[pt] Inicialmente apresenta-se a estrutura de um sistema Multi- taxa, faz-se uma revisão teórica do mesmo, e descreve-se seu funcionamento. Posteriormente são apresentadas várias formas de implementar este sistema, como realizá-lo no domínio do tempo, parte no domínio do tempo e parte no domínio da freqüência e realizá-lo no domínio da freqüência. Nestas formas de implementação são considerados fatores como: emprego da técnica de overlap- save ou overlap-add para realização da convolução a ser implementada no sistema, e utilização de filtros do tipo IIR ou FIR para realizar o filtro de reconstrução. Nos sistemas implementados das diversas formas são analisados o atraso e a complexidade comparando-se os resultados teóricos obtidos. Nestas diferentes formas de implementação são mostrados sistemas com particularidades interessantes como: implementação da interpolação no domínio da freqüência e utilização de uma FFT inversa voltada para interpolação e decimação. Finalmente são apresentados resultados práticos dos sistemas realizados, sendo estes avaliados, comparados com os teóricos e comentados. / [en] Firstly, a multi-rate system structure is presented, a theorical review is made and its behavior is described. Afterwards, many forms of its implementation are presented: in time domain, mixed time domain and frequency domain; and in the frequency domain only. In those implementation some factors are taken into account, such as the use of the overlap-save and overlap- add techniques for the convolution, and the utilization of IIR or FIR structures as reconstruction filters. In the system implemented, the delay and complexity are analised, by the comparing to the theorical results. Some interesting results are show, such as the implementation of the frequency domain interpolation and the use of an inverse FFT dedicated to decimation and interpolation. Finally, pratical results of the implemented system are presented. These are evaluated, compared to the theorical ones and commented upon.

In the 21st century, the proliferation of steer-by-wire systems has become a central issue in the automobile industry. With such systems there is often an objective to minimise vibrations on the steering wheel to increase driver comfort. Nevertheless, steering wheel vibration is also recognised as an important medium that assists drivers in judging the vehicle's subsystems dynamics as well as to indicate important information such as the presence of danger. This has led to studies of the possible role of vibrational stimuli towards informing drivers of environment conditions such as road surface types. Numerous prior studies were done to identify how characteristics of steering wheel vibrational stimuli might influence driver road surface detection which suggested that there is no single, optimal, acceleration gain that could improve the detection of all road surface types. There is currently a lack of studies on the characteristics of transient vibrations of steering wheel as appear to be an important source of information to the driver road surface detection. Therefore, this study is design to identify the similarity characteristics of transient vibrations for answering the main research question: "What are the time-domain features of transient vibrations that can optimise driver road surface detection?" This study starts by critically reviewing the existing principles of transient vibrations detection to ensure that the identified transient vibrations from original steering wheel vibrations satisfy with the definition of transient vibrations. The study continues by performing the experimental activities to identify the optimal measurement signal for both identification process of transient vibrations and driver road surface detection without taking for granted the basic measurement of signal processing. The studies then identify the similarity of transient vibrations according to their time-domain features. The studies done by performing the high-dimensional reduction techniques associated with clustering methods. Result suggests that the time-domain features of transient vibrations that can optimise driver road surface detection were found to consist of duration (Δt), amplitude (m/s2), energy (r.m.s) and Kurtosis.

[pt] Quando há mudanças no padrão sazonal de uma série temporal, ao longo do tempo, fica caracterizada a presença de sazonalidade móvel. Existem evidências de séries macroeconômicas que apresentam um grau considerável de sazonalidade móvel. Atualmente, para a realização do ajuste sazonal, o programa utilizado pelo IBGE é o X-12-ARIMA, que implementa o método X-11 de ajuste sazonal. O X-11 é um dos métodos mais utilizados no mundo pelos órgãos oficiais de estatística, no entanto, quando existe sazonalidade móvel, ele não consegue tratá-la de forma adequada. Este trabalho propõe dois projetos de filtros de extração da componente sazonal, no domínio da frequência, que são adequados tanto para séries com sazonalidade estável quanto para aquelas que apresentam sazonalidade móvel. O primeiro projeto de filtros, intitulado de filtro sazonal-WLS, utiliza critérios baseados em mínimos quadrados. O desempenho do filtro sazonal-WLS é avaliado com base em sinais sazonais artificiais, para séries mensais e trimestrais, baseados nas características das séries macroeconômicas. Os resultados são comparados com o método X-11 e são identificadas as situações nas quais ele é superior ao X-11. Considerando que o filtro sazonal-WLS é tanto superior ao X-11 quanto maior for a razão entre a variação da sazonalidade e a intensidade da componente irregular, foi desenvolvido o projeto de um segundo filtro. Este novo filtro combina a abordagem de mínimos quadrados ponderados com as características dos filtros de Chebyshev, minimizando simultaneamente o erro na estimativa da sazonalidade e a influência da componente irregular. A ele intitulou-se filtro sazonal-WLS-Chebyshev. Os resultados do filtro sazonal-WLS-Chebyshev são comparados com o filtro sazonal-WLS onde observam-se algumas melhorias. / [en] A time series is said to have moving seasonality when there are changes in the seasonal pattern. There is evidence that macroeconomic series show moving seasonality. Currently, to perform a seasonal adjustment, IBGE uses the program X-12-ARIMA, which implements the seasonal adjustment method X-11. This method is worldwide adopted by official statistical agencies. However, when a time series shows changing seasonal patterns, the X-11 seasonal adjustment method generates unreliable estimates. This thesis proposes two designs of filters to extract seasonal components in the frequency domain, that are suitable for series with stable seasonality and for those with moving seasonality. The first filter, named WLS-seasonal filter, uses criteria based on least squares. The performance of this filter is assessed based on artificial seasonal series for monthly and quarterly data, based on the characteristics of real macroeconomic series. The results are compared with the ones of X-11 method, and the situations in which this filter is superior to X-11 are identified. Taking into account the fact that the performance of the WLS-seasonal filter improves in relation to the one of X-11 the higher the ratio between the variation of seasonality and irregular intensity, the design of a second filter was developed. This new filter combines the approach of weighted least squares with the Chebyshev filters characteristics, simultaneously minimizing the error in estimating the seasonal component and the influence of the irregular component. It was named WLS-Chebyshev-seasonal filter. The performance of this new filter is compared with the one of the WLS-seasonal filter, and some improvements are observed.

[pt] Esta dissertação propõe a Rede Generativa Adversarial por Convolução Rápida de Fourier (FCGAN). Essa abordagem inovadora utiliza convoluções no domínio da frequência para permitir que a rede opere com um campo receptivo de abrangência global. Devido aos seus campos receptivos pequenos, GANs baseadas em convoluções tradicionais enfrentam dificuldades para capturar padrões estruturais e geométricos. Nosso método utiliza Convoluções Rápidas de Fourier (FFCs), que usam Transformadas de Fourier para operar no domínio espectral, afetando globalmente os canais da imagem. Assim, a FCGAN é capaz de gerar imagens considerando informações de todas as localizações dos mapas de entrada. Essa nova característica da rede pode levar a um desempenho errático e instável. Mostramos que a utilização de normalização espectral e injeções de ruído estabilizam o treinamento adversarial. O uso de convoluções espectrais em redes convolucionais tem sido explorado para tarefas como inpainting e super-resolução de imagens. Este trabalho foca no seu potencial para geração de imagens. Nossos experimentos também sustentam a afirmação que features de Fourier são substitutos de baixo custo operacional para camadas de self-attention, permitindo que a rede aprenda informações globais desde camadas iniciais. Apresentamos resultados qualitativos e quantitativos para demonstrar que a FCGAN proposta obtém resultados comparáveis a abordagens estado-da-arte com profundidade e número de parâmetros semelhantes, alcançando um FID de 18,98 no CIFAR-10 e 38,71 no STL-10 - uma redução de 4,98 e 1,40, respectivamente. Além disso, em maiores dimensões de imagens, o uso de FFCs em vez de self-attention permite batch-sizes com até o dobro do tamanho, e iterações até 26 por cento mais rápidas. / [en] This thesis proposes the Fast Fourier Convolution Generative Adversarial Network (FCGAN). This novel approach employs convolutions in the frequency domain to enable the network to operate with a channel-wide receptive field. Due to small receptive fields, traditional convolution-based GANs struggle to capture structural and geometric patterns. Our method uses Fast Fourier Convolutions (FFCs), which use Fourier Transforms to operate in the spectral domain, affecting the feature input globally. Thus, FCGAN can generate images considering information from all feature locations. This new hallmark of the network can lead to erratic and unstable performance. We show that employing spectral normalization and noise injections stabilizes adversarial training. The use of spectral convolutions in convolutional networks has been explored for tasks such as image inpainting and super-resolution. This work focuses on its potential for image generation. Our experiments further support the claim that Fourier features are lightweight replacements for self-attention, allowing the network to learn global information from early layers. We present qualitative and quantitative results to demonstrate that the proposed FCGAN achieves results comparable to state-of-the-art approaches of similar depth and parameter count, reaching an FID of 18.98 on CIFAR-10 and 38.71 on STL-10 - a reduction of 4.98 and 1.40, respectively. Moreover, in larger image dimensions, using FFCs instead of self-attention allows for batch sizes up to twice as large and iterations up to 26 percent faster.

[pt] Algumas máquinas produzem solicitações dinâmicas que são transferidas às fundações por meio de movimentos vibratórios. É necessário analisar e estudar esses movimentos vibratórios durante o projeto da fundação para evitar danos aos equipamentos, à própria fundação e às estruturas vizinhas. Este tipo de analise também é muito importante para a segurança e saúde ocupacional de pessoas que venham a frequentar tais estruturas. Por esta razão as frequências naturais de vibração são importantes para o dimensionamento dinâmico de fundações, para se evitar efeitos de ressonância e amplificação de deslocamentos. Neste trabalho, estacas embutidas em fundações elásticas de Winkler foram estudadas. As vibrações longitudinais e transversais foram investigadas. As frequências naturais foram calculadas pelo método de Rayleigh-Ritz considerando estacas como elementos de barra embutidas em fundação elástica no caso da avaliação de vibrações longitudinais, e como vigas embutidas em base elástica no caso da avaliação das vibrações transversais. Na modelagem dos elementos de barra e viga utilizaram-se para aproximações dos deslocamentos funções convencionais enriquecidas com n funções adicionais. As constantes de mola de Winkler foram determinadas por provas de carga verticais e horizontais. Os resultados das vibrações longitudinais foram comparados com medições in situ com estacas reais. Os valores reais da vibração longitudinal das estacas foram obtidos pela aplicação da técnica Fast Fourier Transform nas acelerações resultantes de ensaios de carregamentos dinâmicos, que impõem uma vibração forçada no sistema estaca-solo. / [en] Some machines create dynamic efforts that are transferred to foundations by vibratory movements. Is necessary to analyze and study these movements during foundation design in order avoid damages in equipment, in foundation and in neighbor structures. This analysis is also important to safety and health of persons. For this reason natural vibration frequencies are important to dynamic design of foundations, to avoid resonance effects and displacements amplification. In this work, piles embedded in Winkler elastic foundations were studied. The longitudinal and transversal vibrations were investigated. The natural frequencies are calculated by Rayleigh-Ritz method modelling piles as bars embedded in elastic foundations in the case of longitudinal vibration evaluation, and beams embedded in elastic foundations in the case of transversal vibration evaluation. The bars and beams elements are modeled using conventional shape function enriched with n additional functions. The Winkler spring constants were evaluated by vertical and horizontal load tests. The longitudinal vibration results were compared with in situ measurements in real piles. The longitudinal vibration real values of piles were obtained by Fast Fourier Transform in acceleration data of dynamic load tests that impose a forced vibration in soil-pile system.

Ces travaux de recherche portent sur la problématique CEM (Compatibilité Electromagnétique) en électronique de puissance. Cette étude s'intéresse particulièrement à la modélisation HF (haute fréquence) des convertisseurs d'énergie dans le but d'étudier la propagation des perturbations conduites vers le réseau avec ou sans l’utilisation du RSIL (Réseau Stabilisateur d'Impédance de Ligne). Une première partie présente les outils de modélisation des sources de perturbation dans les convertisseurs statiques.La seconde partie est consacrée à la modélisation HF d'un système d'entraînement à vitesse variable. L'onduleur de tension triphasé est représenté par trois générateurs équivalents afin de reconstituer le découpage des tensions de sortie. Un modèle HF du câble blindé de 4 conducteurs tenant compte de la dissymétrie est ensuite proposé. Les résultats de simulation ont été validés par des relevés expérimentaux. La comparaison montre qu'il est possible de modéliser correctement le comportement HF du dispositif jusqu'à 20MHz.Une troisième partie porte sur la propagation des perturbations conduites vers le réseau sans RSIL. Il a fallu pour cela déterminer et modéliser l'impédance du réseau. L'influence du pont redresseur à diodes sur la propagation des perturbations de mode commun a ensuite été étudiée. Les résultats de simulation montrent que l'impact de ces perturbations sur la tension réseau est prépondérant lorsque le pont redresseur est à l'état bloqué. Les résultats expérimentaux ont permis de valider cette étude. Ce travail se termine par une étude préliminaire sur l'association de deux convertisseurs connectés localement à une même source d'énergie / These research works focus on EMC (Electromagnetic Compatibility) in power electronics systems. Specifically, the study focuses on HF (high frequency) modeling of the power converters in order to study the conducted emissions toward the power grids with or without the use of an LISN (Line Impedance Stabilisation Network). The first part presents the modeling tools of the electromagnetic emissions sources in power converters.The second part describes the HF modeling method of an adjustable speed drive (ASD) that will be used to estimate the conducted emissions. In a first part, a behavioural model of PWM inverter is proposed as an EMI noise source. Then, a model of the shielded 4-wire energy cable that taking into account of the dissymmetry structure is proposed and validated in the frequency domain for two different lengths. The comparison of measurement and simulation results of conducted emissions (with LISN) shows the good compromise between the simulation duration and accuracy of the results.A third part is dedicated to the propagation of conducted emissions toward the power grid without using a LISN. The first stage consists in proposing a model impedance of the network impedance. The influence of the bridge rectifier diodes on the propagation paths of common mode disturbances has been studied in the second step. The simulation results show that the impact of these disturbances on the network voltage is more important when the bridge rectifier is normally off-state. The experimental results have validated this study. This work concludes with a preliminary study on the association of two converters connected locally to a single energy source

<p>Дисертација истражује најприхватљивије методе мерења стабилности фреквенције у<br />фреквенцијском домену, односно мерење вредности величине &bdquo;фазни шум&ldquo;, приликом<br />процеса еталонирања мерне опреме. Када се говори о мерној опреми која је предмет<br />еталонирања, дисертација је, пре свега, окренута ка еталонирању врхунских<br />комерцијалних стандарда (еталона) фреквенције. Приказана истраживања су била основ<br />за покретање и реализацију истраживачко-развојног пројекта у оквиру Министарства<br />одбране, који је имао за циљ оспособљавање метролошке лабораторије Техничког<br />опитног центра Војске Србије за потпуно еталонирање мерне опреме из области<br />времена и фреквенције. Различите методе мерења подразумевају и различите мерне<br />могућности, као и врло широк спектар еталонске и мерне опреме која се користи при<br />њиховој реализацији. У дисертацији се приказују различити начини мерења фазног шума<br />и описују методе реализоване на основу доступне мерне опреме. Како се ради о<br />методама упоредне анализе референтног и мереног сигнала, посебно критична позиција<br />је одабир референце, односно референтног еталона. Истраживање је довело до<br />закључака којима је предложен метод мерења потребне и довољне мерне несигурности<br />као и начин реализације изабране методе, односно одабир еталонске мерне опреме. У<br />циљу увођења мерне методе у употребу, извршена су и приказана мерења на<br />различитим типовима и врстама мерне опреме. Валидација методе је остварена<br />билатералним поређењем са Дирекцијом за мере и драгоцене метале, Група за време,<br />фреквенцију и дистрибуцију времена. Да је могуће реализовати методу мерења фазног<br />шума приликом еталонирања најквалитетнијих осцилатора је постављена хипотеза<br />дисертације, која је истраживањем доказана и практично спроведена. Крајњи резултат<br />истраживања је довео до увођења признате методе еталонирања фазног шума, први пут</p> / <p>Disertacija istražuje najprihvatljivije metode merenja stabilnosti frekvencije u<br />frekvencijskom domenu, odnosno merenje vrednosti veličine &bdquo;fazni šum&ldquo;, prilikom<br />procesa etaloniranja merne opreme. Kada se govori o mernoj opremi koja je predmet<br />etaloniranja, disertacija je, pre svega, okrenuta ka etaloniranju vrhunskih<br />komercijalnih standarda (etalona) frekvencije. Prikazana istraživanja su bila osnov<br />za pokretanje i realizaciju istraživačko-razvojnog projekta u okviru Ministarstva<br />odbrane, koji je imao za cilj osposobljavanje metrološke laboratorije Tehničkog<br />opitnog centra Vojske Srbije za potpuno etaloniranje merne opreme iz oblasti<br />vremena i frekvencije. Različite metode merenja podrazumevaju i različite merne<br />mogućnosti, kao i vrlo širok spektar etalonske i merne opreme koja se koristi pri<br />njihovoj realizaciji. U disertaciji se prikazuju različiti načini merenja faznog šuma<br />i opisuju metode realizovane na osnovu dostupne merne opreme. Kako se radi o<br />metodama uporedne analize referentnog i merenog signala, posebno kritična pozicija<br />je odabir reference, odnosno referentnog etalona. Istraživanje je dovelo do<br />zaključaka kojima je predložen metod merenja potrebne i dovoljne merne nesigurnosti<br />kao i način realizacije izabrane metode, odnosno odabir etalonske merne opreme. U<br />cilju uvođenja merne metode u upotrebu, izvršena su i prikazana merenja na<br />različitim tipovima i vrstama merne opreme. Validacija metode je ostvarena<br />bilateralnim poređenjem sa Direkcijom za mere i dragocene metale, Grupa za vreme,<br />frekvenciju i distribuciju vremena. Da je moguće realizovati metodu merenja faznog<br />šuma prilikom etaloniranja najkvalitetnijih oscilatora je postavljena hipoteza<br />disertacije, koja je istraživanjem dokazana i praktično sprovedena. Krajnji rezultat<br />istraživanja je doveo do uvođenja priznate metode etaloniranja faznog šuma, prvi put</p> / <p>The dissertation examines the most acceptable methods for measuring the frequency stability<br />in the frequency domain, that is, measuring the value of the &quot;phase noise&quot;, during the process<br />of measuring equipment calibration. When it comes to measuring equipment as a subject of<br />calibration, the dissertation is, first of all, oriented towards the calibration of the highest<br />commercial frequency standards. The presented researches were the basis for initiation and<br />realization of the research and development project within the Ministry of Defense, which was<br />aimed at training the metrology laboratory of the Technical Test Center of the Serbian Army<br />for complete calibration of measuring equipment in the area of time and frequency. Different<br />measurement methods also involve different measurement possibilities, as well as a very wide<br />range of calibration and measuring equipment used in their realization. Different methods of<br />measuring phase noise are described in the dissertation and they describe realized methods<br />based on available measuring equipment. As for methods of comparative analysis of the<br />reference and measured signal, a particularly critical position is the selection of the reference,<br />that is, the reference oscillator. The research led to conclusions suggesting the method of<br />measuring the necessary and sufficient measurement uncertainty, as well as the method of<br />realization of the chosen method, i.e. selection of the standard measuring equipment. In order<br />to put the measurement method into practice, measurements were performed on different<br />types of measuring equipment. Validation of the method was achieved by bilateral<br />comparison with the Directorate of Measures and Precious Metals, Group for time, frequency<br />and time dissemination. The possibility to realize the method of phase noise measurement<br />during the calibration of the highest quality oscillators is a hypothesis of the dissertation that<br />has been set up, and has been proven and practically carried out by the research. The final<br />result of the research led to the introduction of a recognized calibration method of the phase<br />noise, for the first time in the Republic of Serbia.</p>

In automatic speech recognition systems (ASRs), training is a critical phase to the system?s success. Communication media, either analog (such as analog landline phones) or digital (VoIP) distort the speaker?s speech signal often in very complex ways: linear distortion occurs in all channels, either in the magnitude or phase spectrum. Non-linear but time-invariant distortion will always appear in all real systems. In digital systems we also have network effects which will produce packet losses and delays and repeated packets. Finally, one cannot really assert what path a signal will take, and so having error or distortion in between is almost a certainty. The channel introduces an acoustical mismatch between the speaker's signal and the trained data in the ASR, which results in poor recognition performance. The approach so far, has been to try to undo the havoc produced by the channels, i.e. compensate for the channel's behavior. In this thesis, we try to characterize the effects of different transmission media and use that as an inexpensive and repeatable way to train ASR systems.