With the development of mechanical fault diagnosis technology, complex mechanical systems do not need to be shut down periodically for the maintenance. The working condition of the mechanical systems can be monitored by analyzing the wear metal particles in the systems' lubricating oil. However, the output signals of the monitoring sensor are non-stationary. In some case the particle signals are overlapped with each other. The goal of this thesis is to find a method to decompose those overlapped particle signals, and then count the particle number in the lubricating oil. At the beginning EMD method was introduced in the experiment because of the character of the sensor signals. In this project, because EMD method is sensitive to the noise in the original signals, an improved version of EMD, EEMD method was implemented. Finally, a post processing method was used to get a better result.

With the development of mechanical fault diagnosis technology, complex mechanical systems do not need to be shut down periodically for the maintenance. The working condition of the mechanical systems can be monitored by analyzing the wear metal particles in the systems' lubricating oil. However, the output signals of the monitoring sensor are non-stationary. In some case the particle signals are overlapped with each other. The goal of this thesis is to find a method to decompose those overlapped particle signals, and then count the particle number in the lubricating oil. At the beginning EMD method was introduced in the experiment because of the character of the sensor signals. In this project, because EMD method is sensitive to the noise in the original signals, an improved version of EMD, EEMD method was implemented. Finally, a post processing method was used to get a better result.

A novel method to reduce speckle noise from a digital image is presented. Speckle noise is introduced once a coherent light source is used. In this paper, we use the Empirical Mode Decomposition(EMD) method to remove speckles caused by such kind of coherent illumination. Many filter algorithms, such as Band-pass Filter, Enhanced Frost Filter, Gamma Filter, Enhanced Lee Filter, have been extensively studied to remove the speckle. However, they cannot remove noise effectively. The EMD method is able to analysis noise efficiently. This makes it possible to accurately analyze fringes in the frequency domain and to accurately retrieve the signal.

Phase-extraction from fringe patterns is an inevitable procedure in many applications, such as interferometry,Moiré analysis, and profilometry using structured light illumination. However, speckle noises could be introduced when a coherent light source is used. In this thesis, we use the empirical mode decomposition (EMD) to perform the speckle-reduction. It is found that phases can be extracted with high accuracy once speckle-reduction is performed with the EMD.

With the development of mechanical fault diagnosis technology, complex mechanical systems do not need to be shut down periodically for the maintenance. The working condition of the mechanical systems can be monitored by analyzing the wear metal particles in the systems' lubricating oil. However, the output signals of the monitoring sensor are non-stationary. In some case the particle signals are overlapped with each other. The goal of this thesis is to find a method to decompose those overlapped particle signals, and then count the particle number in the lubricating oil. At the beginning EMD method was introduced in the experiment because of the character of the sensor signals. In this project, because EMD method is sensitive to the noise in the original signals, an improved version of EMD, EEMD method was implemented. Finally, a post processing method was used to get a better result.

With the development of mechanical fault diagnosis technology, complex mechanical systems do not need to be shut down periodically for the maintenance. The working condition of the mechanical systems can be monitored by analyzing the wear metal particles in the systems' lubricating oil. However, the output signals of the monitoring sensor are non-stationary. In some case the particle signals are overlapped with each other. The goal of this thesis is to find a method to decompose those overlapped particle signals, and then count the particle number in the lubricating oil. At the beginning EMD method was introduced in the experiment because of the character of the sensor signals. In this project, because EMD method is sensitive to the noise in the original signals, an improved version of EMD, EEMD method was implemented. Finally, a post processing method was used to get a better result.

The compilation of wind resource assessments and the implicit long-term correction ofwind measurements require comprehensive data sets. Commonly employed data sets forthis purpose are wind measurements from weather stations as well as SupervisoryControl and Data Acquisition (SCADA) data from existing wind farms. In addition,reanalysis data are a consistent data source. Reanalysis data are a combination ofmeteorological models with measurements of climatology parameters. To increase theperformance of reanalysis data the corresponding data sets are processed with mesoscalemodels. The present study determines the suitability of the readily accessible EMDConWxEurope Mesoscale Data (EMD-ConWx Data) for wind resource assessments.EMD-ConWx Data include hourly wind velocities at seven heights in the range of 10 mup to 200 m and have a spatial resolution of 3 x 3 km. EMD-ConWx Data are based onthe primary reanalysis data set ERA-Interim. The EMD-ConWx Data are compared toSonic Detecting and Ranging (Sodar) measurements at 22 sites in Germany regardingthe parameters wind speed, wind direction and wind speed frequency distribution. Inaddition, the statistical correlation (R) and linear regression (R²) are determined.It was found, that EMD-ConWx Data reveal a moderate accuracy for wind resourceassessments. The determined average wind speed bias of 1.02 m/s, the average rootmeans square error (RMSE) of 1.91 m/s, the average wind direction bias of -0.89° andthe monthly correlation indicate overall an adequate match with the Sodarmeasurements. However, these results entail considerable uncertainties and variances.To reduce these variances and the wind speed overestimation a height shift of 50 m forthe EMD-ConWx wind velocity heights is introduced. The comparison of the EMDConWxwind data at 50 m to Sodar measurements at 100 m leads to a significantreduction of the wind speed bias, but it improves neither the wind direction accuracy northe wind speed correlation. Overall, the EMD-ConWx Data are suitable for windresource assessments and the implicit long-term correction of wind data. The EMDConWxData with the height shift imply the advantage of a proper representation of thewind profile in relation to common reanalysis data, even at sites with complex terrain.

Le coeur est un muscle. Son fonctionnement mécanique est celui d'une pompe chargée de distribuer et de récupérer le sang dans les poumons et dans le système cardiovasculaire. Son fonctionnement électrique est régulé par le son noeud sinusal, un stimulateur ou régulateur électrique chargé de déclencher les battements naturels du coeur qui rythment le fonctionnement du corps. Les médecins surveillent ce fonctionnement électromécanique du coeur en enregistrant un signal électrique appelé électrocardiogramme (ECG) ou un signal sonore : le phono-cardiogramme (PCG). L'analyse et le traitement de ces deux signaux sont fondamentaux pour établir un diagnostic et aider à déceler des anomalies et des pathologies cardiaques. L’objectif de cette thèse est de développer des techniques de traitement des signaux ECG et notamment PCG afin d’aider le médecin dans son analyse de ces signaux. L’idée de fond est de mettre en point des algorithmes relativement simples et peu coûteux en temps de calcul. Le premier intérêt serait de garantir leur implantation aisée dans un système mobile de surveillance cardiaque à l’usage du médecin, voire du patient. Le deuxième intérêt réside dans la possibilité d’une analyse automatique en temps réel des signaux avec le dispositif mobile, autorisant le choix de la transmission de ces signaux pour une levée de doute. De nombreux travaux ont mené à des avancées significatives dans l’analyse des signaux ECG et la reconnaissance automatiques des pathologies cardiaques. Des bases de données de signaux réels ou synthétiques annotées permettent également d’évaluer les performances de toute nouvelle méthode. Quant aux signaux PCG, ils sont nettement moins étudiés, difficiles à analyser et à interpréter. Même si les grandes familles de méthodes (Fourier, Wigner Ville et ondelettes) ont été testées, elles n’autorisent pas une reconnaissance automatique des signatures, d’en avoir une analyse et une compréhension assez fines.La Transformée en Ondelettes (TO) sur les signaux cardiaques a montré son efficacité pour filtrer et localiser les informations utiles mais elle fait intervenir une fonction externe de traitement (ondelette mère) dont le choix dépend de la connaissance au préalable du signal à traiter. Ce n'est pas toujours adapté aux signaux cardiaques. De plus, la Transformée en ondelettes induit généralement une imprécision dans la localisation due à la fonction externe et éventuellement au sous-échantillonnage des signatures. La nature non stationnaire de l'ECG et du PCG et leur sensibilité aux bruits rendent difficile la séparation d’une transition informative d'une transition due aux bruits de mesure. Le choix de l'outil de traitement doit permettre un débruitage et une analyse de ces signaux sans délocalisation des singularités ni altération de leurs caractéristiques. En réponse à nos objectifs et considérant ces différents problèmes, nous proposons de nous appuyer principalement sur la décomposition modale empirique (EMD) ou transformée de Hilbert Huang (THH) pour développer des solutions. L’EMD est une approche non linéaire capable de décomposer le signal étudié en fonctions modales intrinsèques (IMF), oscillations du type FM-AM, donnant ainsi une représentation temps/échelle du signal. Associée à la transformée de Hilbert (TH), la THH permet de déterminer les amplitudes instantanées (AI) et les fréquences instantanées (FI) de chaque mode, menant ainsi à une représentation temps/fréquence des signaux.Sans faire intervenir une fonction externe, on peut ainsi restaurer (réduction de bruit), analyser et reconstruire le signal sans délocalisation de ses singularités. Cette approche permet de localiser les pics R de l'ECG, déterminer le rythme cardiaque et étudier la variabilité fréquentielle cardiaque (VFC), localiser et analyser les composantes des bruits B1 et B2 du PCG. / The heart is a muscle. Its mechanical operation is like a pump charged for distributing and retrieving the blood in the lungs and cardiovascular system. Its electrical operation is regulated by the sinus node, a pacemaker or electric regulator responsible for triggering the natural heart beats that punctuate the functioning of the body.Doctors monitor the electromechanical functioning of the heart by recording an electrical signal called an electrocardiogram (ECG) or an audible signal : the phonocardiogram (PCG). The analysis and processing of these two signals are essential for diagnosis, to help detect anomalies and cardiac pathologies.The objective of this thesis is to develop signal processing tools on ECG and PCG to assist cardiologist in his analysis of these signals. The basic idea is to develop algorithms of low complexity and having inexpensive computing time. The primary interest is to ensure their easy implementation in a mobile heart monitoring system for use by the doctor or the patient. The second advantage lies in the possibility of automatic real-time analysis of signals with the mobile device, allowing control of the transmission of these signals to a removal of doubt.Numerous studies have led to significant advances in the analysis of ECG signals and the automatic recognition of cardiac conditions. Databases of real or synthetic signals annotated also assess the performance of new methods. PCG signals are much less studied, difficult to analyze and to interpret. The main methods (Fourier, wavelet and Wigner Ville) were tested : they do not allow automatic recognition of signatures, and an accurate understanding of their contents.Wavelet Transform (WT) on cardiac signals showed its effectiveness to filter and locate useful information, but it involves an external processing function (mother wavelet) whose the choice depends on the prior knowledge on the signal to be processed. This is not always suitable for cardiac signals. Moreover, the wavelet transform generally induces inaccuracies in the location due to the external function and optionally due to the sub- sampling of the signatures.The non-stationary nature of the ECG and PCG and their sensitivity to noise makes it difficult to separate an informative transition of a transition due to measurement noise. The choice of treatment tool should allow denoising and analysis of these signals without alteration or the processing tool delocalization of the singularities.In response to our objectives and considering these problems, we propose to rely primarily on empirical mode decomposition (EMD) and Hilbert Huang Transform (HHT) to develop solutions. The EMD is a non linear approach decomposing the signal in intrinsic signal (IMF), oscillations of the type FM-AM, giving a time/scale signal representation. Associated with the Hilbert transform (TH), the THH determines the instantaneous amplitude (IA) and instantaneous frequency (IF) of each mode, leading to a time/frequency representation of the ECG and PCG.Without involving an external function, EMD approach can restore (noise reduction), analyze and reconstruct the signal without relocation of its singularities. This approach allows to locate R peaks of the ECG, heart rate and study the cardiac frequency variability (CFV), locate and analyze the sound components B1 and B2 of the PCG.Among the trials and developments that we made, we present in particular a new method (EDA : empirical denoising approach) inspired by the EMD approach for denoising cardiac signals. We also set out the implementation of two approaches for locating ECG signature (QRS complex, T and P waves). The first is based on the detection of local maxima : in using Modulus Maxima and Lipschitz exponent followed by a classifier. The second uses NFLS, wich an nonlinear approach for the detection and location of unique transitions in the discrete domain.

No description available.

The foreign exchange market is one of the largest and most active financial markets with enormous daily trading volumes. Exchange rates are influenced by the interactions of a large number of agents, each operating with different intentions and on different time scales. This gives rise to nonlinear and non-stationary behaviour which complicates modelling. This research proposes a neural network based model trained on data filtered with a novel Empirical Mode Decomposition (EMD) filtering method for the forecasting of exchange rates. One minor and two major exchange rates are evaluated in this study. Firstly the ideal prediction horizons for trading are calculated for each of the exchange rates. The data is filtered according to this ideal prediction horizon using the EMD-filter. This EMD-filter dynamically filters the data based on the apparent number of intrinsic modes in the signal that can contribute towards prediction over the selected horizon. The filter is employed to filter out high frequency noise and components that would not contribute to the prediction of the exchange rate at the chosen timescale. This results in a clearer signal that still includes nonlinear behaviour. An artificial neural network predictor is trained on the filtered data using different sampling rates that are compatible with the cut-off frequency. The neural network is able to capture the nonlinear relationships between historic and future filtered data with greater certainty compared to a neural network trained on unfiltered data. Results show that the neural network trained on EMD-filtered data is significantly more accurate at prediction of exchange rates compared to the benchmark models of a neural network trained on unfiltered data and a random walk model for all the exchange rates. The EMD-filtered neural network’s predicted returns for the higher sample rates show higher correlations with the actual returns, and significant profits can be made when applying a trading strategy based on the predictions. Lower sample rates that just marginally satisfy the Nyquist criterion perform comparably with the neural network trained on unfiltered data; this may indicate that some aliasing occurs for these sampling rates as the EMD low-pass filter has a gradual cut-off, leaving some high frequency noise within the signal. The proposed model of the neural network trained on EMD-filtered data was able to uncover systematic relationships between the filtered inputs and actual outputs. The model is able to deliver profitable average monthly returns for most of the tested sampling rates and forecast horizons of the different exchange rates. This provides evidence that systematic predictable behaviour is present within exchange rates, and that this systematic behaviour can be modelled if it is properly separated from high frequency noise. / MIng (Computer and Electronic Engineering), North-West University, Potchefstroom Campus, 2015