abstract: This work examines two main areas in model-based time-varying signal processing with emphasis in speech processing applications. The first area concentrates on improving speech intelligibility and on increasing the proposed methodologies application for clinical practice in speech-language pathology. The second area concentrates on signal expansions matched to physical-based models but without requiring independent basis functions; the significance of this work is demonstrated with speech vowels. A fully automated Vowel Space Area (VSA) computation method is proposed that can be applied to any type of speech. It is shown that the VSA provides an efficient and reliable measure and is correlated to speech intelligibility. A clinical tool that incorporates the automated VSA was proposed for evaluation and treatment to be used by speech language pathologists. Two exploratory studies are performed using two databases by analyzing mean formant trajectories in healthy speech for a wide range of speakers, dialects, and coarticulation contexts. It is shown that phonemes crowded in formant space can often have distinct trajectories, possibly due to accurate perception. A theory for analyzing time-varying signals models with amplitude modulation and frequency modulation is developed. Examples are provided that demonstrate other possible signal model decompositions with independent basis functions and corresponding physical interpretations. The Hilbert transform (HT) and the use of the analytic form of a signal are motivated, and a proof is provided to show that a signal can still preserve desirable mathematical properties without the use of the HT. A visualization of the Hilbert spectrum is proposed to aid in the interpretation. A signal demodulation is proposed and used to develop a modified Empirical Mode Decomposition (EMD) algorithm. / Dissertation/Thesis / Doctoral Dissertation Electrical Engineering 2016

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Turbulent flow high complexity makes it difficult to describe complex phenomena, such as the transport of vector and scalar quantities at the lower atmosphere, making the analysis of experimental data, such as time series, largely employed. The method mostly used by the micrometeorological community to quantify such turbulent transport is associated with the determination of the statistical covariance between two variables. It is known that the determination of statistical quantities for very long temporal windows leads to a large flux uncertainty. At the same time, the theory indicates that the association between fluxes and statistical covariance is only valid for temporally stationary series. The aim of the present study is to test the hypothesis that the estimate uncertainty is directly related to the series non-stationarity. To better understand this issue, we use a methodology based on a group of parametric and nonparametric statistical tests. The tests considered here are the T-test, F-test, median test, U-test and run test. Furthermore, the test results are compared with the outputs of two signal decomposition procedures: multiresolution analysis and empirical mode decomposition. The results suggest that the flux variability over large temporal scales characterizes the existence of temporal trends and low frequency components in the time series considered, so that it is more associated with an observational limitation of the analysis than with non-stationarity, as this concept should be the property of an ensemble, rather than of a single realization. Such limitation suggests the definition of a practical single order stationarity, associated with temporal trends and low frequency components whose energy is similar or larger to that of the turbulent fluctuations. For that reason, we affirm that the interactions test is, among all considered, the best suited for analyzing atmospheric data, because it is the most sensible to the existence of temporal trends. Furthermore, such test allows obtaining a temporal scale beyond which mesoscale events become important. / A complexidade de escoamentos turbulentos causa dificuldade para a descrição de fenômenos complexos, como o transporte de grandezas vetoriais e escalares na baixa atmosfera, fazendo com que a análise de dados experimentais, principalmente séries temporais, seja amplamente utilizada. O método mais utilizado pela comunidade micrometeorológica para quantificar esse transporte pela turbulência está associado à determinação da covariância entre duas variáveis. Sabe-se que a determinação de quantidades estatísticas para janelas temporais muito longas resulta em uma grande incerteza nos valores dos fluxos obtidos através desse método. Ao mesmo tempo, a teoria indica que o procedimento de associar fluxos a covariâncias estatísticas só vale para séries temporalmente estacionárias. O objetivo deste trabalho é testar a hipótese de que a incerteza das estimativas esteja relacionada diretamente com a não-estacionariedade das séries temporais. Para entendermos melhor isso, usamos uma metodologia baseada em um conjunto de testes estatísticos paramétricos e não-paramétricos de hipótese nula. Os testes considerados são o teste-T, teste-F, teste da mediana, teste-U e o teste das interações. Os resultados dos testes são ainda comparados com os obtidos com dois métodos de decomposição de sinais: a análise de multiresolução e a Decomposição Empírica de Modos. Os resultados sugerem que a variabilidade dos fluxos nas grandes escalas temporais está associada diretamente com a presença de tendências e componentes de baixa frequência nas séries analisadas, e que este fato está mais ligado à limitação observacional em que a análise é realizada do que propriamente com a não-estacionariedade, já que esta última é uma propriedade de ensemble e não de apenas uma realização. Esta limitação sugere a definição de um conceito mais prático de estacionariedade de primeira ordem, que seja associado à presença de tendências ou componentes de baixa frequência com energias da ordem ou maiores que a energia das escalas turbulentas. Por esse motivo podemos afirmar que na análise de dados atmosféricos o teste das interações mostrou-se, entre todos os considerados, o mais sensível à presença de tendências, permitindo inclusive a obtenção de uma escala temporal na qual os eventos de meso/submesoescala ganham importância.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Turbulent fluxes may be directly determined as the statistical covariance between quantities locally observed. Besides environmental and instrumental difficulties associated with taking high frequency measurements over the ocean, there is a source of uncertainty inherent to the estimation of turbulent fluxes in the atmosphere, and it is their contamination by nonturbulent motion. This problem is directly related to the time window over which the covariances are determined and to the cospectral gap that, in theory, separates turbulent and nonturbulent events. In this work, we use a methodology based in the Empirical Mode Decomposition, which allows the precise identification of the cospectral gap for each temporal interval over which the fluxes are determined. Furthermore, this novel methodology allows filtering out oscillation modes associated with nonturbulent events, therefore allowing the use of a time window over which the large turbulent eddies are completely sampled. To test the method, data from two oceanic cruises have been used. One is from project HalocAST-2010 (over Eastern Pacific), and the other is from project Acex 2012 (over Southwestern Atlantic). The use of the new method in 4-h time series resulted in an increase of the absolute values of the fluxes of sensible heat, latent heat and momentum, with respect to those determined with the traditionally used 10-minute time series. For CO2 fluxes, it has been observed a large reduction of the average absolute fluxes, suggesting that such measurement may be largely contaminated by nonturbulent fluxes. When compared to bulk estimates, fluxes obtained by the new methodology show reduced scatter with respect to those determined from fixed 10-minute windows. The scatter reduction of the CO2 flux estimates allowed the determination of a functional relationship between piston velocity and wind speed, which is not possible to be obtained from the 10-minute estimates. / Fluxos turbulentos são determinados diretamente através da covariância estatística de medidas localmente obtidas. Além das dificuldades ambientais e instrumentais encontradas na realização de medidas de alta frequência em regiões oceânicas, existe uma fonte de incerteza inerente às estimativas de fluxos turbulentos na atmosfera que é a contaminação desses pelos movimentos de mesoescala. Esse problema está diretamente relacionado com a janela temporal em que as covariâncias são calculadas e a lacuna espectral que separa os eventos turbulentos dos não-turbulentos. Nesse trabalho, utilizamos uma metodologia baseada na Decomposição em Modos Empíricos que permite a identificação da lacuna coespectral para cada intervalo em que os fluxos são calculados. Além disso, essa nova metodologia possibilita a filtragem dos modos de oscilação associados aos eventos não-turbulentos, permitido que seja usada uma janela temporal em que os grandes turbilhões sejam suficientemente amostrados. Foram utilizadas as medidas obtidas nos cruzeiros realizados pelos projetos HalocAST-2010 (leste do Pacífico) e ACEx-2012 (Atlântico Sudoeste). O uso da nova metodologia em séries de 4 h resultou em um aumento nos valores absolutos dos fluxos médios de calor sensível, latente e momento em comparação aos tradicionalmente calculados a partir de séries de 10 min. Isso mostra que, além da remoção da contribuição dos eventos de mesoescala, uma melhor representação do transporte associado aos grandes turbilhões também foi obtida. No caso do CO2, foi observada uma grande redução no valor absoluto dos fluxos médios, sugerindo que essa medida possa estar sendo fortemente contaminada pelos eventos não-turbulentos. Quando comparados com estimativas de bulk, os fluxos obtidos pela nova metodologia apresentam menor espalhamento que os calculados a partir de janelas de 10 min. A redução no espalhamento das medidas dos fluxos de CO2, possibilitou a determinação de uma relação funcional da velocidade de transferência com a velocidade do vento, que não pôde ser observada de maneira clara a partir das medidas de 10 min.

The study concerns the analysis of vocal cycle length perturbations in normophonic and dysphonic speakers.A method for tracking cycle lengths in voiced speech is proposed. The speech cycles are detected via the saliences of the speech signal samples, defined as the length of the temporal interval over which a sample is a maximum. The tracking of the cycle lengths is based on a dynamic programming algorithm that does not request that the signal is locally periodic and the average period length known a priori.The method is validated on a corpus of synthetic stimuli. The results show a good agreement between the extracted and the synthetic reference length time series. The method is able to track accurately low-frequency modulations and ast cycle-to-cycle perturbations of up to 10% and 4% respectively over the whole range of vocal frequencies. Robustness with regard to the background noise has lso been tested. The results indicate that the tracking is reliable for signal-to-noise ratios higher than 15dB.A method for analyzing the size of the cycle length perturbations as well as their frequency is proposed. The cycle length time series is decomposed into a sum of oscillating components by empirical mode decomposition the instantaneous envelopes and frequencies of which are obtained via AM-FM decomposition. Based on their average instantaneous frequencies, the empirical modes are then assigned to four categories (declination, physiological tremor, neurological tremor as well as cycle length jitter) and added within each. The within-category size of the cycle length perturbations is estimated via the standard deviation of the empirical mode sum divided by the average cycle length. The neurological tremor modulation frequency and bandwidth are obtained via the instantaneous frequencies and amplitudes of empirical modes in the neurological tremor category and summarized via a weighted instantaneous frequency probability density, compensating for the effects of mode mixing.The method is applied to two corpora of vowels comprising 123 and 74 control and 456 and 205 Parkinson speaker recordings respectively. The results indicate that the neurological tremor modulation depth is statistically significantly higher for female Parkinson speakers than for female control speakers. Neurological tremor frequency differs statistically significantly between male and female speakers and increases statistically significantly for the pooled Parkinson speakers compared to the pooled control speakers. Finally, the average vocal frequency increases for male Parkinson speakers and decreases for female Parkinson speakers, compared to the control speakers. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

Dans cette thèse, il est question d'une étude sur les méthodes d'analyse temps fréquence, temps échelle et plus particulièrement sur la décomposition modale empirique en faisant d'abord un parcours sur les méthodes traditionnelles, de l'analyse de Fourier à la transformée en ondelettes, notamment la représentation multi-résolution. Le besoin d'une précision sur les mesures aussi bien dans l'espace temporel que dans l'espace fréquentiel a toujours été une préoccupation majeure. En fait, la transformation de Fourier ne permet pas de concilier la description fréquentielle et la localisation dans le temps. La transformée de Fourier à court terme (TFCT) et ses dérivées - notamment le spectrogramme - ont depuis longtemps été les méthodes temps-fréquence les plus utilisées dans les applications pratiques. Il faut cependant reconnaître que malgré ses nombreux aspects séduisants, ces techniques sont naturellement limitées par le fait qu'elles se sont montrées inefficaces pour l'analyse de signaux non-stationnaires. La transformée en ondelettes a connu un grand succès ces dernières décennies avec le nombre important de ses applications en traitement du signal et de l'image. Malgré son efficacité dans la représentation et la manipulation des signaux, même non-stationnaires, une connaissance a priori sur le signal à décomposer est nécessaire pour un choix d'ondelette adéquat à chaque type de signal. La décomposition modale empirique - EMD pour Empirical Mode Decomposition - est une méthode de décomposition de signaux non-stationnaires ou issus de systèmes non linéaires, en une somme de modes, chaque mode étant localisé en fréquence. Cette décomposition est associée à une transformation de Hilbert-Huang (HHT) dans le but d'extraire localement une fréquence instantanée et une amplitude instantanée. Elle s'apparente à la décomposition en ondelettes avec l'avantage supplémentaire que constitue son auto-adaptabilité. Dans la suite de ces travaux, nous avons introduit une nouvelle méthode de décomposition basée sur une décomposition spectrale d'un opérateur d'interpolation basé sur les équations aux dérivées partielles. La nouvelle méthode appelée Décomposition Spectrale Intrinsèque, - SID, pour Spectrale Intrinsic Decomposition - est auto-adaptative et est plus générale que le principe de base de la Décomposition Modale Empirique. La méthode SID permet de produire un dictionnaire de Fonction Mode Spectrale Propre, en - anglais Spectral Proper Mode Function (SPMF) - qui sont semblables à des atomes dans les représentations parcimonieuses / In this thesis, it is about a study on the time-frequency, time-scale analysis methods and more particularly on Empirical Mode Decomposition (EMD), by first a course on traditional methods from Fourier analysis to wavelets, including the multiresolution representation. The need for precision measurements both in time space and in frequency space has always been a major preoccupation. In fact, the Fourier transformation does not reconcile the frequency description and location in time. The Short-Term Fourier Transform (STFT) and its derivatives - including the spectrogram - have long been the most used in practical applications. It must be recognized that despite its many attractive aspects, these technics are naturally limited by the fact that they were ineffective for non-stationary signals analysis. The wavelet transform has been very successful in recent decades with the large number of its applications in signal and image processing. Despite its effectiveness in the representation and manipulation of signals, even non-stationary, a priori knowledge about the signal to be decomposed is necessary for an appropriate wavelet choice for each type of signal. The empirical mode decomposition (EMD) is a decomposition method of non-stationary or from non-linear systems signals, in an amount of modes, each mode being localized in frequency. This decomposition is associated with a Hilbert-Huang transformation (HHT) to locally extract instantaneous amplitude and instantaneous frequency. It is similar to the wavelet decomposition with the added benefit that constitutes its auto-adaptability. In the remainder of this work, we introduced a new decomposition method based on a spectral decomposition of an interpolation intrinsic operator. The new method called Spectral Decomposition Intrinsic (SID) is auto-adaptive and is more general than the basic principle of Empirical Mode Decomposition. The SID method can produce a dictionary of Spectral Proper Mode Functions (SPMF) that are similar to atoms in sparse representations

Conventional rotating machine vibration monitoring techniques are based on the assumption that changes in the measured structural response are caused by deterioration in the condition of the rotating machine. However, due to variations of the rotational speed, the measured signal may be non-stationary and difficult to interpret. For this reason, the order tracking technique is introduced. One of main advantages of order tracking over traditional vibration monitoring lies in its ability to clearly identify non-stationary vibration data and to a large extent exclude the influences of varying rotational speed. In recent years, different order tracking techniques have been developed. Each of these has their own pros and cons in analyzing rotating machinery vibration signals. In this research, three existing order tracking techniques are extensively investigated and combined to further explore their abilities in the context of condition monitoring. Firstly, computed order tracking is examined. This allows non-stationary effects due to the variation of rotational speed to be largely excluded. However, this technique was developed to deal with the entire raw signal and therefore looses the ability to focus on each individual order of interest. Secondly, Vold-Kalman filter order tracking is considered. It is widely reported that this technique overcomes many of the limitations of other order tracking methods and extracts order signals into the time domain. However because of the adaptive nature of the Vold-Kalman filter, the non-stationary effects due to the rotational speed will remain in the extracted order waveform, which is not ideal for conventional signal processing methods such as Fourier analysis. Yet, the strict mathematical filter (the Vold-Kalman filter is based upon two rigorous mathematical equations, namely the data equation and the structural equation, to realize the filter) gives this technique an excellent ability to focus on the orders of interest. Thirdly, the empirical mode decomposition method is studied. In the literature, this technique is claimed to be an effective diagnostic tool for various kinds of applications including diagnosis of rotating machinery faults. Its unique empirical way of extracting non-stationary and non-linear signals allows it to capture machine fault information which is intractable by other order tracking methods. But since there is no precise mathematical definition for an intrinsic mode function in empirical mode decomposition and – as far as could be ascertained – no published assessment of the relationship between an order and an intrinsic mode function, this technique has not been properly considered by analysts in terms of order tracking. As a result, its abilities have not really been explored in the context of order related vibrations in rotating machinery. In this research, the relationship between an order and an intrinsic mode function is discussed and it is treated as a special kind of order tracking method. In stead of focusing individually on each order tracking technique, the current work synthesizes different order tracking techniques. Through combination, exchange and reconciliation of ideas between these order tracking techniques, three improved order tracking techniques are developed for the purpose of enhancing order tracking analysis in condition monitoring. The techniques are Vold-Kalman filter and computed order tracking (VKC-OT), intrinsic mode function and Vold-Kalman filter order tracking (IVK-OT) and intrinsic cycle re-sampling (ICR). Indeed, these improved approaches contribute to current order tracking practice, by providing new order tracking methods with new capabilities for condition monitoring of systems which are intractable by traditional order tracking methods, or which enhances results obtained by these traditional methods. The work commences with a discussion of the inter-relationship between the order tracking methods which are considered in the thesis, and exposition of the scope of the work and an explanation of the way these independent order tracking techniques are integrated in the thesis. To demonstrate the abilities of the improved order tracking techniques, two simulation models are established. One is a simple single-degree-of-freedom (SDOF) rotor model with which VKC-OT and IVK-OT techniques are demonstrated. The other is a simplified gear mesh model through which the effectiveness of the ICR technique is proved. Finally two experimental set-ups in the Sasol Laboratory for Structural Mechanics at the University of Pretoria are used for demonstrating the improved approaches for real rotating machine signals. One test rig was established to monitor an automotive alternator driven by a variable speed motor. A stator winding inter-turn short was artificially introduced. Advantages of the VKC-OT technique are presented and features clear and clean order components under non-stationary conditions. The diagnostic ability of the IVK-OT technique of further decomposing an intrinsic mode function is also demonstrated via signals from this test rig, so that order signals and vibrations that modulate orders in IMFs can be separated and used for condition monitoring purposes. The second experimental test rig is a transmission gearbox. Artificially damaged gear teeth were introduced. The ICR technique provides a practical alternative tool for fault diagnosis. It proves to be effective in diagnosing damaged gear teeth. / Thesis (PhD)--University of Pretoria, 2011. / Mechanical and Aeronautical Engineering / unrestricted

This thesis presents a study on time series analysis, which was conducted based on the following hypothesis: time series influenced by additive noise can be decomposed into stochastic and deterministic components in which individual models permit obtaining a hybrid one that improves accuracy. This hypothesis was confirmed in two steps. In the first one, we developed a formal analysis using the Nyquist-Shannon sampling theorem, proving Intrinsic Mode Functions (IMFs) extracted from the Empirical Mode Decomposition (EMD) method can be combined, according to their frequency intensities, to form stochastic and deterministic components. Considering this proof, we designed two approaches to decompose time series, which were evaluated in synthetic and real-world scenarios. Experimental results confirmed the importance of decomposing time series and individually modeling the deterministic and stochastic components, proving the second part of our hypothesis. Furthermore, we noticed the individual analysis of both components plays an important role in detecting patterns and extracting implicit information from time series. In addition to these approaches, this thesis also presents two new measurements. The first one is used to evaluate the accuracy of time series modeling in forecasting observations. This measurement was motivated by the fact that existing measurements only consider the perfect match between expected and predicted values. This new measurement overcomes this issue by also analyzing the global time series behavior. The second measurement presented important results to assess the influence of the deterministic and stochastic components on time series observations, supporting the decomposition process. Finally, this thesis also presents a Systematic Literature Review, which collected important information on related work, and two new methods to produce surrogate data, which permit investigating the presence of linear and nonlinear Gaussian processes in time series, irrespective of the influence of nonstationary behavior / Esta tese apresenta um estudo sobre análise de séries temporais, a qual foi conduzida baseada na seguinte hipótese: séries temporais influenciadas por ruído aditivo podem ser decompostas em componentes estocásticos e determinísticos que ao serem modelados individualmente permitem obter um modelo híbrido de maior acurácia. Essa hipótese foi confirmada em duas etapas. Na primeira, desenvolveu-se uma análise formal usando o teorema de amostragem proposto por Nyquist-Shannon, provando que IMFs (Intrinsic Mode Functions) extraídas pelo método EMD (Empirical Mode Decomposition) podem ser combinadas de acordo com suas intensidades de frequência para formar os componentes estocásticos e determinísticos. Considerando essa prova, duas abordagens de decomposição de séries foram desenvolvidas e avaliadas em aplicações sintéticas e reais. Resultados experimentais confirmaram a importância de decompor séries temporais e modelar seus componentes estocásticos e determinísticos, provando a segunda parte da hipótese. Além disso, notou-se que a análise individual desses componentes possibilita detectar padrões e extrair importantes informações implícitas em séries temporais. Essa tese apresenta ainda duas novas medidas. A primeira é usada para avaliar a acurácia de modelos utilizados para predizer observações. A principal vantagem dessa medida em relação às existentes é a possibilidade de avaliar os valores individuais de predição e o comportamento global entre as observações preditas e experadas. A segunda medida permite avaliar a influência dos componentes estocásticos e determinísticos sobre as séries temporais. Finalmente, essa tese apresenta ainda resultados obtidos por meio de uma revisão sistemática da literatura, a qual coletou importantes trabalhos relacionados, e dois novos métodos para geração de dados substitutos, permitindo investigar a presença de processos Gaussianos lineares e não-lineares, independente da influência de comportamento não-estacionário

The aim of the thesis is a realization of advanced analysis of signals from gait laboratory. The introductory part deals with the gait cycle and its relation to the joints kinematic is discussed. Additionally, the work is focused on the description of the gait laboratory and the definition of the indexes in order to quantify patient´s overall gait in kinematic analysis. In the practical part, kinematic data analysis was implemented in the MATLAB environment and the results of healthy individuals and patients with cerebral palsy were compared. Kinematic analysis included peak detection in specific kinematic variables. In the last part a graphical user interface for visualization was implemented.

In this semestral thesis, realizations of chosen linear filters for baseline wander are described. These filters are then used on artificial ECG signals from CSE database with added baseline wander. These methods are compared and results are evaluated. After that, literature search of Empirical mode decomposition method is utilized. Realization of designed filters in MATLAB programming language are described, then results are evaluated with respect to filtration success.

La simulation aux grandes échelles est devenue un outil d’analyse incontournable pour l’étude des écoulements turbulents dans des géométries complexes. Cependant, à cause de l’augmentation constante des ressources de calcul, le traitement des grandes quantités de données générées par les simulations hautement résolues est devenu un véritable défi qu’il n’est plus possible de relever avec des outils traditionnels. En mécanique des fluides numérique, cette problématique émergente soulève les mêmes questions que celles communément rencontrées en informatique avec des données massives. A ce sujet, certaines méthodes ont déjà été développées telles que le partitionnement et l’ordonnancement des données ou bien encore le traitement en parallèle mais restent insuffisantes pour les simulations numériques modernes. Ainsi, l’objectif de cette thèse est de proposer de nouveaux formalismes permettant de contourner le problème de volume de données en vue des futurs calculs exaflopiques que l’informatique devrait atteindre en 2020. A cette fin, une méthode massivement parallèle de co-traitement, adaptée au formalisme non-structuré, a été développée afin d’extraire les grandes structures des écoulements turbulents. Son principe consiste à introduire une série de grilles de plus en plus grossières réduisant ainsi la quantité de données à traiter tout en gardant intactes les structures cohérentes d’intérêt. Les données sont transférées d’une grille à une autre grâce à l’utilisation de filtres et de méthodes d’interpolation d’ordre élevé. L’efficacité de cette méthodologie a pu être démontrée en appliquant des techniques de décomposition modale lors de la simulation 3D d’une pale de turbine turbulente sur une grille de plusieurs milliards d’éléments. En outre, cette capacité à pouvoir gérer plusieurs niveaux de grilles au sein d’une simulation a été utilisée par la suite pour la mise en place de calculs basés sur une stratégie multi-niveaux. L’objectif de cette méthode est d’évaluer au cours du calcul les erreurs numériques et celles liées à la modélisation en simulant simultanément la même configuration pour deux résolutions différentes. Cette estimation de l’erreur est précieuse car elle permet de générer des grilles optimisées à travers la construction d’une mesure objective de la qualité des grilles. Ainsi, cette méthodologie de multi-résolution tente de limiter le coût de calcul de la simulation en minimisant les erreurs de modélisation en sous-maille, et a été appliquée avec succès à la simulation d’un écoulement turbulent autour d’un cylindre. / Large-Eddy Simulation (LES) has become a major tool for the analysis of highly turbulent flows in complex geometries. However, due to the steadily increase of computational resources, the amount of data generated by well-resolved numerical simulations is such that it has become very challenging to manage them with traditional data processing tools. In Computational Fluid Dynamics (CFD), this emerging problematic leads to the same "Big Data" challenges as in the computer science field. Some techniques have already been developed such as data partitioning and ordering or parallel processing but still remain insufficient for modern numerical simulations. Hence, the objective of this work is to propose new processing formalisms to circumvent the data volume issue for the future 2020 exa-scale computing objectives. To this aim, a massively parallel co-processing method, suited for complex geometries, was developed in order to extract large-scale features in turbulent flows. The principle of the method is to introduce a series of coarser nested grids to reduce the amount of data while keeping the large scales of interest. Data is transferred from one grid level to another using high-order filters and accurate interpolation techniques. This method enabled to apply modal decomposition techniques to a billion-cell LES of a 3D turbulent turbine blade, thus demonstrating its effectiveness. The capability of performing calculations on several embedded grid levels was then used to devise the multi-resolution LES (MR-LES). The aim of the method is to evaluate the modeling and numerical errors during an LES by conducting the same simulation on two different mesh resolutions, simultaneously. This error estimation is highly valuable as it allows to generate optimal grids through the building of an objective grid quality measure. MR-LES intents to limit the computational cost of the simulation while minimizing the sub-grid scale modeling errors. This novel framework was applied successfully to the simulation of a turbulent flow around a 3D cylinder.