresses the efficient data detection and channel-code decoding problem under THz-band channel conditions and Tbps baseband processing limitations. We propose ultra-massive multiple-input multiple-output THz channel models, then investigate the corresponding performance of several candidate data detection and coding schemes. We further investigate the complexity of different detectors and decoders, motivating parallelizability at both levels. We recommend which detector to combine best with which channel code decoder under specific THz channel characteristics.

The low-voltage power distribution network is considered today as a serious candidate to provide residential customers with a high-speed access to communication services such as Internet. Outdoor Power-Line Communications (PLC) systems represent an alternative to the other classical 'last-mile solutions' such as ADSL, cable modems, or wireless access systems. We developed an accurate powerline channel simulation tool based on the Multiconductor Transmission Line theory. This tool is able to predict the end-to-end channel responses on the basis of the multiconductor cable structure and the network topology. Then the issue of optimal resource allocation in a multiuser environment was addressed in the light of the Multiuser Information Theory. Simultaneously active users are in competition for the limited resources that are the power (constrained by electro-magnetic compatibility restrictions) and the bandwidth (in the range of 1 to 10 MHz for outdoor PLC). The concept of multiuser balanced capacity was introduced to characterize the optimal resource allocation providing the maximum data rates with fairness constraints among the subscribers. The optimal PLC system was shown to require the shaping of the signal spectrum in the transmitters, and successive decoding in the receiver. A generic multiple access scheme based on Filter Banks (FB) was proposed, which offers the required spectral shaping with limited degrees of freedom. Classical multiple-access techniques (TDMA, CDMA, OFDMA) can be obtained by selecting the appropriate FB. The Minimum-Mean-Square-Error Decision-Feedback Joint Detector was shown to approach the performance of the optimal successive decoding receiver. Finally, the robustness of the proposed system against channel estimation and timing synchronization errors was addressed. The problem of multiuser timing synchronization was introduced, and practical multiuser timing error detectors were proposed.

Soft iterative detection and decoding techniques have been shown to be able to achieve near-capacity performance in multiple-antenna systems. To obtain the optimal soft information by marginalization over the entire observation space is intractable; and the current literature is unable to guide us towards the best way to obtain the suboptimal soft information. In this thesis, several existing soft-input soft-output (SISO) detectors, including minimum mean-square error-successive interference cancellation (MMSE-SIC), list sphere decoding (LSD), and Fincke-Pohst maximum-a-posteriori (FPMAP), are examined. Prior research has demonstrated that LSD and FPMAP outperform soft-equalization methods (i.e., MMSE-SIC); however, it is unclear which of the two scheme is superior in terms of performance-complexity trade-off. A comparison is conducted to resolve the matter. In addition, an improved scheme is proposed to modify LSD and FPMAP, providing error performance improvement and a reduction in computational complexity simultaneously. Although list-type detectors such as LSD and FPMAP provide outstanding error performance, issues such as the optimal initial sphere radius, optimal radius update strategy, and their highly variable computational complexity are still unresolved. A new detection scheme is proposed to address the above issues with fixed detection complexity, making the scheme suitable for practical implementation. / Thesis (Master, Electrical & Computer Engineering) -- Queen's University, 2008-07-08 19:29:17.66

Joint detection is vital for characterizing human pose and serves as a foundation for a wide range of computer vision applications such as physical training, health care, entertainment. This dissertation proposed two methods to detect joints in the human body for pose analysis. The first method detects joints by combining body model and automatic feature points detection together. The human body model maps the detected extreme points to the corresponding body parts of the model and detects the position of implicit joints. The dominant joints are detected after implicit joints and extreme points are located by a shortest path based methods. The main contribution of this work is a hybrid framework to detect joints on the human body to achieve robustness to different body shapes or proportions, pose variations and occlusions. Another contribution of this work is the idea of using geodesic features of the human body to build a model for guiding the human pose detection and estimation. The second proposed method detects joints by segmenting human body into parts first and then detect joints by making the detection algorithm focusing on each limb. The advantage of applying body part segmentation first is that the body segmentation method narrows down the searching area for each joint so that the joint detection method can provide more stable and accurate results.

Cette thèse traite d'un des problèmes les plus stimulants dans le traitement de la parole concernant la prothèse auditive, où seulement un capteur est disponible avec de faibles coûts de calcul, de faible utilisation d'énergie et l'absence de bases de données. Basée sur les récents résultats dans les deux estimations statistiques paramétriques et non-paramétriques, ainsi que la représentation parcimonieuse. Cette étude propose quelques techniques non seulement pour améliorer la qualité et l'intelligibilité de la parole, mais aussi pour s'attaquer au débruitage du signal audio en général.La thèse est divisée en deux parties ; Dans la première partie, on aborde le problème d'estimation de la densité spectrale de puissance du bruit, particulièrement pour le bruit non-stationnaire. Ce problème est une des parties principales du traitement de la parole du mono-capteur. La méthode proposée prend en compte le modèle parcimonieux de la parole dans le domaine transféré. Lorsque la densité spectrale de puissance du bruit est estimée, une approche sémantique est exploitée pour tenir compte de la présence ou de l'absence de la parole dans la deuxième partie. En combinant l'estimation Bayésienne et la détection Neyman-Pearson, quelques estimateurs paramétriques sont développés et testés dans le domaine Fourier. Pour approfondir la performance et la robustesse de débruitage du signal audio, une approche semi-paramétrique est considérée. La conjointe détection et estimation peut être interprétée par Smoothed Sigmoid-Based Shrinkage (SSBS). Ainsi, la méthode Bloc-SSBS est proposée afin de prendre en compte les atomes voisinages dans le domaine temporel-fréquentiel. De plus, pour améliorer fructueusement la qualité de la parole et du signal audio, un estimateur Bayésien est aussi dérivé et combiné avec la méthode Bloc-SSBS. L'efficacité et la pertinence de la stratégie dans le domaine transformée cosinus pour les débruitages de la parole et de l'audio sont confirmées par les résultats expérimentaux. / This PhD thesis deals with one of the most challenging problem in speech enhancement for assisted listening where only one micro is available with the low computational cost, the low power usage and the lack out of the database. Based on the novel and recent results both in non-parametric and parametric statistical estimation and sparse representation, this thesis work proposes several techniques for not only improving speech quality and intelligibility and but also tackling the denoising problem of the other audio signal. In the first major part, our work addresses the problem of the noise power spectrum estimation, especially for non-stationary noise, that is the key part in the single channel speech enhancement. The proposed approach takes into account the weak-sparseness model of speech in the transformed model. Once the noise power spectrum has been estimated, a semantic road is exploited to take into consideration the presence or absence of speech in the second major part. By applying the joint of the Bayesian estimator and the Neyman-Pearson detection, some parametric estimators were developed and tested in the discrete Fourier transform domain. For further improve performance and robustness in audio denoising, a semi-parametric approach is considered. The joint detection and estimation can be interpreted by Smoothed Sigmoid-Based Shrinkage (SSBS). Thus, Block-SSBS is proposed to take into additionally account the neighborhood bins in the time-frequency domain. Moreover, in order to enhance fruitfully speech and audio, a Bayesian estimator is also derived and combined with Block-SSBS. The effectiveness and relevance of this strategy in the discrete Cosine transform for both speech and audio denoising are confirmed by experimental results.

Abstract Biology and medicine rely heavily on images to understand how the body functions, for diagnosing diseases and to test the effects of treatments. In recent decades, microscopy has experienced rapid improvements, enabling imaging of fixed and living cells at higher resolutions and frame rates, and deeper inside the biological samples. This has led to rapid growth in the image data. Automated methods are needed to quantitatively analyze these huge datasets and find statistically valid patterns. Cell segmentation and tracking is critical for automated analysis, yet it is a challenging problem due to large variations in cell shapes and appearances caused by various factors, including cell type, sample preparation and imaging setup. This thesis proposes novel methods for segmentation and tracking of cells, which rely on machine learning based approaches to improve the performance, generalization and reusability of automated methods. Cell proposals are used to efficiently exploit spatial and temporal context for resolving detection ambiguities in high-cell-density regions, caused by weak boundaries and deformable shapes of cells. This thesis presents two cell proposal methods: the first method uses multiple blob-like filter banks for detecting candidates for round cells, while the second method, Cell Proposal Network (CPN), uses convolutional neural networks to learn the cell shapes and appearances, and can propose candidates for cells in a wide variety of microscopy images. CPN first regresses cell candidate bounding boxes and their scores, then, it segments the regions inside the top ranked boxes to obtain cell candidate masks. CPN can be used as a general cell detector, as is demonstrated by training a single model to segment images from histology, fluorescence and phase-contrast microscopy. This work poses segmentation and tracking as proposal selection problems, which are solved optimally using integer linear programming or approximately using iterative shortest cost path search and non-maximum suppression. Additionally, this thesis presents a method which utilizes graph-cuts and an off-the-shelf edge detector to accurately segment highly deformable cells. The main contribution of this thesis is a cell tracking method which uses CPN to propose cell candidates, represents alternative tracking hypotheses using a graphical model, and selects the globally optimal sub-graph providing cell tracks. It achieves state-of-the-art tracking performance on multiple public benchmark datasets from both phase-contrast and fluorescence microscopy containing cells of various shapes and appearances. / Tiivistelmä Biologia ja lääketiede nojaavat vahvasti kuvatietoon solujen ja kehon toimintojen ymmärtämiseksi sairauksien diagnostiikassa ja hoitojen vaikutusten seuraamisessa. Viime vuosikymmeninä mikroskopiassa on tapahtunut nopeaa teknistä kehitystä, mikä on mahdollistanut elävien solujen kuvantamisen tarkemmin, nopeammin sekä syvemmältä automatisoidusti useasta näytteestä. Tämä taas on johtanut kuvadatan nopeaan kasvuun ja suurempaan määrään biologisia kysymyksiä, joihin voidaan vastata. Kuvadatan räjähdysmäisen kasvun vuoksi kaikkia tuloksia ei voida enää tulkita pelkästään ihmistyövoimaa käyttämällä, mikä on johtanut tarpeeseen kehittää automaattisia menetelmiä analysoimaan kvantitatiivisesti suuria datajoukkoja ja löytämään tilastollisesti kelvollisia malleja. Solujen erottaminen niiden ympäristöstä ja toisista soluista (segmentointi) ja solujen seuranta ovat kriittisiä alkuvaiheen osia onnistuneessa automaattisessa analyysissä. Automaattisten menetelmien kehittämisessä solusegmentointi on kuitenkin osoittautunut hyvin haastavaksi ongelmaksi solujen muodon ja ulkonäön suurten muutosten vuoksi solutyypistä, näytteen valmistelusta ja kuvantamisjärjestelmästä johtuen. Tämä väitöskirja esittää uusia menetelmiä solujen segmentointiin ja seurantaan keskittyen koneoppimiseen perustuviin lähestymistapoihin, jotka parantavat automaattisten menetelmien suorituskykyä ja uudelleenkäytettävyyttä. Spatiaalista ja ajallista kontekstia tehokkaasti hyödyntäviä soluehdotelmia käytetään ratkaisemaan solujen heikosti erottuvista reunoista ja joustavista muodoista johtuvaa solujen muodon monitulkintaisuutta erityisesti silloin kun tutkittava solutiheys on suuri. Tämä väitöskirja esittää kaksi menetelmää soluehdotelmille; ensimmäinen menetelmä käyttää useita läikkätyyppisiä suodatinpankkeja ilmaisemaan kandidaatteja pyöreänmuotoisille soluille, kun taas toinen menetelmä nimeltään soluehdotelmaverkko (Cell Proposal Network, CPN) käyttää konvoluutionaalisia neuroverkkoja oppiakseen tunnistamaan solut niiden muodon sekä ulkonäön perusteella erityyppisissä mikroskooppikuvissa. CPN regressoi ensin solukandidaatteja ympäröivät suorakaiteet ja niiden pistemäärän, jonka jälkeen se segmentoi alueet parhaiten sijoittuneiden suorakaiteiden joukosta tuottaen solukandidaattimaskit. CPN:ää voidaan mahdollisesti käyttää yleisenä soluilmaisimena erityyppisilla kuvantamistekniikoilla tuotetuissa kuvissa mukaan lukien histologisen valo-, fluoresenssi- ja vaihekontrastimikroskooppian. Väitöskirja esittää solujen segmentoinnin ja seurannan soluehdotelmien valintaongelmina, mitkä ratkaistaan joko optimaalisesti käyttämällä kokonaislukuoptimointia tai likimääräisesti käyttämällä iteratiivista lyhimmän kustannuspolun hakua sekä ei-maksimien vaimennusta. Tämä väitöskirja esittää myös verkon leikkaukseen (graph cut) perustuvan menetelmän, mikä hyödyntää valmiiksi saatavilla olevaa reunanilmaisinta segmentoimaan tarkasti muotoaan voimakkaasti muuttavia soluja. Väitöskirjatutkimuksen keskeinen tulos on uusi solujen seurantamenetelmä, mikä käyttää CPN:ää solukandidaattien ehdottamiseen, esittää vaihtoehtoiset seurantahypoteesit verkkomallia hyödyntämällä, ja valitsee globaalisti optimaalisen aliverkon solujen kulkemille reitille. Verrattuna useisiin muihin julkisesti saatavilla oleviin kuva-analyysiohjelmistoihin tässä väitöskirjassa kehitetyt menetelmät olivat suorituskyvyltään parhaita vaihekontrasti- ja fluoresenssimikroskopialla tuotettujen kuva-aineistojen analyyseissa, joissa solujen ulkomuoto oli hyvin vaihteleva.

As drones are becoming ubiquitous, robust detection and tracking of potentially hostile drones is becoming a necessity. Among the many approaches being investigated in this relatively new research area, one cost effective option is the use of optical cameras equipped with computer vision algorithms. With the use of deep learning, it becomes possible to achieve high accuracy by generalizing from large datasets. However, drones are small and visually similar to birds, which has proven to be a major difficulty for purely vision based systems. This thesis investigates the utility of trajectory information (velocity and acceleration) in addition to temporal appearance features for detection and tracking of drones. While both kinds of information has been used in a variation of ways, work combining the two is largely lacking. Our approach uses background subtraction to generate candidate objects that initialize an LSTM which in turn combines trajectory and appearance information over multiple frames for joint detection and tracking of drones. While our specific implementation fails to outperform a traditional object detector in the form of YOLOv8, this could change with the solution of two problems identified with our approach. First problem being how to effectively incorporate large amounts of background data into the training of our network. Second being how to avoid repeatedly proposing the same non-drone candidates, while still being able to quickly resume tracking of a lost drone. / I takt med att drönare blir allt vanligare stiger kraven på robusta system som kan upptäcka och spåra hotfulla drönare. Bland de flertal tillvägagångssätt som undersöks i detta relativt nya forskningsområde är användandet av optiska kameror utrustade med datorseende-algoritmer ett kostnadseffektivt val. Genom användningen av djupinlärning har det blivit möjligt att uppnå hög pricksäkerhet genom att generalisera utifrån stora dataset. Men, drönare är små och utseendemässigt sett lika fåglar vilket är ett svåröverkomligt problem för system som endast förlitar sig på datorseende. I detta examensarbete undersöks vilken nytta som kan fås om man även tar hänsyn till information om drönarens trajektoria i form av hastighet och acceleration. Trots att både visuellt utseende och trajektoria är välstuderat när det kommer till drönardetektering, saknas det till stor del forskning som behandlar båda tillsammans. Vi använder bakgrundssubtraktion för att generera kandidater som startpunkt för en LSTM för att sedan kombinera trajektoria med utseende för förenad detektering och spårning av drönare. Fastän vår specifika implementation inte lyckas överträffa en traditionell objektdetekterare i form av YOLOv8, skulle detta kunna ändras givet en lösning på två identifierade problem med vårt tillvägagångssätt. Det första problemet är att hitta ett effektivt sätt att inkorporera stora mängder bakgrundsdata i träningen av vårt nätverk. Det andra är att undvika att gång på gång föreslå samma kandidater och samtidigt kunna snabbt återuppta spårningen av en förlorad drönare.

In the context of broadband wireless communications using code division multiple access (CDMA), the main multiple access (MA) options include single-carrier direct sequence CDMA (SC-DS-CDMA) using time-domain direct sequence spreading [1, p. 728], multicarrier CDMA (MC-CDMA) using frequency-domain spreading [2, 3] and multicarrier DS-CDMA (MC-DS-CDMA) using time-domain direct sequence spreading of the individual sub-carrier signals [4, 5]. It was shown in [6] that MC-DS-CDMA has the highest degrees of freedom in the family of CDMA schemes that can be beneficially exploited during the system design and reconfiguration procedures. An amalgam of MC-CDMA and MC-DS-CDMA known as time and frequency domain spreading (TF-domain spreading) MC-DS-CDMA was proposed in [6]. TF-domain spreading MC-DS-CDMA has several benefits over conventional MC-DS-CDMA with regard to both capacity and performance [7]. However, in contrast to conventional MC-DS-CDMA, TF-domain spreading MC-DS-CDMA introduces MUI, which necessitates the use of joint detection at the receiver. Recently, multiple input multiple output (MIMO) or multi-antenna TF-domain spreading MC-DS-CDMA schemes have been proposed in the literature that e ciently exploit both the spatial and frequency diversity available in MIMO frequency-selective channels [8, 9]. Although an extensive amount of research has been done on single and multi-antenna TF-domain spreading MC-DS-CDMA schemes that achieve both spatial and frequency diversity in frequency-selective slow fading channels [6–9], very little research considers the time-selectivity of the wireless channels encountered. Thus, the above-mentioned schemes may not be su ciently e cient, when communicating over wireless channels exhibiting both frequency-selective and time-selective fading. There are very few MC-DS-CDMA schemes in the literature that consider the time-selectivity of the wireless channels encountered. This study considers the design of single and multi-antenna TF-domain spreading MC-DS-CDMA, for frequency-selective block-fading channels, which are capable of exploiting the full diversity available in the channel (i.e. spatial, frequency and temporal diversity), using various methods of joint detection at the receiver. It has been shown that the diversity gain in block-fading channels can be improved by coding across multiple fading blocks [10–12]. Single-antenna TF-domain spreading MC-DS-CDMA is considered for the quasi-synchronous uplink channel, and multi-antenna TF-domain spreading MC-DS-CDMA is considered for the synchronous downlink channel. Numerous simulated bit error rate (BER) performance curves, obtained using a triply selective MIMO channel platform, are presented in this study using optimal and sub-optimal joint detection algorithms at the receiver. In addition, this study investigates the impact of spatial correlation on the BER performance of the MC-DS-CDMA schemes considered. From these simulated results, one is able to conclude that TF-domain spreading MC-DS-CDMA designed for frequency-selective block-fading channels performs better than previously proposed schemes designed for frequency-selective slow fading channels, owing to the additional temporal diversity exploited under the block-fading assumption. AFRIKAANS : In die konteks van bre¨eband- draadlose kommunikasie deur die gebruik van kodeverdelingveelvuldige toegang (KVVT) behels die belangrikste veelvuldigetoegang- (VT) opsies enkel-draer direkte-sekwensie KVVT (ED-DS-KVVT), deur die gebruik van tyd-domein direkte sekwensie-verspreiding [1, p. 728], veelvuldigedraer-KVVT (VD-KVVT) deur die gebruik van frekwensiedomein-verspreiding [2, 3] en VD-DS- KVVT deur die gebruik van tyd-domein direkte sekwensie-verspreiding van die individuele sub-draerseine [4, 5]. Daar is in [6] aangetoon dat VD-DS-KVVT die hoogste vlakke van vryheid in die familie KVVT-skemas het wat voordelig benut kan word gedurende sisteemontwerp en rekonfigurasieprosedures. ’n Amalgaam van VD-KVVT en VD-DS-KVVT bekend as tyd-en-frekwensiedomeinverspreiding (TF-domeinverspreiding) VD-DS-KVVT is voorgestel in [6]. TF-domeinverspreiding VD-DS-KVVT het verskeie voordele bo konvensionele VD-DS-KVVT wat sowel kapasiteit as werkverrigting betref [7]. In teenstelling met konvensionele VD-DS-KVVT benut TF-domeinverspreiding VD-DS-KVVT multi-gebruiker-interferensie, wat die gebruik van gesamentlike opsporing by die ontvanger noodsaak. In die onlangse verlede is in die literatuur veelvuldige-inset-veelvuldige-uitset- (VIVU) of veelvuldige-antenna TF-omeinverspreiding VD-DS-KVVT-skemas voorgestel wat sowel die ruimtelike as frekwensiediversiteit wat in VIVU frekwensie-selektiewe kanale beskikbaar is, e ektief gebruik [8, 9]. Hoewel uitgebreide navorsing onderneem is oor enkel- en multi-antenna TF-domeinverspreiding VD-DS-KVVT-skemas wat sowel ruimtelike as frekwensie diversiteit in frekwensie-selektiewe stadig deinende kanale bereik [6–9], oorweeg baie min navorsing die tyd-selektiwiteit van die draadlose kanale wat betrokke is. Bogenoemde skemas mag dus nie e ektief genoeg wees nie wanneer kommunikasie plaasvind oor draadlose kanale wat sowel frekwensie-selektiewe as tyd-selektiewe wegsterwing toon. Baie min VD-DS-KVVT-skemas in die literatuur skenk aandag aan die tyd-selektiwiteit van die betrokke draadlose kanale. Die studie ondersoek die ontwerp van enkel- en multi-antenna TF-domeinverspreiding VD-DS-KVVT vir frekwensie-selektiewe blokwegsterwingkanale, wat in staat is om die volle diversiteit wat in die kanaal beskikbaar is, te benut (i.e. ruimtelike, frekwensie- en tyddiversiteit), deur die gebruik van verskeie metodes van gesamentlike opsporing by die ontvanger. Daar is aangetoon dat die diversiteitwins in blokwegsterwingkanale verbeter kan word deur kodering oor veelvuldige deinende blokke [10–12]. Enkel-antenna TF-domeinverspreiding VD-DS-KVVT word oorweeg vir die kwasi-sinchroniese opverbinding-kanaal, en multi-antenna TF-domeinverspreiding VD-DS-KVVT vir die sinchroniese afverbinding-kanaal. Talryke gesimuleerde bisfouttempo (BFT) werkverrigtingkurwes wat verkry is deur die gebruik van ’n drie-voudige selektiewe VIVU-kanaalplatform, word in hierdie studie aangebied, deur die gebruik van optimale en sub-optimale gesamentlike opsporingsalgoritmes by die ontvanger. Daarbenewens ondersoek hierdie studie die impak van ruimtelike korrelasie op die BFT-werkverrigring van die VD-DS-KVVT-skemas wat oorweeg word. Uit hierdie gesimuleerde resultate is dit moontlik om tot die gevolgtrekking te kom dat TF-domeinverspreiding VD-DS-KVVT wat ontwerp is vir frekwensie-selektiese blokwegsterwingkanale beter werkverrigting toon as vroe¨er voorgestelde skemas wat ontwerp is vir frekwensie-selektiewe stadig deinende kanale, te danke aan die ekstra tyddiversiteit wat deur die blokwegsterwing-aanname benut word. Copyright / Dissertation (MEng)--University of Pretoria, 2010. / Electrical, Electronic and Computer Engineering / unrestricted