Dissertations / Theses on the topic 'Decorrelating'

International Telemetering Conference Proceedings / October 18-21, 2004 / Town & Country Resort, San Diego, California
The decorrelating detector can afford good data estimates because it does not need to know many parameters of the received signal. However, it shows great performance deprivation when the background noise is high. On the other hand, partial parallel interference canceller(PPIC) has the potential to combat the near-far problem and have much lower computation complexity. But its performance depends on the initial data estimate. An improved PPIC scheme is proposed in this paper to combat the near-far problem. It utilizes the advantages of the two detectors by combining them. The focus of this paper is on the BER performance and the near-far resistance capability of the proposed scheme. Computer simulations demonstrate that the proposed detector has good BER performance and near-far resistance capability.

The study of the spatial aspects of multipath in wireless communications environments is an increasingly important addition to the study of the temporal aspects in the search for ways to increase the utilization of the available wireless channel capacity. Traditionally, multipath has been viewed as an encumbrance in wireless communications, two of the major impairments being signal fading and intersymbol interference. However, recently the potential advantages of the diversity offered by multipath rich environments in multiuser communications have been recognised. Space time coding, for example, is a recent technique which relies on a rich scattering environment to create many practically uncorrelated signal transmission channels. Most often, statistical models have been used to describe the multipath environments in such applications. This approach has met with reasonable success but is limited when the statistical nature of a field is not easily determined or is not readily described by a known distribution.¶ Our primary aim in this thesis is to probe further into the nature of multipath environments in order to gain a greater understanding of their characteristics and diversity potential. We highlight the shortcomings of beamforming in a multipath multiuser access environment. We show that the ability of a beamformer to resolve two or more signals in angle directly limits its achievable capacity.¶ We test the probity of multipath as a source of spatial diversity, the limiting case of which is co-located users. We introduce the concept of separability to define the fundamental limits of a receiver to extract the signal of a desired user from interfering users’ signals and noise. We consider the separability performances of the minimum mean square error (MMSE), decorrelating (DEC) and matched filter (MF) detectors as we bring the positions of a desired and an interfering user closer together. We show that both the MMSE and DEC detectors are able to achieve acceptable levels of separability with the users as close as λ/10.¶ In seeking a better understanding of the nature of multipath fields themselves, we take two approaches. In the first we take a path oriented approach. The effects on the variation of the field power of the relative values of parameters such as amplitude and propagation direction are considered for a two path field. The results are applied to a theoretical analysis of the behaviour of linear detectors in multipath fields. This approach is insightful for fields with small numbers of multipaths, but quickly becomes mathematically complex.¶ In a more general approach, we take a field oriented view, seeking to quantify the complexity of arbitrary fields. We find that a multipath field has an intrinsic dimensionality of (πe)R/λ≈8.54R/λ, for a field in a two dimensional circular region, increasing only linearly with the radius R of the region. This result implies that there is no such thing as an arbitrarily complicated multipath field. That is, a field generated by any number of nearfield and farfield, specular and diffuse multipath reflections is no more complicated than a field generated by a limited number of plane waves. As such, there are limits on how rich multipath can be. This result has significant implications including means: i) to determine a parsimonious parameterization for arbitrary multipath fields and ii) of synthesizing arbitrary multipath fields with arbitrarily located nearfield or farfield, spatially discrete or continuous sources. The theoretical results are corroborated by examples of multipath field analysis and synthesis.

Hoy en día, los datos de teledetección son esenciales para muchas aplicaciones dirigidas a la observación de la tierra. El potencial de los datos de teledetección en ofrecer información valiosa permite entender mejor las características de la tierra y las actividades humanas. Los desarrollos recientes en los sensores de satélites permiten cubrir amplias áreas geográficas, produciendo imágenes con resoluciones espaciales, espectrales y temporales sin precedentes. Esta cantidad de datos producidos implica una necesidad requiere técnicas de compresión eficientes para mejorar la transmisión y la capacidad de almacenamiento. La mayoría de estas técnicas se basan en las transformadas o en los métodos de predicción. Con el fin de entender la independencia no lineal y la compactación de datos para las imágenes hiperespectrales, empezaos por investigar la mejora de la transformada “Principa Component Analysis” (PCA) que proporciona una decorrelación optima para fuentes Gausianas. Analizamos la eficiencia en compresión sin perdida de “Principal Polynomial Analysis” (PPA) que generaliza PCA con la eliminación de las dependencias non lineales a través de regresión polinomial. Mostramos que las componentes principales no son capaces de predecirse con la regresión polinomial y por tanto no se mejora la independencia del PCA. Este análisis nos permite entender mejor el concepto de la predicción en el dominio de la transformada para fines de compresión. Por tanto, en lugar de utilizar transformadas sofisticadas y costosas como PCA, centramos nuestro interés en transformadas más simples como “CDiscrete Wavelet Transform”(DWT). Mientras tanto, adoptamos técnicas de predicción para explotar cualquier dependencia restante entre las componentes transformadas. Así, introducimos un nuevo esquema llamado “Regression Wavelet Analysis” (RWA) para aumentar la independencia entre los coeficientes de las imágenes hiperespectrales. El algoritmo utiliza la regresión multivariante para explotar las relaciones entre los coeficientes de las transformada DWT. El algoritmo RWA ofrece muchas ventajas, como el bajo coste computacional y la no expansión del rango dinámico. Sin embargo, la propiedad más importante es la eficiencia en compresión sin perdida. Experimentaos extensivos sobre un conjunto amplio de imanes indican que RWA supera las técnicas mas competitivas en el estado del arte com. PCA o el estándar CCSDS-123. Extendemos los beneficios de RWA para la compresión progresiva “ Lossy-to-lossless “. Mostramos que RWA puede alcanzar una relación rate-distorsión mejor que las obtenidas por otras técnicas del estado del arte como PCA. Para este fin, proponemos un esquema de pesos que captura la significancia predictiva de las componentes. Para un análisis más profundo, también analizamos el sesgo en los parámetros de regresión cuando se aplica una compresión con perdida. Mostramos que los parámetros de RWA no son sesgados cuando los modelos de regresión se aplican con los datos recuperados que carecen información. Finalmente, introducimos una versión del algoritmo RWA de muy bajo coste computacional. Con este nuevo enfoque, la predicción solo se basa en muy pocas componentes, mientras que el rendimiento se mantiene. Mientras que la complejidad de RWA se lleva a su bajo extremo, un método de selección eficiente es necesario. A diferencia de otros métodos de selección costosos, proponemos una estrategia simple pero eficiente llamada “ neighbor selection” para seleccionar modelos con pocas componentes predictivas. Sobre un amplio conjunto de imágenes hiperespectrales, estos modelos mantienen el excelente rendimiento de RWA con el modelo máximo, mientras que el coste computacional es reducido al
Today remote sensing is essential for many applications addressed to Earth Observation. The potential capability of remote sensing in providing valuable information enables a better understanding of Earth characteristics and human activities. Recent advances in satellite sensors allow recovering large areas, producing images with unprecedented spatial, spectral and temporal resolution. This amount of data implies a need for efficient compression techniques to improve the capabilities of storage and transmissions. Most of these techniques are dominated by transforms or prediction methods. This thesis aims at deeply analyzing the state-of-the-art techniques and at providing efficient solutions that improve the compression of remote sensing data. In order to understand the non-linear independence and data compaction of hyperspectral images, we investigate the improvement of Principal Component Analysis (PCA) that provides optimal independence for Gaussian sources. We analyse the lossless coding efficiency of Principal Polynomial Analysis (PPA), which generalizes PCA by removing non-linear relations among components using polynomial regression. We show that principal components are not able to predict each other through polynomial regression, resulting in no improvement of PCA at the cost of higher complexity and larger amount of side information. This analysis allows us to understand better the concept of prediction in the transform domain for compression purposes. Therefore, rather than using expensive sophisticated transforms like PCA, we focus on theoretically suboptimal but simpler transforms like Discrete Wavelet Transform (DWT). Meanwhile, we adopt predictive techniques to exploit any remaining statistical dependence. Thus, we introduce a novel scheme, called Regression Wavelet Analysis (RWA), to increase the coefficient independence in remote sensing images. The algorithm employs multivariate regression to exploit the relationships among wavelet-transformed components. The proposed RWA has many important advantages, like the low complexity and no dynamic range expansion. Nevertheless, the most important advantage consists of its performance for lossless coding. Extensive experimental results over a wide range of sensors, such as AVIRIS, IASI and Hyperion, indicate that RWA outperforms the most prominent transforms like PCA and wavelets, and also the best recent coding standard, CCSDS-123. We extend the benefits of RWA to progressive lossy-to-lossless. We show that RWA can attain a rate-distortion performance superior to those obtained with the state-of-the-art techniques. To this end, we propose a Prediction Weighting Scheme that captures the prediction significance of each transformed components. The reason of using a weighting strategy is that coefficients with similar magnitude can have extremely different impact on the reconstruction quality. For a deeper analysis, we also investigate the bias in the least squares parameters, when coding with low bitrates. We show that the RWA parameters are unbiased for lossy coding, where the regression models are used not with the original transformed components, but with the recovered ones, which lack some information due to the lossy reconstruction. We show that hyperspectral images with large size in the spectral dimension can be coded via RWA without side information and at a lower computational cost. Finally, we introduce a very low-complexity version of RWA algorithm. Here, the prediction is based on only some few components, while the performance is maintained. When the complexity of RWA is taken to an extremely low level, a careful model selection is necessary. Contrary to expensive selection procedures, we propose a simple and efficient strategy called \textit{neighbor selection} for using small regression models. On a set of well-known and representative hyperspectral images, these small models maintain the excellent coding performance of RWA, while reducing the computational cost by about 90\%.

Torque measurement of a rotating shaft is a method of monitoring machine performance. Steady transmission of mechanical power from the prime mover to the load is vital to avoid gear and bearing wear, shaft fatigue failure, bearing and coupling failure and noise. Mean and fluctuating torque are fundamental quantities of interest. The laser torquemeter is capable of providing a non-contact measurement of time-resolved torque through a driven system. The laser torquemeter depends upon laser speckle produced from coherent light illuminating a diffuse object and the speckle pattern may be used in determining the angular position of a rotating object. When the object rotates the backscattered speckle pattern, which changes continuously but repeats exactly with every revolution, is sampled by a suitably positioned photo detector. The photo detector output signal is periodic and one period is recorded in memory as a reference and the angular position of a shaft can then be determined by a comparison of this recorded reference signal with the current photo detector output signal. The speckle pattern from two axially separated points on the shaft are monitored and under low or, ideally, zero torque the photo detector outputs are recorded into the laser torquemeter electronics. The laser torquemeter then tracks the live photo detector output and determines the angle at the two points on the shaft. Relative angular displacement in the two angle outputs appears when torque is applied and the shaft twists. When the shaft is displaced, for example by vibration, the backscattered speckle pattern changes on the photo detector and the similarity between the recorded, reference signal and the live, current signal is reduced. In this thesis, the cross-correlation of the real-time photodetector output signal and the recorded reference signal as a function of shaft position is examined. The effects of various shaft motions - rotation, axial translation, pitch and yaw, and radial translation are theoretically and experimentally examined and the results can then be used in the design of an optical head for the laser torquemeter. A review of the current torquemeter technology allows for discussion of the broad spectrum of typical torquemeter operating conditions. The optical head of the laser torquemeter may vary significantly for various torque measurement scenarios. A design procedure for the optical head of the laser torquemeter is summarised. The holy grail of torquemeter manufacturers is to produce a cheap, easy to use, robust, accurate, reliable and non-contacting torquemeter. The laser torque meter has great potential to meet these requirements.

Approved for public release; distribution is unlimited
Historically, studies on decorrelation scales have been conducted in the deep ocean waters. As the Navy shifts its interest toward the less understood shallow water regions, decorrelation scales need to be computed in order to use formerly deep water models such as the Optimum Thermal Interpolation System (OTIS) for shallow water regions such as the Yellow Sea. A data set containing over 35,000 temperature profiles from 1929 to 1991 was obtained from the Naval Oceanographic Office's MOODS data set. The winter and summer seasons provide realistic results. Winter has the smallest decorrelation scales of all the seasons, approximately 15 days and 165 km. Summer shows that there are different decorrelation scales between the surface and at depth. The surface has scales of 12.3 days and 251 km while at depth the scales are approximately 16.5 days and 163 km. An observational sampling network design is suggested for future sampling of the region. Spring and fall provide mixed results which may be due to the irregularities in time and space of the data set or to the very complex forcing mechanisms found in the region. Overall, this study gives a ground work for better refinement of decorrelation scales and thus, the ability to assess the conversion of deep water models to shallow water regions

Thesis (M.S. in Meteorology and Physical Oceanography) Naval Postgraduate School, September 1994.
Thesis advisor(s): Peter Chu, Steven D. Haeger. "September 1994." Bibliography: p. 45-47. Also available online.

Thesis (Ph.D.)--City University of Hong Kong, 2004.
"Submitted to Department of Computer Science in partial fulfillment of the requirements for the degree of Doctor of Philosophy" Includes bibliographical references (leaves 145-158)

The use of three-dimensional (3D) surround sound systems has seen a rapid increase over recent years. In two-dimensional (2D) loudspeaker formats (i.e. two-channel stereophony (stereo) and 5.1 Surround), horizontal interchannel decorrelation is a well-established technique for controlling the horizontal spread of a phantom image. Use of interchannel decorrelation can also be found within established two-to-five channel upmixing methods (stereo to 5.1). More recently, proprietary algorithms have been developed that perform 2D-to-3D upmixing, which presumably make use of interchannel decorrelation as well; however, it is not currently known how interchannel decorrelation is perceived in the vertical domain. From this, it is considered that formal investigations into the perception of vertical interchannel decorrelation are necessary. Findings from such experiments may contribute to the improved control of a sound source within 3D surround systems (i.e. the vertical spread), in addition to aiding the optimisation of 2D-to-3D upmixing algorithms. The current thesis presents a series of experiments that systematically assess vertical interchannel decorrelation under various conditions. Firstly, a comparison is made between horizontal and vertical interchannel decorrelation, where it is found that vertical decorrelation is weaker than horizontal decorrelation. However, it is also seen that vertical decorrelation can generate a significant increase of vertical image spread (VIS) for some conditions. Following this, vertical decorrelation is assessed for octave-band pink noise stimuli at various azimuth angles to the listener. The results demonstrate that vertical decorrelation is dependent on both frequency and presentation angle – a general relationship between the interchannel cross-correlation (ICC) and VIS is observed for the 500 Hz octave-band and above, and strongest for the 8 kHz octave-band. Objective analysis of these stimuli signals determined that spectral changes at higher frequencies appear to be associated with VIS perception – at 0° azimuth, the 8 and 16 kHz octave-bands demonstrate potential spectral cues, at ±30°, similar cues are seen in the 4, 8 and 16 kHz bands, and from ±110°, cues are featured in the 2, 4, 8 and 16 kHz bands. In the case of the 8 kHz octave-band, it seems that vertical decorrelation causes a ‘filling in’ of vertical localisation notch cues, potentially resulting in ambiguous perception of vertical extent. In contrast, the objective analysis suggests that VIS perception of the 500 Hz and 1 kHz bands may have been related to early reflections in the listening room. From the experiments above, it is demonstrated that the perception of VIS from vertical interchannel decorrelation is frequency-dependent, with high frequencies playing a particularly important role. A following experiment explores the vertical decorrelation of high frequencies only, where it is seen that decorrelation of the 500 Hz octave-band and above produces a similar perception of VIS to broadband decorrelation, whilst improving tonal quality. The results also indicate that decorrelation of the 8 kHz octave-band and above alone can significantly increase VIS, provided the source signal has sufficient high frequency energy. The final experimental chapter of the present thesis aims to provide a controlled assessment of 2D-to-3D upmixing, taking into account the findings of the previous experiments. In general, 2D-to-3D upmixing by vertical interchannel decorrelation had little impact on listener envelopment (LEV), when compared against a level-matched 2D 5.1 reference. Furthermore, amplitude-based decorrelation appeared to be marginally more effective, and ‘high-pass decorrelation’ resulted in slightly better tonal quality for sources that featured greater low frequency energy.

Particle image velocimetry (PIV) has become one of the most important tools for experimentally investigating the physics of fluid flows. In PIV, image-processing algorithms estimate flow velocity by measuring the displacements of flow-tracer particles suspended in a fluid. The fundamental operation in PIV is the cross correlation (CC), which measures the displacement between two similar patterns. These measurements can fail under circumstances that arise due to the nature of the underlying flow field (e.g., vortices and boundary layers, where particle patterns not only translate but also rotate, stretch, and shear) or of the images (e.g., X-ray images, with comparatively low signal to noise ratios). Despite these shortcomings, fairly little attention has been paid to fundamentally improving measurements at the level of the CC. The objective of this dissertation is to demonstrate specific modifications to the correlation kernel of PIV that increase its accuracy and in certain cases extend its utility to classes of flows and image types that were previously unresolvable. First, we present a new PIV correlation algorithm called the Fourier-Mellin correlation (FMC) that reduces velocity errors by an order of magnitude in rotating flows (chapter 1). Second, we develop a model of PIV cross correlations that explains the fundamental sources of several major drivers of error in these measurements. We show how the shapes of the tracer particles and the distributions of their individual displacements affect the correlation signal to noise ratio (SNR), whose effects have previously been described only heuristically. We use this insight to create an algorithm that automatically creates a Fourier-based weighting filter, and demonstrate that our algorithm reduces bias and RMS errors in multiple types of PIV experiments (chapter 2). Finally, we apply principles from our insights to measure blood flows in the hearts of grasshoppers using X-ray PIV, and discovered flow kinematics that were unexpected according to the current prevailing understanding of the heart as a peristaltic pump that produces directional flows. Our results suggest that flow production in insect hearts may be more complex than once thought (chapter 3).
Ph. D.

Abstract Multiuser demodulation algorithms for centralized receivers of asynchronous direct-sequence (DS) spread-spectrum code-division multiple-access (CDMA) systems in frequency-selective fading channels are studied. Both DS-CDMA systems with short (one symbol interval) and long (several symbol intervals) spreading sequences are considered. Linear multiuser receivers process ideally the complete received data block. The approximation of ideal infinite memory-length (IIR) linear multiuser detectors by finite memory-length (FIR) detectors is studied. It is shown that the FIR detectors can be made near-far resistant under a given ratio between maximum and minimum received power of users by selecting an appropriate memory-length. Numerical examples demonstrate the fact that moderate memory-lengths of the FIR detectors are sufficient to achieve the performance of the ideal IIR detectors even under severe near-far conditions. Multiuser demodulation in relatively fast fading channels is analyzed. The optimal maximum likelihood sequence detection receiver and suboptimal receivers are considered. The parallel interference cancellation (PIC) receiver is demonstrated to achieve better performance in known channels than the decorrelating receiver, but it is observed to be more sensitive to channel coefficient estimation errors than the decorrelator. At high channel loads the PIC receiver suffers from bit error rate (BER) saturation, whereas the decorrelating receiver does not. Choice of channel estimation filters is shown to be crucial if low BER is required. Data-aided channel estimation is shown to be more robust than decision-directed channel estimation, which may suffer from BER saturation caused by hang-ups at high signal-to-noise ratios. Multiuser receivers for dynamic CDMA systems are studied. Algorithms for ideal linear detector computation are derived and their complexity is analyzed. The complexity of the linear detector computation is a cubic function of KL, where K and L are the number of users and multipath components, respectively. Iterative steepest descent, conjugate gradient, and preconditioned conjugate gradient algorithms are proposed to reduce the complexity. The computational requirements for one iteration are a quadratic function of KL. The iterative detectors are also shown to be applicable for parallel implementation. Simulation results demonstrate that a moderate number of iterations yields the performance of the corresponding ideal linear detectors. A quantitative analysis shows that the PIC receivers are significantly simpler to implement than the linear receivers and only moderately more complex than the conventional matched filter bank receiver.

Cette thèse comporte deux parties qui correspondent à deux domaines distincts : les opérateurs aléatoires et les opérateurs périodiques en dimension 1. Dans la première partie, nous prouvons une estimée de décorrélation pour un opérateur aléatoire avec désordre hors diagonal en dimension 1. En se servant de cette estimée, nous déduisons l'indépendance asymptotique des statistiques locales des valeurs propres près d'énergies distinctes positives dans le régime localisé. Finalement, nous donnons une démonstration alternative de l'estimée de décorrélation pour le modèle d'Anderson discret unidimensionnel. La deuxième partie de cette thèse est liée à un problème de résonances pour l'opérateur de Schrödinger discret en dimension 1 avec potentiel périodique tronqué [...]
This thesis consists of two parts : te random and periodic operators in dimension 1. In this part, we prove the decorrelation estimate for a 1D lattice Hamiltonian with off-diagonal disorder. Consequently, we deduce the asymptotic independance of the local level statistics near distinct positive energies in the localized regime. Finally, we revisit a known result on the decorrelation estimate for the 1D discret Anderson model. The second part on my thesis adresses questions on resonances for a 1D Schrödinger operators with truncated periodic potential [...]

De nombreux travaux ont montre que la distance et la direction sont programmees et controlees differemment. En telemanipulation, les operateurs effectuant des gestes a partir d'un ecran video, doivent s'adapter a des perturbations en direction et en distance generees par l'utilisation de la camera. Toutefois, les recherches experimentales ont porte essentiellement sur l'effet des perturbations en direction et ont delaisse l'analyse des distorsions en distance. Le but de ce travail est de definir les mecanismes qui permettent de s'adapter a une perturbation en distance introduite par un ecran video. Les sujets devaient effectuer des mouvements de pointage au moyen des informations fournies par un ecran. La taille de l'image variait en fonction de l'echelle utilisee (grossissement ou reduction). Les resultats obtenus montrent que l'adaptation a une distorsion en distance est bien plus rapide que celle survenant pour les perturbations directionnelles. Les corrections different selon que la perturbation porte sur la distance ou la direction. Toutefois, cette adaptation se generalise mal a d'autres situations experimentales. Il n'y a pas de reorganisation generale de l'espace de controle et le mecanisme adaptatif ne se transfere pas a d'autres echelles. L'adaptation releve plutot d'un controle realise dans l'espace de la tache, c'est a dire celui definissant la trajectoire de la main. On observe une predominance des informations visuelles sur les informations proprioceptives et la presence d'indices visuels permet d'ameliorer l'adaptation. L'ensemble de ces resultats suggere l'existence de deux mecanismes permettant de s'adapter a une perturbation en distance. Le premier mecanisme entraine une adaptation visuomotrice locale. Le second mecanisme, necessaire pour un transfert de l'adaptation, est beaucoup plus lent et implique une recoordination generale de l'espace de travail
Several studies on motor control suggest that amplitude and direction parameters involve different processes. In remote operation, operators carry out various visual-motor tasks while viewing their hands by means of video feedback. This situation usually entails a discrepancy between the information provided by visual and proprioceptive systems. Among numerous studies that analyzed the adaptation of the perceptual-motor system to visual spatial distortions, many focused on perceptual-motor adaptation to directional distortions. The present study concerns the adaptation of motor system to the distortion of distance when the production of movement is visually controlled through a video display. The subjects had to perform a pointing movement with information provided by video screen. By changing the zoom of the camera, the amplitude of the movement perceived on the screen was different from the actual movement. Results showed that the movement adapted more quickly to the distortions of distance than to directional distortions. However, adaptation did not transfer to other situations since it was specific to the target and scale. Adaptation occurred at task level, i. E. , was based on information relating to the displacement of the hand. Most of subjects tended to use mainly the information provided by the visual system. The spatial cues (e. G. Reference frame, known objects) facilitated this adaptation. In conclusion, adaptation to distance distortion depends on two processes. The first one leads to a local visual-motor adaptation. The second one, which is necessary for the transfer of adaptation, is slower and involves a general remapping of the whole grasping space

Les travaux effectués durant cette thèse ont pour but de pallier le problème des corrélations au sein des bases de données, particulièrement fréquentes dans le cadre industriel. Une modélisation explicite des corrélations par un système de sous-régressions entre covariables permet de pointer les sources des corrélations et d'isoler certaines variables redondantes. Il en découle une pré-sélection de variables sans perte significative d'information et avec un fort potentiel explicatif (la structure de sous-régression est explicite et simple). Un algorithme MCMC (Monte-Carlo Markov Chain) de recherche de structure de sous-régressions est proposé, basé sur un modèle génératif complet sur les données. Ce prétraitement ne dépend pas de la variable réponse et peut donc être utilisé de manière générale pour toute problématique de corrélations. Par la suite, un estimateur plug-in pour la régression linéaire est proposé pour ré-injecter l'information résiduelle de manière séquentielle sans souffrir des corrélations entre covariables. Enfin, le modèle génératif complet peut être utilisé pour gérer des valeurs manquantes dans les données. Cela permet l'imputation multiple des données manquantes, préalable à l'utilisation de méthodes classiques incompatibles avec la présence de valeurs manquantes. Le package R intitulé CorReg implémente les méthodes développées durant cette thèse
This thesis was motivated by correlation issues in real datasets, in particular industrial datasets. The main idea stands in explicit modeling of the correlations between covariates by a structure of sub-regressions, that simply is a system of linear regressions between the covariates. It points out redundant covariates that can be deleted in a pre-selection step to improve matrix conditioning without significant loss of information and with strong explicative potential because this pre-selection is explained by the structure of sub-regressions, itself easy to interpret. An algorithm to find the sub-regressions structure inherent to the dataset is provided, based on a full generative model and using Monte-Carlo Markov Chain (MCMC) method. This pre-treatment does not depend on a response variable and thus can be used in a more general way with any correlated datasets. In a second part, a plug-in estimator is defined to get back the redundant covariates sequentially. Then all the covariates are used but the sequential approach acts as a protection against correlations. Finally, the generative model defined here allows, as a perspective, to manage missing values both during the MCMC and then for imputation. Then we are able to use classical methods that are not compatible with missing datasets. Once again, linear regression is used to illustrate the benefits of this method but it remains a pre-treatment that can be used in other contexts, like clustering and so on. The R package CorReg implements the methods created during this thesis

In current mammography practice, radiologists typically view mammograms in a symmetric, side-by-side, configuration in the belief that abnormalities will be made salient because they break the perceived symmetry. The literature on the use of symmetry as an aid to signal detection is limited and this thesis has taken a psychophysical approach to investigate the radiologist’s task of detecting a small mass (a blob) in paired mammogram backgrounds. Initial experiments used Gaussian white noise and synthetic mammogram backgrounds to test observer performance for the radiologist’s task using symmetric (side-by-side) displays and animated (the two images of a pair alternated sequentially in the same location) displays. The use of animated displays was then tested using real mammogram backgrounds in the subsequent experiments. The results showed that side-by-side presentation of paired images does not provide any benefit for the detection of a blob, whereas, alternated presentation enabled the observer to use the correlation present between the paired images to improve detection performance. The effect of alternation was not evident when applied to the task of detecting a small mass in real mammogram pairs and subsequent investigation suggested that the loss of effect resulted from the lack of scale invariance of real images. This meant that, regardless of the level of global correlation between two images, the localised correlation, at a region size reflecting the visual angle subtended by the fovea, was much lower. Thus, decorrelation by the visual system was ineffective and performance for the detection of a blob in the paired images was also ineffective. This thesis suggests that, whilst animated displays can be a powerful tool for the identification of differences between paired images, the underpinning mechanism of decorrelation makes them unsuited for mammograms where scale invariance means that correlation at local levels is a fraction of the global correlation level.

The objective of the research is to achieve a systematic combination of acoustic echo reduction components that together achieve a robust performance of the MCAEC system as a whole. Conventional approaches to the acoustic echo reduction system typically assume that individual components would perform ideally. For example, the adaptive algorithm for AEC is often developed in the absence of strong near-end signal, the algorithm for RES is often an added module that is developed as a separate noise reduction component, and the decorrelation procedure for MCAEC is yet another add-on module that simply introduces some form of distortion to the reference signal. The main challenge is in designing a consistent criterion across all modules that can be jointly optimized to form a more consistent framework for acoustic echo reduction. The decorrelation procedure can potentially benefit from the system approach as well if it is designed by taking the near-end listener into account. The MCAEC system should be optimized not only for the echo cancellation and suppression performance, but also for the reference signal quality after the added distortion from the decorrelation procedure. Finally, a tuning strategy is presented to jointly optimize the parameters across all modules using object criteria.

The major application of active noise control (ANC) has been focused on using a single reference signal; the work on multiple reference ANC is very scarce. Here, the behavior of multiple reference ANC is analyzed in both the frequency and time domain, and the coherence functions are provided to evaluate the effectiveness of multiple reference ANC. When there are multiple noise sources, multiple reference sensors are needed to generate complete reference signals. A simplified method combines those signals from multiple reference sensors into a single reference signal. Although this method could result in satisfactory noise control effects under special circumstances, the performance is generally compromised. A widely adopted method feeds each reference signal into a different control filter. This approach suffers from the problem of ill-conditioning when the reference signals are correlated. The problem of ill-conditioning results in slow convergence rate and high sensitivity to measurement error especially when the FXLMS algorithm is applied. To handle this particular problem, the decorrelated Filtered-X LMS (DFXLMS) algorithm is developed and studied in this thesis. Both simulations and experiments have been conducted to verify the DFXLMS algorithm and other issues associated with multiple reference ANC. The results presented herein are consistent with the theoretical analysis, and favorably indicate that the DFXLMS algorithm is effective in improving the convergence speed. To take the maximum advantage of the TMS320C30 DSP board used to implement the controller, several DSP programming issues are discussed, and assembly routines are given in the appendix. Furthermore, a graphical user interface (GUI) running under Windows' environment is introduced. The main purpose of the GUI is to facilitate parameters modification, real time data monitoring and DSP process control.
Master of Science

Diskrečiosioms bangelių transformacijoms galima panaudoti daugybę „motininių“ bangelių, tokių kaip Daubechies, Morlet‘o, Le Gall‘o arba Haar‘o. Daubechies 9/7 (CDF 9/7) bangelės istoriškai yra pirmoji bangelių šeima, kuri buvo išpopuliarinta Ingridos Daubechies (1987 m. Ingrid Daubechies suformavo vieną iš pagrindinių bangelių bazių). CDF 9/7 bangelės yra ypač veiksmingos ortogonaliosios bangelės, plačiai naudojamos praktikoje (FTB pirštų atspaudų glaudinimas, vaizdų kodavimo standartas JPEG2000 ir pan.). Šiame darbe yra aptariamos ir algoritmizuojamos dvi CDF 9/7 transformacijos versijos, būtent: bazinė CDF 9/7 ir CDF 9/7 su daline blokų dekoreliacija. Darbo tikslas – atlikti palyginamąją bazinės CDF 9/7 ir CDF 9/7 su daline blokų dekoreliacija analizę vaizde sukauptos energijos „pakavimo“ spektrinėje bangelių srityje savybės kontekste. Palyginimo kriterijumi yra pasirenkamas dvimatis hiperbolinis skaitmeninių vaizdų filtras, t.y. apdorojamas vaizdas pervedamas į spektrų bangelių sritį, taikant abi CDF 9/7 transformacijos versijas. Gautieji spektrai apdorojami fiksuoto lygmens hiperboliniais filtrais. Po filtravimo įvertinama atkurtų vaizdų kokybė.
Many kernels (“mother” wavelets) can be used for the discrete wavelet transform DWT, like those of Daubechies, Morlet, discrete Le Gall transform (DLGT) or the discrete Haar transform (HT). Cohen-Daubechies-Feauveau (CDF 9/7) wavelet are the historically first family of biorthogonal wavelets, which was made popular by Ingrid Daubechies. These are not the same as the orthogonal Daubechies wavelets, and also not very similar in shape and properties. However their construction idea is the same. The JPEG 2000 compression standard uses the biorthogonal CDF 5/3 wavelet (also called the LeGall 5/3 wavelet) for lossless compression and a CDF 9/7 wavelet for lossy compression. In this paper, two distinct DWT (CDF 9/7 and CDF 9/7 with decorrelation) as well as their computational algorithms are discussed, analyzed and compared. Comparison criteria are chosen to be one – dimensional hyperbolic filters and smoothness level of the digital signal under processing.

Com o aumento das ameaças na Internet, firewalls tornaram-se mecanismos de defesa cada vez mais utilizados. No entanto, sua configuração é notadamente complexa, podendo resultar em erros. Vários estudos foram realizados com o intuito de resolver tais problemas, mas a grande maioria deles se concentrou em trabalhar diretamente no nível de configuração, o que possui limitações. O presente trabalho investiga maneiras de extrair políticas em mais alto nível a partir de regras de firewall em baixo nível, o que é mais intuitivo. A fim de extrair as políticas reais a partir de regras de firewall, o problema do descorrelacionamento é estudado e algoritmos anteriormente propostos para resolvê-lo são apresentados e discutidos. É apresentado, também, um tipo de grafo para a melhor visualização e análise de correlacionamento entre regras. Além disso, é pesquisado o agrupamento de regras descorrelacionadas, que tem o objetivo de elevar o nível das mesmas. São apresentados dois algoritmos para realizar o agrupamento, sendo um deles novo. A seguir, é proposta uma nova metodologia de extração de políticas de firewall. A primeira parte desta consiste na utilização de um novo tipo de descorrelacionamento, o descorrelacionamento hierárquico. Este é acompanhado por uma nova maneira de agrupar regras descorrelacionadas hierarquicamente, o agrupamento hierárquico. A segunda parte é uma nova modelagem de regras de firewall que fazem parte de blacklist ou whitelist, separando-as das demais regras na extração de políticas. Algumas maneiras de realizar esta separação também são discutidas. Por fim, são relatadas as conclusões e possibilidades de trabalhos futuros.
As the number of threats in the Internet grows, firewalls have become a very important defense mechanism. However, configuring a firewall is not an easy task and is prone to errors. Several investigations have been made towards solving these issue. However, most of them have focused on working directly at the configuration level and have a number of limitations. This work investigates methods to extract higher level policies from low level firewall rules. Aiming at extracting real policies from firewall rules, we analyse the firewall decorrelation problem and previously proposed algoritmhs to solve it. In addition, a new type of graph is presented aiming at better visualising and analysing rules’ correlation. We also investigate the merging of decorrelated rules, with the goal of defining more abstract rules. Two algorithms are then presented and a new methodology for the extraction of firewall policies is proposed. This methodology is twofold. The first part consists of the use a new type of decorrelation: the hierachical decorrelation, which is introduced along with a new way of hierarchically merging decorrelated rules. The second part is a new model for blacklist or whitelist firewall rules, separating them from the other rules in the policy extraction. We also present alternatives for accomplishing this separation. Finally, we conclpude and point out directions for future work.

Cette thèse étudie, théoriquement et numériquement, la décorrélation verticale d’un tourbillon initialement vertical par un écoulement externe cisaillé sinusoïdalement dans un fluide stratifié. Il a été proposé qu’un tel mécanisme devrait déclencher des instabilités de cisaillement et ainsi contribuer à la production de petites échelles en turbulence fortement stratifiée, que l’on rencontre dans l’atmosphère et les océans dans une gamme d’échelles où la force de Coriolis est négligeable.La première partie de la thèse étudie les évolutions des énergies et enstrophie totales du tourbillon calculées au moyen de Simulations Numériques Directes (DNS), en fonction des paramètres de contrôle. Cette analyse montre que la dynamique est différente de celle des écoulements libres en déclin : du fait de la présence du cisaillement ambiant, la condition de saturation entre les termes d’étirement et de dissipation dans le bilan d’enstrophie globale implique que l’enstrophie maximale du tourbillon sature proportionnellement à $Re^{2/3}$, où $Re$ est le nombre de Reynolds, au lieu de $Re$. Néanmoins, cette condition de saturation ne rend pas compte de l’effet observé de la stratification.Afin de l’expliquer, la dynamique locale du tourbillon a été étudiée au travers de deux analyses asymptotiques présentées dans une deuxième partie. Une étude pour temps courts prouve que la réponse initiale du tourbillon est non-hydrostatique quelle que soit la stratification. Une autre analyse pour grande longueur d’onde fournit les équations qui décrivent l’évolution de la vitesse angulaire du tourbillon et des déformations de son axe. L’excitation d’ondes internes au début de l’évolution est à l’origine du régime non-hydrostatique initial. Le tourbillon est principalement advecté dans la direction du cisaillement ambiant mais aussi perpendiculairement du fait de son auto-induction. Sa vitesse angulaire décroît à cause d’effets dynamique et visqueux. La décroissance dynamique est dûe à un resserrement des lignes isopycnes dans le cœur du tourbillon, qui implique une atténuation de la vorticité verticale afin que la vorticité potentielle se conserve.Dans une troisième partie, des DNS révèlent que l’instabilité de cisaillement se développe seulement lorsque la stratification est modérée et la longueur d’onde du cisaillement ambiant suffisamment petite. Les résultats numériques sont comparés aux prédictions asymptotiques. En particulier, les évolutions du cisaillement vertical de vitesse horizontale et du gradient vertical de flottabilité sont prédites de manière fine et exhaustive par l’analyse asymptotique pour grande longueur d’onde lorsque le nombre de Froude est petit. Le nombre de Richardson asymptotique admet un minimum qui n’est pratiquement jamais inférieur au seuil critique $1/4$ nécessaire au déclenchement de l’instabilité de cisaillement. La saturation du cisaillement vertical est dûe au déclin du tourbillon dans les régions où le cisaillement ambiant est maximal. Ces résultats suggèrent que l’instabilité de cisaillement est difficilement déclenchée par des processus de décorrélation dans les écoulements fortement stratifiés, contredisant ainsi les conjectures précédemment formulées dans la littérature
This thesis investigates, theoretically and numerically, the vertical decorrelation of an initially vertical vortex by an ambient sinusoidal shear flow in a stratified fluid. It has been conjectured that such process should trigger the shear instability and, as such, contribute to the generation of small scales in strongly stratified turbulence. This type of turbulence is encountered in the atmosphere and the oceans in an intermediate range of scaleswhere Coriolis effects are negligible.The first part analyses the evolutions of the total energy and enstrophy of the vortex in Direct Numerical Simulations (DNS) as functions of the control parameters. This study reveals that the dynamics differs from freely decaying flows: because of the presence of the ambient shear flow, the balance between stretching and dissipation terms in the global enstrophy budget implies that the maximum enstrophy of the vortex scales as $Re^{2/3}$, where $Re$ is the Reynolds number, instead of simply $Re$. However, such simplified balance does not account for the observed effect of the stratification.In order to overcome this difficulty, the local dynamics of the vortex has been investigated by means of two asymptotic analyses, presented in the second part. A short-time analysis first proves that the initial response of the vortex is non-hydrostatic regardless of the stratification. A long-wavelength analysis provides governing equations for the evolution of the angular velocity of the vortex and the deformations of its axis. Internal waves are excited at the start-up of the motion, explaining the initial non-hydrostatic regime. The vortex is mostly advected in the direction of the shear flow but also perpendicularly owing to the self-induced motion. Its angular velocity decays because of dynamic and viscous effects. The former effect is due to the squeezing of the isopycnals in the vortex core which implies a decrease of the vertical vorticity to conserve potential vorticity.In a third part, the DNS show that the shear instability is only triggered if the Froude number is moderate and the wavelength of the shear small enough. The numerical results are compared to the asymptotic predictions. In particular, the evolutions of the vertical shear of horizontal velocity and of the vertical buoyancy gradient for small Froude number are comprehensively and finely captured by the long-wavelength asymptotic predictions. The minimum value of the asymptotic Richardson number almost never goes below the critical threshold $1/4$ necessary for the development of the shear instability. The saturation of the vertical shear is due to the decay of the vortex in the regions of high ambient shear. These results suggest that the shear instability is not easilytriggered by decorrelation processes in strongly stratified flows in contradiction with previous conjectures

This thesis presents phenomenological studies of a state-of-the-art NNLL+NLO theoretical calculation of a novel collider observable known as 'phi star'. In these studies the 'phi star' observable, a measure of azimuthal decorrelation, is applied directly to the leptons in the production of massive lepton pairs in hadron collisions (the Drell-Yan process). This provides an alternate measure of the recoil of the massive vector boson (Z/gamma) against initial state QCD radiation, but with distinct experimental advantages over the traditional boson transverse momentum. Attention is focused on the small-'phi star' regime (the quasi-back-to-back regime) where the infrared dynamics of soft/collinear gluon emissions become important. These phenomenological studies are followed up with the presentation of a measurement of 'phi star' in 'Z to mu mu' events using 20.3 fb^-1 of collision data collected by the ATLAS experiment in 2012. Finally, studies directly related to the ATLAS absolute luminosity calibration by the van der Meer (vdM) method are presented, with the objective of elucidating the role of transverse linear beam correlation. In particular, I present studies using an analytical method I have developed in order to precisely extract individual beam information by way of studying phenomena pertaining to the luminous region during vdM scans. In addition, a dedicated study of the long- and short-term stabilities of the principal detectors for luminosity monitoring is also presented, along with an appropriate recalibration of these detectors.

Darbe analizuojama diskrečioji Le Gall transformacija (DLGT) yra viena iš bangelių transformacijų, kurios naudojamos skaitmeninių vaizdų glaudinime. Vaizdų glaudinimas ypač svarbus, kai kalbama apie vaizdų siuntimą mažo pralaidumo kompiuteriniu kanalu. Šiai problemai spręsti pasitelkiama diskrečioji Le Gall transformacija, bei didelis dėmesys darbe skirtas vaizdo fragmento DLGT spektrui. Šio spektro radimui sukurtas ir realizuotas naujas algoritmas − greitoji procedūra. Palyginamosios analizės metu įrodyta, kad sukurta greitoji procedūra randa vaizdo fragmento DLGT spektrą daugiau nei 50 kartų greičiau nei tiesiogiai taikant DLGT vaizdo fragmentui. Darbo pradžioje aptariamos pagrindinės bangelių ir jų transformacijų savybės. Šių savybių svarba iliustruojama pateiktais pavyzdžiais ir sąsajomis su taikymu praktikoje. Viena iš svarbesnių aptariamų savybių būtų transformacijos lokalizavimas erdvėje. Nors DLGT pilnai netenkina šios savybės, tačiau darbe apžvelgtas dekoreliacijos metodas leidžia išspręsti šią problemą. Be to, pateikiami išsamūs DLGT algoritmai vienmačiams ir dvimačiams vaizdams. Šių algoritmų dėka buvo sukurta programinė įranga, padedanti analizuoti DLGT. Galiausiai darbo pabaigoje pateikiama sukurtos programinės įrangos instrukcija.
In this paper the main attention is dedicated to discrete Le Gall transformation (DLGT), the family of discrete wavelets transformations, which usage in the image compression is quite popular. Image compression is very important when it comes to low bandwidth network and computer channels. To deal with this problem the discrete Le Gall wavelets are brought up and focused on the image fragment DLGT spectrum. For calculation of this spectrum the new fast procedure is suggested and realized. In comparative analysis the fast procedure is proved to work more than 50 times faster than applying DLGT direct to image fragment. In the beginning of this paper the main concepts of wavelets and their transformations are analyzed. An importance of these properties is illustrated by examples and applications in practice. One of the most important properties is localization in space. Even though DLGT does not fully meet this property decorrelation technique, introduced in paper, benefits in solving localization problem. Moreover, DLGT algorithms are proposed for one-dimensional and 2-dimensional images. According to these algorithms the new program was created, which is dedicated for image analysis. Finally, in the end of paper the instructions of program are presented.

A white Gaussian noise measurement model is widely used in target tracking problem formulation. In practice, the measurement noise may not be white. This phenomenon is due to the scintillation of the target. In many radar systems, the measurement frequency is high enough so that the correlation cannot be ignored without degrading tracking performance. In this thesis, target tracking problem with correlated measurement noise is considered. The correlated measurement noise is modeled by a first-order Markov model. The effect of correlation is thought as interference, and Optimum Decoding Based Smoothing Algorithm is applied. For linear models, the estimation performances of Optimum Decoding Based Smoothing Algorithm are compared with the performances of Alpha-Beta Filter Algorithm. For nonlinear models, the estimation performances of Optimum Decoding Based Smoothing Algorithm are compared with the performances of Extended Kalman Filter by performing various simulations.

Thesis (M.S.)--Worcester Polytechnic Institute.
Keywords: angular decorrelation function; impulse response; BUA; BMD; ultrasound; coral; trabecular bone; osteoporosis. Includes bibliographical references (p. 189-191).

Existe uma variedade de modelos de dispersão de poluentes, em geral, os modelos gaussianos são usados em todo o mundo por agências ambientais com intuito de regulação. O modelo CALPUFF é um deles. Neste estudo, _e avaliada a influência de escalas de tempo de descorrelação no sistema de modelagem CALPUFF, sob condições atmosféricas neutras. Para fazer isso uma nova parametrização das escalas de tempo descorrelação é proposta. Uma distribuição espectral de um perfil de velocidade euleriana e uma for- mutação da evolução das escalas temporais de descorrelação Lagrangiana são utilizadas como os mecanismos forçantes na dispersão turbulenta, numa camada limite dominada pelo cisalhamento do vento. O desempenho do modelo foi estabelecido com a comparação das concentrações superficiais do experimento Over-Land Alongwind Dispersion. Emissões de fontes em linha foram avaliadas com o modelo CALPUFF com distintas formas de inicialização. Um segundo modelo também foi testado, normalmente utilizado para estudar e prever o impacto ambiental e validar parametrizações turbulentas. É o modelo estocástico de partículas Lagrangiano LAMBDA. Também neste estudo, os resultados do modelo LAMBDA e CALPUFF enfatizam a capacidade da nova derivação de escalas de tempo em representar o comportamento estocástico desconhecido do fenômeno da dispersão de poluentes.
There exists a variety of pollution of dispersion models and in general, Gaussian models are used worldwide by environmental agencies in regulatory applications. The CALPUFF model is one of them. In this study, the in uence of decorrelation time scales in the CALPUFF modeling system under neutral conditions is evaluated. To do this a new parameterization of decorrelation time scales is proposed. A spectral distribution of an Eulerian velocity pro le and a formulation of the evolution of the Lagrangian decorrela- tion timescales are used as the forcing mechanisms (shear-dominated boundary layer) for the turbulent dispersion. The model performance was established by comparing ground- level concentrations with Over-Land Alongwind Dispersion experimental results. Line source emissions was evaluated using the CALPUFF model with different forms of the initialization. A second model was also tested, normally used to study and predict the environmental impact and validate turbulent parameterizations. Is the stochastic La- grangian dispersion model LAMBDA (Ferrero and Anfossi, 1998). Also in this study the model LAMBDA and CALPUFF results emphasized the ability of the new derivation of decorrelation time scales to represent the unknown stochastic behavior.

Dans les milieux fortement hétérogènes, les ondes ne se propagent pas de façon balistique. Elles peuvent interagir de nombreuses fois avec les hétérogénéités du milieu et entrer ainsi dans le régime de diffusion multiple. Dans ce régime, les méthodes classiques d'imagerie basées sur les trajets des ondes directes ou simplement diffusées sont inefficaces. Les formes d'ondes multiplement diffusées (coda) sont trop complexes pour être modélisées exactement mais elles sont parfaitement reproductibles et très sensibles aux variations du milieu de propagation. Des travaux récents ont démontré la possibilité de mesurer de faibles changements de vitesse d'un milieu grâce aux ondes diffuses, à l'échelle de la croûte terrestre (coda sismique) comme à l'échelle des matériaux (coda ultrasonore). Ces travaux s'intéressent majoritairement à des changements globaux ou régionaux des différents milieux. La problématique de cette thèse concerne la possibilité d'utiliser la coda pour étudier des changements locaux du milieu. Deux études complémentaires sont développées : Le problème direct consiste à modéliser les variations de la coda engendrées par un changement local. Nous distinguons le cas d'un changement local de structure (fort contraste d'impédance) du cas d'un changement local de vitesse (faible contraste d'impédance). Le problème inverse consiste à utiliser les mesures de variations de la coda pour tenter de localiser et de caractériser les changements survenus. Les applications potentielles de ces travaux concernent entre autres le suivi temporel de structures géologiques ainsi que le contrôle non destructif de matériaux hétérogènes. Avec cet objectif, les différentes méthodes développées sont illustrées par des simulations numériques d'ondes acoustiques et sismiques ainsi que par des expériences en ultrasons dans des éléments en béton.

We live in a dynamic world full of noises and interferences. The conventional acoustic echo cancellation (AEC) framework based on the least mean square (LMS) algorithm by itself lacks the ability to handle many secondary signals that interfere with the adaptive filtering process, e.g., local speech and background noise. In this dissertation, we build a foundation for what we refer to as the system approach to signal enhancement as we focus on the AEC problem. We first propose the residual echo enhancement (REE) technique that utilizes the error recovery nonlinearity (ERN) to "enhances" the filter estimation error prior to the filter adaptation. The single-channel AEC problem can be viewed as a special case of semi-blind source separation (SBSS) where one of the source signals is partially known, i.e., the far-end microphone signal that generates the near-end acoustic echo. SBSS optimized via independent component analysis (ICA) leads to the system combination of the LMS algorithm with the ERN that allows for continuous and stable adaptation even during double talk. Second, we extend the system perspective to the decorrelation problem for AEC, where we show that the REE procedure can be applied effectively in a multi-channel AEC (MCAEC) setting to indirectly assist the recovery of lost AEC performance due to inter-channel correlation, known generally as the "non-uniqueness" problem. We develop a novel, computationally efficient technique of frequency-domain resampling (FDR) that effectively alleviates the non-uniqueness problem directly while introducing minimal distortion to signal quality and statistics. We also apply the system approach to the multi-delay filter (MDF) that suffers from the inter-block correlation problem. Finally, we generalize the MCAEC problem in the SBSS framework and discuss many issues related to the implementation of an SBSS system. We propose a constrained batch-online implementation of SBSS that stabilizes the convergence behavior even in the worst case scenario of a single far-end talker along with the non-uniqueness condition on the far-end mixing system. The proposed techniques are developed from a pragmatic standpoint, motivated by real-world problems in acoustic and audio signal processing. Generalization of the orthogonality principle to the system level of an AEC problem allows us to relate AEC to source separation that seeks to maximize the independence, hence implicitly the orthogonality, not only between the error signal and the far-end signal, but rather, among all signals involved. The system approach, for which the REE paradigm is just one realization, enables the encompassing of many traditional signal enhancement techniques in analytically consistent yet practically effective manner for solving the enhancement problem in a very noisy and disruptive acoustic mixing environment.

Cephalopods constitute an important faunistic group in marine ecosystems, representing keystone elements in marine food webs. Additionally, they are an important fishing resource, and their contribution to worldwide fishery landings has been rising significantly over the last decades. In the Mediterranean Sea, several local short-term studies on cephalopod abundance and distribution exist, but few integrate longer time scales and larger areas. The present thesis work combines large-scale, long-term studies on cephalopods community composition, diversity and abundance (Part I) with small-scale, local case studies around the Balearic Islands (Part II). First, a community analysis was conducted at the whole Mediterranean scale in order to describe the present status and reveal possible differences between the western and the eastern Mediterranean sub-basins (Chapter 2). The aim was to compare assemblage structures between sub-basins together with the influence of large-scale environmental gradients on species composition. While basins did not differ significantly in cephalopod species composition, relative species contributions varied. Cephalopods are very sensitive to environmental changes, and adapt quickly due to their short life cycle and typically seasonal semelparity impeding generation buffering. Therefore sea surface temperature (SST) and chlorophyll a (Chl a) were hypothesised and documented as important drivers of cephalopod assemblages. While their importance differed between sub-basins, depth was confirmed as the most important factor structuring cephalopod communities. At the same spatial scale, cephalopod diversity was analysed during two decades (Chapter 3). Results revealed the absence of clear longitudinal or latitudinal gradients in cephalopod diversity, combined with no distinct temporal trends over the last twenty years. Besides, temporal changes in diversity varied between various Mediterranean sub-regions. Location, depth, SST and productivity regime resulted to influence diversity. To assess whether environmental variability impacts cephalopod populations at species level, the influence of SST and surface primary production on the spatial structuring and the population dynamics of two commercially important cephalopod species was investigated (Chapter 4). Although Octopus vulgaris and Illex coindetii exhibit contrasting life history traits and inhabit different habitats, long-term abundance trends and synchronic behaviour of their populations was similar, and population sub-structuring across the Mediterranean Sea were revealed. In consistency with former chapters, SST and Chl a regimes were shown to be significant drivers of distribution patterns and inter-annual abundance fluctuations. The motivation for the second part of the thesis was fuelled by the present situation in the Mediterranean Sea, where cephalopod stocks are presumably overexploited, but where their ad-hoc assessment and management is far from being implemented. To contribute to existing knowledge of the current situation in the Balearic Sea, a case study on cuttlefish fishery, life-history and population dynamics was conducted using short-term and intra-annual data (Chapter 5). Possible drivers of cuttlefish abundance fluctuations were analysed, revealing significant influences of SST and local climatic variability. While the study provided an overview of possible intrinsic and extrinsic drivers of cuttlefish population fluctuations in the Balearic Sea, the next step was applying stock assessment techniques to cuttlefish and squid fisheries of this area (Chapter 6). Being the first application of depletion methods to Mediterranean cephalopod stocks, the study combined information of those two fisheries with the evaluation of this stock assessment method for short life cycle species. Results uncovered existing problems and suggested that the current sampling scheme under the EU-Data Collection Framework is inadequate for cephalopods in view of the assessment´s data requirements. Consequently, a modification of the existing scheme was proposed.
Los cefalópodos constituyen un grupo faunístico importante en los ecosistemas marinos, representando elementos clave en la estructura y dinámica de las redes tróficas. Además, son un importante recurso pesquero y su contribución a los desembarcos pesqueros mundiales se ha incrementado significativamente en las últimas décadas. En el Mar Mediterráneo existen varios estudios locales sobre la abundancia y distribución de cefalópodos, pero pocos integran escalas temporales largas y áreas extensas. Esta tesis combina estudios sobre la estructura de la comunidad, la diversidad y la abundancia de cefalópodos a gran escala espacio-temporal (Parte I) con estudios a escala local alrededor de las Islas Baleares (Parte II). En primer lugar, se realizó un análisis de la comunidad de cefalópodos a escala de todo el Mediterráneo para describir el estado actual y determinar posibles diferencias entre las sub-cuencas occidental y oriental (Capítulo 2). El objetivo fue comparar la estructura de las comunidades faunísticas entre estas dos sub-cuencas y analizar la influencia de gradientes ambientales a gran escala sobre la composición específica. Aunque no se observaron diferencias significativas en la composición específica, las contribuciones relativas de las diferentes especies variaron entre ambas cuencas. Debido a su corto ciclo vital y estrategia reproductiva semélpara, los cefalópodos son muy sensibles a los cambios ambientales, adaptándose a ellos con relativa rapidez. Se esperaba, por tanto, que factores ambientales como la temperatura superficial del mar (SST) y el contenido de clorofila a (Chl a) fueran factores importantes en la estructuración de las comunidades de cefalópodos. Si bien su importancia difirió entre las dos sub-cuencas, se confirmó que la profundidad era el factor estructural más importante en dichas comunidades. Se analizaron también cambios en la diversidad específica de las comunidades de cefalópodos a escala de todo el Mediterráneo durante los últimos 20 años (Capítulo 3). Los resultados no mostraron ninguna tendencia temporal ni espacial (latitud, longitud) clara y homogénea en la diversidad. Las únicas tendencias temporales observadas variaron entre diferentes sub-regiones, siendo la posición geográfica, profundidad, SST y productividad los principales factores determinantes de estas tendencias. Para confirmar la importancia de la variabilidad ambiental a nivel específico, se analizó la influencia de la SST y la producción primaria en la estructura espacial y la dinámica poblacional de dos especies de cefalópodos de importancia comercial (Capítulo 4). Aunque Octopus vulgaris e Illex coindetii exhiben diferentes estrategias vitales y viven en diferentes hábitats, las tendencias temporales a largo plazo en su abundancia y el comportamiento sincrónico de sus poblaciones fueron similares, observándose sub-poblaciones temporalmente persistentes a lo largo de todo el Mediterráneo. En consonancia con los capítulos anteriores, la SST y Chla resultaron ser factores significativos en la definición de patrones de distribución y fluctuaciones interanuales de la abundancia. La segunda parte de la tesis estaba motivada por la situación actual de los cefalópodos en el Mar Mediterráneo, donde no existen medidas de evaluación y gestión específica para este grupo y algunas de sus poblaciones se encuentran sobreexplotadas. Para contribuir al conocimiento de la situación actual en el Mar Balear, se analizó la pesquería, ciclo vital y dinámica poblacional de la sepia utilizando series temporales intra-anuales (Capítulo 5). El análisis de diferentes factores ambientales reveló que la SST y la variabilidad climática local influían significativamente en las fluctuaciones de abundancia de la especie. En un trabajo posterior (Capítulo 6) se aplicaron técnicas de evaluación de stocks a las pesquerías de sepia y calamar de las Islas Baleares utilizando datos procedentes de la EU Data Collection Framework (DCF). Hasta donde sabemos, este trabajo constituye la primera aplicación de los Lesly-Delury Depletion Methods a stocks de cefalópodos del Mediterráneo. Los resultados demostraron que el esquema actual de muestreo de la DCF es inadecuado para la evaluación de cefalópodos y se proponen posibles modificaciones.
Els cefalòpodes constitueixen un grup faunístic important als ecosistemes marins, essent elements clau a les xarxes tròfiques. A més, són importants recursos pesquers i la seva contribució als desembarcs pesquers mundials s’ha incrementat en les últimes dècades. representant al voltant del 30% de la captura desembarcada a nivell mundial. Al Mar Mediterrani existeixen varis estudis locals sobre l’abundància i distribució de cefalòpodes, però pocs integren escales temporals llargues i àrees extenses. Aquesta tesi combina estudis sobre la composició de la comunitat, la diversitat i l’abundància de cefalòpodes a gran escala espaciotemporal (Part I) amb estudis a escala local al voltant de les Illes Balears (Part II). En primer lloc, es va realitzar un anàlisi de les comunitats de cefalòpodes a escala de tot el Mediterrani per descriure l’estat actual i determinar possibles diferències entre les sub-conques occidental i oriental (Capítol 2). L’objectiu fou comparar l’estructura de les comunitats faunístiques entre sub-conques i analitzar la influència de gradients ambientals a gran escala sobre la composició específica. Tot i que no s’observaren diferències significatives en la composició específica, les contribucions relatives de les diferents espècies variaren entre sub-conques. Degut al seu cicle vital curt i estratègia reproductiva semèlpara els cefalòpodes són molt sensibles als canvis ambientals, adaptant-se a ells amb relativa rapidesa. S’esperava, per tant, que factors ambientals com la temperatura superficial del mar (SST) i el contingut de clorofil·la a (Chla) fossin factors importants en l’estructuració de les comunitats de cefalòpodes. Si bé la seva importància va diferir entres les dues sub-conques, es va confirmar que la profunditat era el factor estructural més important en aquestes comunitats. S’analitzaren també canvies en la diversitat específica de les comunitats de cefalòpodes a escala de tot el Mediterrani durant els últims 20 anys (Capítol 3). Els resultats no mostraren cap tendència temporal ni espacial (latitud, longitud) clara i homogènia en la diversitat. Les úniques tendències temporals observades variaren entre diferents sub-regions, essent la posició geogràfica, profunditat, SST i productivitat els principals factors determinants d’aquestes tendències. Per tal de confirmar la importància de la variabilitat ambiental a nivell específic, es va investigar la influència de la SST i la producció primària en l’estructuració espacial i la dinàmica poblacional de dues espècies de cefalòpodes d’importància comercial (Capítol 4). Tot i que Octopus vulgaris i Illex coindetii tenen característiques vitals diferents i viuen en hàbitats diferents, les tendències temporals a llarg plaç en la seva l’abundància i el comportament sincrònic de les seves poblacions va ser similar, observant-se sub-poblacions temporalment persistents al llarg del Mediterrani. En consonància amb els capítols anteriors, la SST i Chla resultaren ser factors significatius en la definició de patrons de distribució i fluctuacions interanuals de l’abundància. La segona part de la tesi estava motivada per la situació actual dels cefalòpodes al Mar Mediterrani, on no existeixen mesures d’avaluació i gestió específica per aquest grup i algunes de les seves poblacions es troben sobre-explotades. Per contribuir al coneixement de la situació actual al Mar Balear, es va analitzar la pesqueria, cicle vital i dinàmica poblacional de la sípia utilitzant sèries temporals intra-anuals (Capítol 5). L’anàlisi de diferents factors ambientals va revelar que la SST i la variabilitat climàtica local influïen significativament en les fluctuacions d’abundància de l’espècie. En un treball posterior (Capítol 6) s’aplicaren tècniques d’avaluació d’stocks a les pesqueries de sípia i calamar de les Illes Balears utilitzant dades procedents de la EU Data Collection Framework (DCF). Fins on sabem, aquest treball constitueix la primera aplicació dels Lesly-Delury Depletion Methods a stocks de cefalòpodes del Mediterrani. Els resultats demostraren que l’esquema actual de mostratge de la DCF és inadequat per a l’avaluació de cefalòpodes i es proposen possibles modificacions.

The prediction of recoverable resources at an operating manganese mine is currently undertaken using univariate ordinary kriging of the target variable manganese, and 5 deleterious variables. Input data densities at the time of this calculation are considerably lower than at the time of final selection (grade control), and the potential for unnacceptable conditional bias to be introduced through the use of linear geostatistical methods when determining grade estimates over a small support has led to assessment of the potential benefit of employing the local change of support methods Localised Uniform Conditioning (LUC) and Conditional Simulation (CS). Allowances for the operating conditions, including time frames for estimation / simulation, and likely software limitations are accounted for by also requiring decorrelation to be used in instances where the data are considered in a multivariate sense. A novel method for decorrelation of geostatistical datasets, Independent Components Analysis (ICA), is compared against the more common method of Minimum-Maximum Autocorrelation Factorisation (MAF). ICA performs comparably against MAF in terms of its ability to diagonalise the variance-covariance matrix of the test dataset over multiple lags, for a variety of input data densities and treatments (log-ratio transformed and raw oxide data). Based on these results, ICA decorrelated data were incorporated into a comparative study of LUC and CS against block ordinary kriging (BOK), using an input dataset of reduced density, treated variously as raw univariate oxide data, decorrelated oxide data, and log-ratio transformed decorrelated data. The use of the log-ratio transform, designed to account for the 100% sum constraint inherent to the input data, proved impractical for LUC due to difficulties associated with the discrete Gaussian model change of support method employed by this technique. Log-ratio data transformation was restricted to use with CS where back transformation to raw oxide space could take place on a pseudo-equivalent support to the input data, prior to change of support. While use of the log-ratio transformation for CS guaranteed adherence to the sum constraint for results (the only method to do so) it resulted in distortion to both the spatial and grade distribution of results. Decorrelation by ICA also posed difficulties, with biases introduced to final back transformed results as a result of the decorrelation algorithm in both log-ratio transformed and oxide data, which in some instances caused impossible negative values to be returned for some variables in the final results. In a comparison of net profit calculations for each method, the distortions introduced from both log-ratio transformation, and decorrelation become evident in either overly optimistic or conservative profit distributions for methods in which they were used. Of the results presented, only BOK, CS and LUC of non-decorrelated oxide data appear to show results similar to those which would be used at the operation during final selection (based on ordinary kriging of a complete dataset). Based on the comparison of spatial grade distributions and both net profit spatial distribution and summary, the decision to employ a non-linear method of recoverable resource calculation at the operation under question would be questionable in terms of its reward for effort, given that the current method of BOK appears to produce equivalent results.

Cette thèse concerne la restauration aveugle d’images (formulée comme un problème inverse mal-posé et mal-conditionné), en considérant particulièrement les systèmes SIMO. Dans un premier temps une technique d’identification aveugle de ce système où l’ordre du canal est inconnu (surestimé) est introduite. Nous introduisons d’abord une version simplifiée à coût réduit SCR de la méthode des relations croisées (CR). Ensuite, une version robuste R-SCR basée sur la recherche d’une solution parcimonieuse minimisant la fonction de coût CR est proposée. La restauration d’image est ensuite assurée par une nouvelle approche inspirée des techniques de décodage des signaux 1D et étendue ici aux cas de la restauration d’images en se basant sur une recherche arborescente efficace (algorithme ‘Stack’). Plusieurs améliorations de la méthode ‘Stack’ ont été introduites afin de réduire sa complexité et d’améliorer la qualité de restauration lorsque les images sont fortement bruitées. Ceci en utilisant une technique de régularisation et une approche d’optimisation all-at-once basée sur la descente du gradient qui permet de raffiner l’image estimée et mieux converger vers la solution optimale. Ensuite, les mesures de la qualité d’images sont utilisées comme fonctions de coûts (intégrées dans le critère global) et ce afin d’étudier leur potentiel pour améliorer les performances de restauration. Dans le contexte où l’image d’intérêt est corrompue par d’autres images interférentes, sa restauration nécessite le recours aux techniques de séparation aveugle de sources. Pour cela, une étude comparative de certaines techniques de séparation basées sur la propriété de décorrélation au second ordre et la parcimonie est réalisée
This thesis concerns the blind restoration of images (formulated as an ill-posed and illconditioned inverse problem), considering a SIMO system. Thus, a blind system identification technique in which the order of the channel is unknown (overestimated) is introduced. Firstly, a simplified version at reduced cost SCR of the cross relation (CR) method is introduced. Secondly, a robust version R-SCR based on the search for a sparse solution minimizing the CR cost function is proposed. Image restoration is then achieved by a new approach (inspired from 1D signal decoding techniques and extended here to the case of 2D images) based on an efficient tree search (Stack algorithm). Several improvements to the ‘Stack’ method have been introduced in order to reduce its complexity and to improve the restoration quality when the images are noisy. This is done using a regularization technique and an all-at-once optimization approach based on the gradient descent which refines the estimated image and improves the algorithm’s convergence towards the optimal solution. Then, image quality measurements are used as cost functions (integrated in the global criterion), in order to study their potential for improving restoration performance. In the context where the image of interest is corrupted by other interfering images, its restoration requires the use of blind sources separation techniques. In this sense, a comparative study of some separation techniques based on the property of second-order decorrelation and sparsity is performed

碩士
國立臺灣大學
電信工程學研究所
89
In a code division multiple access (CDMA) system, as the number of active users increases, multiple access interference (MAI) will degrade the system performance severely. Therefore, the CDMA system is an inference limited multiple access system, and the capacity is dominant by MAI, i.e., soft capacity. Multiuser detection (MUD) is investigated to suppress MAI, improve the performance and then increase the capacity of CDMA systems. Typically, for an asynchronous CDMA system, the first step of MUD is to transform the asynchronous system into an equivalent synchronous system. Furthermore, if the signals are transmitted in a multipath environment, the signal in each path is treated as a different user’s signal. Then, at the receiver, the multipath signals belonging to the same user will be combined to attain a better decision, i.e., RAKE receiver. The disadvantages of this method are the large complexity and long detection delay as the number of users and paths increase. In this thesis, we’ll propose a linear decorrelating detector (LDD), one type of MUD that can reduce the detection delay to be less than one bit duration. Furthermore, by exploiting an adaptive algorithm, we can reduce the complexity of the proposed scheme to the simplest as same as the complexity of the LDD used in synchronous CDMA systems and maintain the performance.

碩士
國立中山大學
通訊工程研究所
98
In this thesis, we consider the uplink of a cooperative code division multiple access (CDMA) network, where users cooperate by relaying each other’s messages to the base station. The sources adopt CDMA to share the informations offered by relays. In general, spreading waveforms of sources are not orthogonal to each others due to pratical design issues of CDMA network. Therefore, the source signals will suffer from multiple access interference (MAI) at the relays and destination. The MAI results in the increase of bit error rate increased, and diminishes the cooperative network diveristy gains. In order to mitigate MAI, the decorrelating multiuser detection and zero-forcing precoder have been commonly adopted. But, the decorrelating multiuser detection causes noise enhancement, while the zero-forcing precoder causes power expansion. In this thesis, we adopt relay-assisted decorrelating multiuser detector (RAD-MUD) to mitigate MAI as proposed in [1].In this scheme, the relays perform half of decorrelating operation and the destination performs the other half. In this way, neither noise enhancement nor power expansion will occur. However, in the reference [1], the authers assume the transmission is synchronous between sources and relays. The assumption is unrealistic, because it is difficult to achieve synchronization between all sources and relays due to various propagation delays. In this thesis, we extend the research in [1] and relax restriction of synchronization between all sources and relays. Besides, we also adopt cooperative strategies such as transmit beamforming or selective relaying to enhance system performance. Compared with other’s multiuser detection schemes, we show that the proposed schemes can effectively reduce the bit error rate.

碩士
國立中央大學
土木工程研究所
90
Phase Ambiguity plays an important role in GPS positioning using phase Data. There are two major ways to deal with the ambiguity resolution, vie to solve for and to eliminate. In order to solve for high correlation between ambiguities and geometry parameters, the decorrelation method is used, and then to fix the integer ambiguities. Dual reference GPS positioning is then going on after the ambiguities fixed. In the dual reference GPS positioning, there is redundant information, i.e. ambiguity constraint. By adding ambiguity constraint into the dual reference GPS positioning model, the theoretical parameter accuracy can be improved. After parameter decorrelating ,the correct integer ambiguities can be solved and single epoch positioning can then be carried on. While doing single epoch positioning, there are different appearances in the models by adding ionosphere effecting parameter or introducing variance and covariance component estimation. Base on the analysis between the models, the results coming form the dual reference positioning shows its superior to the single reference positioning.

碩士
國立中興大學
電機工程學系
90
We propose a multi-user detection scheme for downlink channel and uplink link channel, including multi-path effect. The Quasi-Decorrelating detector (QDD) is a linear detector but the decorrelating algorithm is performed without inverse matrix transform, as an effort to overcome the problem associated with the DD. The QDD can be implemented by multi-stage feed-forward filters, the number os which can be made scalable to trade MAI suppression for noise enhancement reduction according to varying channel conditions. On the contrary, the DD has a rigid structure and is unable to adapt to changing operation environment. For downlink channel, we considering the multi-path and fading effect, and we assume that there are 12 kinds of channels to simulation. To improve the system performance, two different types of RAKE receiver structures, equal gain combining RAKE and maximum ratio combining RAKE, are also considered for CDMA systems to mitigate multi-path fading effects of mobile channels. For uplink channel, the assumption are the same with downlink channel, the different is just the users are asynchronous each other.

碩士
國立海洋大學
電機工程學系
92
The second generation mobile communication system has shor-tcomings, such as low data rate, poor quality of service(Qos), limited service types -- speech and data only, and low system capacity. But the third generation mobile communication system overcomes the shortcomings above, and direct-sequence code division multiple access (DS-CDMA) is one of the best candi-dates.It has advantages; for example, anti-mulpath, anti-noise, keeping a secret well and high capacity at the cost of band-width. However, multiple access interference (MAI) between the users contributes to the limitation of Qos and capacity. Trad-itionally, the multiuser detection (MUD) scheme is applied to reduce or eliminate the MAI. Although the decorrelating detector can remove the MAI comp-letely, has linear complexity in the number of users, and does not need the knowledge of users' signal energy, but it has noise enhancement, causing poor performance. A new structure that combines the overlapping window with decision feedback to improve the performance of the detector is proposed in the thesis. When removing the correlation between users, we replace the inverse matrix with the algorithm of recursive least square(RLS) in order to reduce the computation complexity of the detector. The overlapping window has the properties that allow the states of data bits undecided in the previous window to be decided in the next window. Decision feedback has the charact- eristic that in the environment of near-far effect, the perform-ance of the weakest SNR user can be improved due to the remove of MAI of the stronger SNR users. Simulation results have shown that the bit-error-rate of the detector using overlapping window and decision feedback is better. Therefore, it can make the communication quality perfect.

碩士
國立交通大學
電信工程研究所
84
This paper investigates the application of linear reinforcement learning stochastic approximation to the blind adaptive energy estimation for a decorrelating decision-feedback (DDF) multiuser detector over synchronous CDMA radio channel in the presence of multiple access interference (MAI) and additive Gaussian noise. The decision feedback incorporated into the structure of linear decorrelating detector is able to significantly improve the weaker users' performance by cancelling the MAI from the stronger users. However, the DDF receiver requires knowledge of the received energies. In this paper, a new novel blind estimation mechanism is proposed to estimate all the users' energies using a stochastic approximation algorithm without training data. In order to increase the convergence speed of the energy estimation, a linear reinforcement learning technique is conducted to accelerate the stochastic approximation algorithms. Results show that our blind adaptation mechanism is able to accurately estimate all the users' energies even the users of DDF detector are not ranked properly. After performing the blind energy estimation and then re-ordering the users in a nonincreasing order, numerical simulations show that the DDF detector for the weakest user performs closely to the maximum likelihood detector, whose complexity grows exponentically with the number of users.

碩士
國立中山大學
通訊工程研究所
100
In this paper, we explore a cooperative code-division-multiple-access(CDMA) network. Users cooperate by forwarding each other’s messages toward the destination. For simplicity, we assume that signal reception at the destination is well-synchronized. Due to practical design issues of CDMA systems, spreading waveforms allocated to users are not perfectly orthogonal in general. This results in multiple-access interference(MAI) at relays and destination. In CDMA uplink networks one common approach is to adopt decorrelating multi-user detection, but it will lead to noise amplification[16,17]. Therefore, we employ relay-assisted decorrelating multiuser detector(RAD-MUD) to mitigate MAI[1] by performing half of decorrelation at the relay and destination respectively. Based on the availability of CSI at relays, we can further adopt cooperative strategies to improve performance, e.g., transmit beamforming and selective relaying. The destination side will use minimum mean-square error(MMSE) detector to demodulate source symbols. In the existing literatures, channel state information(CSI) is assumed to be perfectly known at relay and destination. Actually, CSI is obtained from channel estimation, which usually contains estimation errors. In order to alleviate effects of channel estimation, one goal of this thesis is to design a robust system. Using estimated CSI and statistical property channel estimation errors, we design robust precoder and detector for the relay and destination. It shows that, even with distortion on channel estimations, the system still achieve excellent transmission efficiency. From the simulation results, it shows that the robust design is better than the system without consider channel estimation errors. Finally, we can see that the stable robust design can effectively mitigate effects of imperfect CSI.

碩士
大葉大學
電信工程學系碩士班
96
In this paper, we propose a two-stage linear multiuser detector (LMD) for Ultra wideband (UWB) Multiple Single-input Multiple-output (M-SIMO) system and multipath fading environment. Time-hopping (TH) and antipodal pulse amplitude modulation (PAM) are employed for the multiple access system. The decorrelating detector is first employed at the front end of each receive antenna to eliminate the multi-user interference (MUI), then a set of maximum-ratio-combiners (MRC) are proceeded to maximize the signal-to-noise power ratio (SNR) for each user. Since the channel information is crucial for the Decorrelating-MRC (D-MRC) receiver, we develop a subspace-based blind M-SIMO channel estimation method. The effect of channel estimation error on system performance is extensively evaluated. It is also verified from the analytical and numerical results that by exploiting both spatial and temporal diversities, the D-MRC outperforms the conventional single-antenna decorrelating detector. Moreover, we demonstrate that both the decorrelating-MRC (D-MRC) receiver and subspace based M-SIMO channel estimators are computationally feasible and near-far resistant.