Master of Science / Department of Electrical and Computer Engineering / Alexei Nikitin / Balasubramaniam Natarajan / It has been shown that the performance of communications systems can be severely limited by non-Gaussian and impulsive interference from a variety of sources. The non-Gaussian nature of this interference provides an opportunity for its effective mitigation by nonlinear filtering. In this thesis, we describe blind adaptive analog nonlinear filters, referred to as Adaptive Nonlinear Differential Limiters (ANDLs), that are characterized by several methodological distinctions from the existing digital solutions. When ANDLs are incorporated into a communications receiver, these methodological differences can translate into significant practical advantages, improving the receiver performance in the presence of non-Gaussian interference. A Nonlinear Differential Limiter (NDL) is obtained from a linear analog filter by introducing an appropriately chosen feedback-based nonlinearity into the response of the filter, and the degree of nonlinearity is controlled by a single parameter. ANDLs are similarly controlled by a single parameter, and are suitable for improving quality of non-stationary signals under time-varying noise conditions. ANDLs are designed to be fully compatible with existing linear devices and systems (i.e., ANDLs’ behavior is linear in the absence of impulsive interference), and to be used as an enhancement, or as a simple low-cost alternative, to state-of-the-art interference mitigation methods. We provide an introduction to the NDLs and illustrate their potential use for noise mitigation in communications systems. We also develop a digital implementation of an ANDL. This allows for rapid prototyping and performance analysis of various ANDL configurations and use cases.

Submitted by Fernanda Rodrigues de Lima (fernanda.rlima@ufpe.br) on 2018-08-03T20:52:13Z No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) TESE Marcos Antonio Martins de Almeida.pdf: 11555397 bytes, checksum: db589d39915a5dda1d8b9e763a9cf4c0 (MD5) / Approved for entry into archive by Alice Araujo (alice.caraujo@ufpe.br) on 2018-08-09T20:49:00Z (GMT) No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) TESE Marcos Antonio Martins de Almeida.pdf: 11555397 bytes, checksum: db589d39915a5dda1d8b9e763a9cf4c0 (MD5) / Made available in DSpace on 2018-08-09T20:49:01Z (GMT). No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) TESE Marcos Antonio Martins de Almeida.pdf: 11555397 bytes, checksum: db589d39915a5dda1d8b9e763a9cf4c0 (MD5) Previous issue date: 2016-12-21 / Statistical analysis is a tool of wide applicability in several areas of scientific knowledge. This thesis makes use of statistical analysis in two different applications: data classification and image processing targeted at document image binarization. In the first case, this thesis presents an analysis of several aspects of the consistency of the classification of the senior researchers in computer science of the Brazilian research council, CNPq - Conselho Nacional de Desenvolvimento Científico e Tecnológico. The second application of statistical analysis developed in this thesis addresses filtering-out the back to front interference which appears whenever a document is written or typed on both sides of translucent paper. In this topic, an assessment of the most important algorithms found in the literature is made, taking into account a large quantity of parameters such as the strength of the back to front interference, the diffusion of the ink in the paper, and the texture and hue of the paper due to aging. A new binarization algorithm is proposed, which is capable of removing the back-to-front noise in a wide range of documents. Additionally, this thesis proposes a new concept of “intelligent” binarization for complex documents, which besides text encompass several graphical elements such as figures, photos, diagrams, etc. / Análise estatística é uma ferramenta de grande aplicabilidade em diversas áreas do conhecimento científico. Esta tese faz uso de análise estatística em duas aplicações distintas: classificação de dados e processamento de imagens de documentos visando a binarização. No primeiro caso, é aqui feita uma análise de diversos aspectos da consistência da classificação de pesquisadores sêniores do CNPq - Conselho Nacional de Desenvolvimento Científico e Tecnológico, na área de Ciência da Computação. A segunda aplicação de análise estatística aqui desenvolvida trata da filtragem da interferência frente-verso que surge quando um documento é escrito ou impresso em ambos os lados da folha de um papel translúcido. Neste tópico é inicialmente feita uma análise da qualidade dos mais importantes algoritmos de binarização levando em consideração parâmetros tais como a intensidade da interferência frente-verso, a difusão da tinta no papel e a textura e escurecimento do papel pelo envelhecimento. Um novo algoritmo para a binarização eficiente de documentos com interferência frente-verso é aqui apresentado, tendo se mostrado capaz de remover tal ruído em uma grande gama de documentos. Adicionalmente, é aqui proposta a binarização “inteligente” de documentos complexos que envolvem diversos elementos gráficos (figuras, diagramas, etc).

[pt] Um método satisfatório para a caracterização de problemas não determinísticos é a identificação de modelos dinâmicos representativos destes problemas. Faz-se inicialmente uma análise comparativa quanto ao domínio, equivalência e adequação de modelos de parâmetro discreto da classe ARMA, de função de Transferência (FT) e de estado, não necessariamente escalares ou invariantes. A seguir, examinam-se aspectos dos procedimentos usuais de identificação destes modelos. O problema de estimação de processos, abordado através do processo de inovações, objetiva um desenvolvimento gradual dos conceitos, no que se refere à determinação da estrutura do modelo. Seguem-se comparações entre algorítmos recursivos de estimação (Kalman e outros), abordando-se o problema da propriedade finitamente recursiva e de convergência. Em geral, as técnicas de identificação conduzem a mais de um modelo passível de ser utilizado na caracterização do processo. O problema de se escolher entre estes modelos é formulado como um problema de teste de hipóteses, ao qual se aplica a técnica de Máxima Verossimilhança, indistintamente para modelos ARMA, FT e de estado. A resolução do teste é imediata a partir do processo de inovações, tendo-se, no caso de modelos ARMA/FT, algumas alternativas bastante simplificadas. A aplicação do teste de hipóteses, no caso não-Gaussiana, é também enfocada. / [en] Non-deterministic problems can be adequately characterized by identifying dynamic models that can represent them. Early in the study, a comparative analysis of the range, equivalence and adequacy of the models is initially performed. Types of models considered in this work are ARMA, transfer function and State models of discrete parameter, not necessarilly scalar or invariant. The usual identification methods are then succinttly examined and compared. The estimation problem of stochastic processes, using the innovation processes, using the innovation process approach, is also analyzed, with a view to a gradual development of concepts as regards the determination of the model structure. Recursive estimation algorithms (Kalman and others) are then compared, and the problem of finitely recursive properties and convergence is examined. Identification thecniques usually leod to more than one model capable of characterizing the stochastic process. The problem of choosing between these models is formulated as a hypothesis testing problem, to which the Maximum Likelyhood thecnique is applied. Test resolution follows immediately from the innovation process, and can indistinctly be applied to ARMA, Transfer Function or state models. In the case of ARMA and Transfer Function models, an even more simplified result can be obtained. The application of hypothesis testing to the non-Gausian assumption is also brought to focus.

M.Ing. / Vehicle attitude is the most significant of the navigational parameters in terms of its influence on accumulated dead reckoning errors. To determine the attitude of the host vehicle body, with respect to the earth, it is necessary to keep track of the orientation of the body axes with respect to the local earth navigational frame (north, east and down). The aim of this research is to investigate the feasibility and the enhancement of low cost inertial sensors (such as gyroscopes) by the addition of magnetometer and pitch and roll angle sensors. The focus of this research is on the use of low cost inertial measurement systems to determine the attitude of a vehicle body. Strapdown system principles and the estimation theory are applied to achieve this goal. Both Euler angles and Quatemions as attitude representation are implemented and compared with one another. Work is concentrated around the mathematical models for low cost sensors and the attitude system dynamics. A sensor cluster is constructed using three gyroscopes, a magnetometer and two inclinometers. These inertial sensors were integrated using a Kalman filter. The mathematics, calculations and principles used are universal for all attitude systems. Practical data was recorded after which it was filtered to illustrate the working of the Kalman filter. The addition of a magnetometer and two inclinometers are indeed feasible for enhancing the attitude obtained from the inertial sensors. The benefit associated with the gyroscopes, when the magnetometer readings are disturbed by external magnetic anomalies, where small and of little significance. This thesis fully describes the theory and approach followed to implement the Kalman filter, making this a good example of a Kalman filter implementation, especially with the MATLAB software realisation presented in the appendix.

Graphics Processing Units (GPU) and their development tools have advanced recently, and industry has become more interested in using them. Among several development frameworks for GPU(s), OpenCL provides a programming environment to write portable code that can run in parallel. This report describes two case studies of algorithm implementations in OpenCL. The first algorithm is Median Filtering which is a widely used image processing algorithm. The other algorithm is RSA which is a popular algorithm used in encryption. The CPU and GPU implementations of these algorithms are compared in method and speed. The GPU implementations are also evaluated by efficiency, stability, scalability and portability. We find that the GPU implementations perform better overall with some exceptions. We see that a pure GPU solution is not always the best and that a hybrid solution with both CPU and GPU may be to prefer in some cases.

This thesis describes the process of conceptualizing and developing a recommendersystem for a peer-to-peer commerce application. The application in question is calledPlick and is a vintage clothes marketplace where private persons and smaller vintageretailers buy and sell secondhand clothes from each other. Recommender systems is arelatively young field of research but has become more popular in recent years withthe advent of big data applications such as Netflix and Amazon. Examples ofrecommender systems being used in e-marketplace applications are however stillsparse and the main contribution of this thesis is insight into this sub-problem inrecommender system research. The three main families of recommender algorithmsare analyzed and two of them are deemed unfitting for the e-marketplace scenario.Out of the third family, collaborative filtering, three algorithms are described,implemented and tested on a large subset of data collected in Plick that consistsmainly of clicks made by users on items in the system. By using both traditional andnovel evaluation techniques it is further shown that a user-based collaborative filteringalgorithm yields the most accurate recommendations when compared to actual userbehavior. This represents a divergence from recommender systems commonly usedin e-commerce applications. The paper concludes with a discussion on the cause andsignificance of this difference and the impact of certain data-preprocessing techniqueson the results.

Rekommendationssystem används överallt. På populära plattformar såsom Netflix och Amazon får du alltid rekommendationer på vad som är nästa lämpliga film eller inköp, baserat på din personliga profil. Detta sker genom korsreferering mellan användare och produkter för att finna sannolika mönster. Syftet med studien har varit att jämföra de två prevalenta tillvägagångssätten att skapa rekommendationer, på en annorlunda datamängd, där ”best practice” inte nödvändigtvis är tillämpbart. Som följd därav, har jämförelse gjorts på effektiviteten av Content-based Filtering kontra Collaborative Filtering, på Steams spelplattform, i syfte att etablera potential för en bättre lösning. Detta angreps genom att samla in data från Steam; Bygga en Content-based Filtering motor som baslinje för att representera Steams nuvarande rekommendationssystem, samt en motsvarande Collaborative Filtering motor, baserad på en standard-implementation, att jämföra mot. Under studiens gång visade det sig att Content-based Filtering prestanda initiellt växte linjärt medan spelarbasen på ett givet spel ökade. Collaborative Filtering däremot hade en exponentiell prestationskurva för spel med få spelare, för att sedan plana ut på en nivå som prestationsmässigt överträffade jämförelsmetoden. Den praktiska signifikansen av dessa resultat torde rättfärdiga en mer utbredd implementering av Collaborative Filtering även där man normalt avstår till förmån för Content-based Filtering då det är enklare att implementera och ger acceptabla resultat. Då våra resultat visar på såpass stor avvikelse redan vid basmodeller, är det här en attityd som mycket väl kan förändras. Collaborative Filtering har varit sparsamt använt på mer mångfacetterade datamängder, men våra resultat visar på potential att överträffa Content-based Filtering med relativt liten insats även på sådana datamängder. Detta kan gynna alla inköps- och community-kombinerade plattformar, då det finns möjlighet att övervaka användandet av inköpen i realtid, vilket möjliggör för justeringar av de faktorer som kan visa sig resultera i felrepresentation. / The use of recommender systems is everywhere. On popular platforms such as Netflix and Amazon, you are always given new recommendations on what to consume next, based on your specific profiling. This is done by cross-referencing users and products to find probable patterns. The aims of this study were to compare the two main ways of generating recommendations, in an unorthodox dataset where “best practice” might not apply. Subsequently, recommendation efficiency was compared between Content Based Filtering and Collaborative Filtering, on the gaming-platform of Steam, in order to establish if there was potential for a better solution. We approached this by gathering data from Steam, building a representational baseline Content-based Filtering recommendation-engine based on what is currently used by Steam, and a competing Collaborative Filtering engine based on a standard implementation. In the course of this study, we found that while Content-based Filtering performance initially grew linearly as the player base of a game increased, Collaborative Filtering’s performance grew exponentially from a small player base, to plateau at a performance-level exceeding the comparison. The practical consequence of these findings would be the justification to apply Collaborative Filtering even on smaller, more complex sets of data than is normally done; The justification being that Content-based Filtering is easier to implement and yields decent results. With our findings showing such a big discrepancy even at basic models, this attitude might well change. The usage of Collaborative Filtering has been used scarcely on the more multifaceted datasets, but our results show that the potential to exceed Content-based Filtering is rather easily obtainable on such sets as well. This potentially benefits all purchase/community-combined platforms, as the usage of the purchase is monitorable on-line, and allows for the adjustments of misrepresentational factors as they appear.

Harmonics in systems are conventionally suppressed using passive tuned filters, which have practical limitations in terms of the overall cost, size and performance, and these are particularly unsatisfactory when large number of harmonics are involved Active power filtering is an alternative approach in which the filter injects suitable compensation currents to cancel the harmonic currents, usually through the use of power electronic converters. This type of filter does not exhibit the drawbacks normally associated with its passive counterpart, and a large number of harmonics can be compensated by a single unit without incurring additional cost or performance degradation. This thesis investigates an active power filter configuration incorporating instantaneous reactive power theory to calculate the compensation currents. Since the original equations for determining the reference compensation currents are defined in two imaginary phases, considerable computation time is necessary to transform them from the real three-phase values. The novel approach described in the thesis minimises the required computation time by calculating the equations directly in terms of the phase values i. e. three-phase currents and voltages. Furthermore, by utilising a sufficiently fast digital signal processor ( DSP ) to perform the calculation, real-time compensation can be achieved with greater accuracy. The results obtained show that the proposed approach leads to further harmonic suppression in both the current and voltage waveforms compared to the original approach, due to considerable reduction in the computation time of the reference compensation currents.

Abstract Subsampling can be used in a radio receiver to perform signal downconversion and sample-and-hold operations in order to relieve the operation frequency and bandwidth requirements of the subsequent discrete-time circuitry. However, due to the inherent aliasing behaviour of wideband noise and interference in subsampling, and the difficulty of implementing appropriate bandpass anti-aliasing filtering at high frequencies, straightforward use of a low subsampling rate can result in significant degradation of the receiver dynamic range. The aim of this thesis is to investigate and implement methods for integrating filtering into high-frequency signal sampling and downconversion by subsampling to alleviate the requirements for additional front-end filters and to mitigate the effects of noise and out-of-band signal aliasing, thereby facilitating use in integrated high-quality radio receivers. The charge-domain sampling technique studied here allows simple integration of both continuous-and discrete-time filtering functions into high-frequency signal sampling. Gated current integration results in a lowpass sin(x)/x(sinc(x)) response capable of performing built-in anti-aliasing filtering in baseband signal sampling. Weighted integration of several successive current samples can be further used to obtain an embedded discrete-time finite-impulse-response (FIR) filtering response, which can be used for internal anti-aliasing and image-rejection filtering in the downconversion of bandpass signals by subsampling. The detailed analysis of elementary charge-domain sampling circuits presented here shows that the use of integrated FIR filtering with subsampling allows acceptable noise figures to be achieved and can provide effective internal anti-aliasing rejection. The new methods for increasing the selectivity of elementary charge-domain sampling circuits presented here enable the integration of advanced, digitally programmable FIR filtering functions into high-frequency signal sampling, thereby markedly relieving the requirements for additional anti-aliasing, image rejection and possibly even channel selection filters in a radio receiver. BiCMOS and CMOS IF sampler implementations are presented in order to demonstrate the feasibility of the charge-domain sampling technique for integrated anti-aliasing and image-rejection filtering in IF signal quadrature downconversion by subsampling. Circuit measurements show that this sampling technique for built-in filtering results in an accurate frequency response and allows the use of high subsampling ratios while still achieving a competitive dynamic range.

Development of heavy rail intra-urban public transportation systems is an economically expensive policy tool for State and Local Governments that is often justified with the promise of economic development and neighborhood revitalization around station areas. However, the literature on the effects of rail intra-urban transit stations on neighborhoods is relatively thin, particularly on the socioeconomic effects. This quasi-experimental study evaluated the effect of heavy rail intra-urban transit stations on surrounding neighborhoods, using Atlanta, Georgia and its transit authority, the Metropolitan Atlanta Rapid Transit Authority (MARTA), as a case study. Atlanta is an expansive American city, with a large public transportation system, but low population density and no large-scale policies promoting growth around MARTA rail stations. The study period, 1970 to 2014, covers the entire period of MARTA’s existence – stations opened between 1979 and 2000. Neighborhood change was operationalized with a neighborhood change index (NCI), built on the Neighborhood Life-Cycle framework, with an adaptation that incorporates both the filtering (negative NCI) and gentrification (positive NCI) models of neighborhood change. The study differentiates between an initial effect of new MARTA rail stations, and a long-term effect. Control groups were formed using one and three mile buffers, as well as a matching strategy. Difference-in-difference (DID) models find very little evidence of a positive relationship of NCI with the opening of new MARTA rail stations. The economic recovery that began in 2010 is of special interest for housing research. To address this time-period this study utilized two models, with mixed results. The DID model suggested a negative effect of stations on the NCI. To control for selection bias in the 2010 to 2014 economic time-period, this study utilized propensity score matching to balance the treatment and control group on observed characteristics. A time and tract fixed effects model using the matched treatment and control groups found a significant positive effect of stations on neighborhood change. To test the long-term effect, a time and tract fixed effects model (1970-2014) with the NCI as the dependent variable found a positive NCI effect of MARTA stations on neighborhoods. Therefore, overall, positive neighborhood change (on the NCI scale) can be attributed to MARTA transit stations. Since 2002 MARTA ridership has slightly declined; therefore, the study concludes that given stagnant ridership, lack of supporting policy, and the finding of a positive relationship between MARTA transit stations and gentrification, the stations are a positive amenity, and are a significant contributor to neighborhood change. However, neighborhoods are heterogeneous on many dimensions, and the effect of rail intra-urban transit stations on neighborhoods may depend on the tract’s location, service characteristics, accessibility, and many other unobserved characteristics. Future research will supplement this methodology with additional data and compare the effect of intra-urban transit stations on neighborhood change in other cities to better address potential neighborhood heterogeneity.