Bearing failure is one of the foremost causes of breakdown in rotating machinery. Such failure can be catastrophic and can result in costly downtime. Bearing condition monitoring has thus played an important role in machine maintenance. In condition monitoring, the observed signal at a measurement point is often corrupted by extraneous noise during the transmission process. It is important to detect incipient faults in advance before catastrophic failure occurs. In condition monitoring, the early detection of incipient bearing signal is often made difficult due to its corruption by background vibration (noise). Numerous advanced signal processing techniques have been developed to detect defective bearing signals but with varying degree of success because they require a high Signal to Noise Ratio (SNR), and the fault components need to be larger than the background noise. Vibration analyses in the time and frequency domains are commonly used to detect machinery failure, but these methods require a relatively high SNR. Hence, it is essential to minimize the noise component in the observed signal before post processing is conducted. In this research, detection of failure in rolling element bearing faults by vibration analysis is investigated. The expected time intervals between the impacts of faulty bearing components signals are analysed using the blind deconvolution technique as a feature extraction technique to recover the source signal. Blind deconvolution refers to the process of learning the inverse of an unknown channel and applying it to the observed signal to recover the source signal of a damaged bearing. The estimation time period between the impacts is improved by using the technique and consequently provides a better approach to identify a damaged bearing. The procedure to obtain the optimum inverse equalizer filter is addressed to provide the filter parameters for the blind deconvolution process. The efficiency and robustness of the proposed algorithm is assessed initially using different kinds of corrupting noises. The result show that the proposed algorithm works well with simulated corrupting periodic noises. This research also shows that blind deconvolution behaves as a notch filter to remove the noise components. This research involves the application of blind deconvolution technique with optimum equalizer design for improving the SNR for the detection of damaged rolling element bearings. The filter length of the blind equalizer needs to be adjusted continuously due to different operating conditions, size and structure of the machines. To determine the optimum filter length a simulation test was conducted with a pre-recorded bearing signal (source) and corrupted with varying magnitude noise. From the output, the modified Crest Factor (CF) and Arithmetic Mean (AM) of the recovered signal can be plotted versus the filter length. The optimum filter length can be selected by observation when the plot converges close to the pre-determined source feature value. The filter length is selected based on the CF and AM plots, and these values are stored in a data training set for optimum determination of filter length using neural network. A pre-trained neural network is designed to train the behaviour of the system to target the optimum filter length. The performance of the blind deconvolution technique was assessed based on kurtosis values. The capability of blind deconvolution with optimum filter length developed from the simulation studies was further applied in a life bearing test rig. In this research, life time testing is also conducted to gauge the performance of the blind deconvolution technique in detecting a growing potential failure of a new bearing which is eventually run to failure. Results from unseeded new bearing tests are different, because seeded defects have certain defect characteristic frequencies which can be used to track a specific damaged frequency component. In this test, the test bearing was set to operate continuously until failures occurred. The proposed technique was then applied to monitor the condition of the test bearing and a trend of the bearing life was established. The results revealed the superiority of the technique in identifying the periodic components of the bearing before final break-down of the test bearing. The results show that the proposed technique with optimum filter length does improve the SNR of the deconvolved signal and can be used for automatic feature extraction and fault classification. This technique has potential for use in machine diagnostics.

No description available.

No description available.

In this thesis, we conduct functional Magnetic Resonance Imaging (fMRI) data analysis with the aim of grouping the brain voxels depending on their responsiveness to a neural task. We mathematically treat the fMRI signals as the convolution of the neural stimulus with the hemodynamic response function (HRF). We first estimate a time series including HRFs for each of the observed fMRI signals from a given set and we cluster them in order to identify the groups of brain voxels. The HRF estimation problem is studied within the Bayesian framework through a blind deconvolution algorithm using MAP approach under completely unsupervised and model-free settings, i.e, stimulus is assumed to be unknown and also no particular shape is assumed for the HRF. Only using a given fMRI signal together with a weak Gaussian prior distribution imposed on HRF favoring &lsquo / smoothness&rsquo / , our method successfully estimates all the components of our framework: the HRF, the stimulus and the noise process. Then, we propose to use a modified version of Hausdorff distance to detect similarities within the space of HRFs, spectrally transform the data using Laplacian Eigenmaps and finally cluster them through EM clustering. According to our simulations, our method proves to be robust to lag, sampling jitter, quadratic drift and AWGN (Additive White Gaussian Noise). In particular, we obtained 100% sensitivity and specificity in terms of detecting active and passive voxels in our real data experiments. To conclude with, we propose a new framework for a mathematical treatment for voxel-based fMRI data analysis and our findings show that even when the HRF is unpredictable due to variability in cognitive processes, one can still obtain very high quality activation detection through the method proposed in this thesis.

A Bussgang based blind deconvolution algorithm called self-correcting multi-channel Bussgang (SCMB) blind deconvolution algorithm was proposed. Unlike the original Bussgang blind deconvolution algorithm where the probability density function (pdf) of the signal being recovered is assumed to be completely known, the proposed SCMB blind deconvolution algorithm relaxes this restriction by parameterized the pdf with a Gaussian mixture model and expectation maximization (EM) algorithm, an iterative maximum likelihood approach, is employed to estimate the parameter side by side with the estimation of the equalization filters of the original Bussgang blind deconvolution algorithm. A feedback loop is also designed to compensate the effect of the parameter estimation error on the estimation of the equalization filters. Application of the SCMB blind deconvolution framework for binary image restoration, multi-pass synthetic aperture radar (SAR) autofocus and inverse synthetic aperture radar (ISAR) autofocus are exploited with great results.

Low-dimensional signal structures naturally arise in a large set of applications in various fields such as medical imaging, machine learning, signal, and array processing. A ubiquitous low-dimensional structure in signals and images is sparsity, and a new sampling theory; namely, compressive sensing, proves that the sparse signals and images can be reconstructed from incomplete measurements. The signal recovery is achieved using efficient algorithms such as \ell_1-minimization. Recently, the research focus has spun-off to encompass other interesting low-dimensional signal structures such as group-sparsity and low-rank structure. This thesis considers low-rank matrix recovery (LRMR) from various structured-random measurement ensembles. These results are then employed for the in depth investigation of the classical blind-deconvolution problem from a new perspective, and for the development of a framework for the efficient sampling of correlated signals (the signals lying in a subspace). In the first part, we study the blind deconvolution; separation of two unknown signals by observing their convolution. We recast the deconvolution of discrete signals w and x as a rank-1 matrix wx* recovery problem from a structured random measurement ensemble. The convex relaxation of the problem leads to a tractable semidefinite program. We show, using some of the mathematical tools developed recently for LRMR, that if we assume the signals convolved with one another live in known subspaces, then this semidefinite relaxation is provably effective. In the second part, we design various efficient sampling architectures for signals acquired using large arrays. The sampling architectures exploit the correlation in the signals to acquire them at a sub-Nyquist rate. The sampling devices are designed using analog components with clear implementation potential. For each of the sampling scheme, we show that the signal reconstruction can be framed as an LRMR problem from a structured-random measurement ensemble. The signals can be reconstructed using the familiar nuclear-norm minimization. The sampling theorems derived for each of the sampling architecture show that the LRMR framework produces the Shannon-Nyquist performance for the sub-Nyquist acquisition of correlated signals. In the final part, we study low-rank matrix factorizations using randomized linear algebra. This specific method allows us to use a least-squares program for the reconstruction of the unknown low-rank matrix from the samples of its row and column space. Based on the principles of this method, we then design sampling architectures that not only acquire correlated signals efficiently but also require a simple least-squares program for the signal reconstruction. A theoretical analysis of all of the LRMR problems above is presented in this thesis, which provides the sufficient measurements required for the successful reconstruction of the unknown low-rank matrix, and the upper bound on the recovery error in both noiseless and noisy cases. For each of the LRMR problem, we also provide a discussion of a computationally feasible algorithm, which includes a least-squares-based algorithm, and some of the fastest algorithms for solving nuclear-norm minimization.

No description available.

In many scenarios, events such as singularities and transients that carry important information about a signal undergo spreading during acquisition or transmission and it is important to localize the events. For example, edges in an image, point sources in a microscopy or astronomical image are blurred by the point-spread function (PSF) of the acquisition system, while in a speech signal, the epochs corresponding to glottal closure instants are shaped by the vocal tract response. Such events can be extracted with the help of techniques that promote sparsity, which enables separation of the smooth components from the transient ones. In this thesis, we consider development of such sparsity promoting techniques. The contributions of the thesis are three-fold: (i) an auditory-motivated continuous wavelet design and representation, which helps identify singularities; (ii) a sparsity-driven deconvolution technique; and (iii) a sparsity-driven deconvolution technique for reconstruction of nite-rate-of-innovation (FRI) signals. We use the speech signal for illustrating the performance of the techniques in the first two parts and super-resolution microscopy (2-D) for the third part. In the rst part, we develop a continuous wavelet transform (CWT) starting from an auditory motivation. Wavelet analysis provides good time and frequency localization, which has made it a popular tool for time-frequency analysis of signals. The CWT is a multiresolution analysis tool that involves decomposition of a signal using a constant-Q wavelet filterbank, akin to the time-frequency analysis performed by basilar membrane in the peripheral human auditory system. This connection motivated us to develop wavelets that possess auditory localization capabilities. Gammatone functions are extensively used in the modeling of the basilar membrane, but the non-zero average of the functions poses a hurdle. We construct bona de wavelets from the Gammatone function called Gammatone wavelets and analyze their properties such as admissibility, time-bandwidth product, vanishing moments, etc.. Of particular interest is the vanishing moments property, which enables the wavelet to suppress smooth regions in a signal leading to sparsi cation. We show how this property of the Gammatone wavelets coupled with multiresolution analysis could be employed for singularity and transient detection. Using these wavelets, we also construct equivalent lterbank models and obtain cepstral feature vectors out of such a representation. We show that the Gammatone wavelet cepstral coefficients (GWCC) are effective for robust speech recognition compared with mel-frequency cepstral coefficients (MFCC). In the second part, we consider the problem of sparse blind deconvolution (SBD) starting from a signal obtained as the convolution of an unknown PSF and a sparse excitation. The BD problem is ill-posed and the goal is to employ sparsity to come up with an accurate solution. We formulate the SBD problem within a Bayesian framework. The estimation of lter and excitation involves optimization of a cost function that consists of an `2 data- fidelity term and an `p-norm (p 2 [0; 1]) regularizer, as the sparsity promoting prior. Since the `p-norm is not differentiable at the origin, we consider a smoothed version of the `p-norm as a proxy in the optimization. Apart from the regularizer being non-convex, the data term is also non-convex in the filter and excitation as they are both unknown. We optimize the non-convex cost using an alternating minimization strategy, and develop an alternating `p `2 projections algorithm (ALPA). We demonstrate convergence of the iterative algorithm and analyze in detail the role of the pseudo-inverse solution as an initialization for the ALPA and provide probabilistic bounds on its accuracy considering the presence of noise and the condition number of the linear system of equations. We also consider the case of bounded noise and derive tight tail bounds using the Hoe ding inequality. As an application, we consider the problem of blind deconvolution of speech signals. In the linear model for speech production, voiced speech is assumed to be the result of a quasi-periodic impulse train exciting a vocal-tract lter. The locations of the impulses or epochs indicate the glottal closure instants and the spacing between them the pitch. Hence, the excitation in the case of voiced speech is sparse and its deconvolution from the vocal-tract filter is posed as a SBD problem. We employ ALPA for SBD and show that excitation obtained is sparser than the excitations obtained using sparse linear prediction, smoothed `1=`2 sparse blind deconvolution algorithm, and majorization-minimization-based sparse deconvolution techniques. We also consider the problem of epoch estimation and show that epochs estimated by ALPA in both clean and noisy conditions are closer to the instants indicated by the electroglottograph when with to the estimates provided by the zero-frequency ltering technique, which is the state-of-the-art epoch estimation technique. In the third part, we consider the problem of deconvolution of a specific class of continuous-time signals called nite-rate-of-innovation (FRI) signals, which are not bandlimited, but specified by a nite number of parameters over an observation interval. The signal is assumed to be a linear combination of delayed versions of a prototypical pulse. The reconstruction problem is posed as a 2-D SBD problem. The kernel is assumed to have a known form but with unknown parameters. Given the sampled version of the FRI signal, the delays quantized to the nearest point on the sampling grid are rst estimated using proximal-operator-based alternating `p `2 algorithm (ALPAprox), and then super-resolved to obtain o -grid (O. G.) estimates using gradient-descent optimization. The overall technique is termed OG-ALPAprox. We show application of OG-ALPAprox to a particular modality of super-resolution microscopy (SRM), called stochastic optical reconstruction microscopy (STORM). The resolution of the traditional optical microscope is limited by di raction and is termed as Abbe's limit. The goal of SRM is to engineer the optical imaging system to resolve structures in specimens, such as proteins, whose dimensions are smaller than the di raction limit. The specimen to be imaged is tagged or labeled with light-emitting or uorescent chemical compounds called uorophores. These compounds speci cally bind to proteins and exhibit uorescence upon excitation. The uorophores are assumed to be point sources and the light emitted by them undergo spreading due to di raction. STORM employs a sequential approach, wherein each step only a few uorophores are randomly excited and the image is captured by a sensor array. The obtained image is di raction-limited, however, the separation between the uorophores allows for localizing the point sources with high precision. The localization is performed using Gaussian peak- tting. This process of random excitation coupled with localization is performed sequentially and subsequently consolidated to obtain a high-resolution image. We pose the localization as a SBD problem and employ OG-ALPAprox to estimate the locations. We also report comparisons with the de facto standard Gaussian peak- tting algorithm and show that the statistical performance is superior. Experimental results on real data show that the reconstruction quality is on par with the Gaussian peak- tting.

abstract: Speech is generated by articulators acting on a phonatory source. Identification of this phonatory source and articulatory geometry are individually challenging and ill-posed problems, called speech separation and articulatory inversion, respectively. There exists a trade-off between decomposition and recovered articulatory geometry due to multiple possible mappings between an articulatory configuration and the speech produced. However, if measurements are obtained only from a microphone sensor, they lack any invasive insight and add additional challenge to an already difficult problem. A joint non-invasive estimation strategy that couples articulatory and phonatory knowledge would lead to better articulatory speech synthesis. In this thesis, a joint estimation strategy for speech separation and articulatory geometry recovery is studied. Unlike previous periodic/aperiodic decomposition methods that use stationary speech models within a frame, the proposed model presents a non-stationary speech decomposition method. A parametric glottal source model and an articulatory vocal tract response are represented in a dynamic state space formulation. The unknown parameters of the speech generation components are estimated using sequential Monte Carlo methods under some specific assumptions. The proposed approach is compared with other glottal inverse filtering methods, including iterative adaptive inverse filtering, state-space inverse filtering, and the quasi-closed phase method. / Dissertation/Thesis / Masters Thesis Electrical Engineering 2018

Dans les problèmes inverses en imagerie, on suppose généralement connu l’opérateur ou matrice décrivant le système de formation de l’image. De façon équivalente pour un système linéaire, on suppose connue sa réponse impulsionnelle. Toutefois, ceci n’est pas une hypothèse réaliste pour de nombreuses applications pratiques pour lesquelles cet opérateur n’est en fait pas connu (ou n’est connu qu’approximativement). On a alors affaire à un problème d’inversion dite “aveugle”. Dans le cas de systèmes invariants par translation, on parle de “déconvolution aveugle” car à la fois l’image ou objet de départ et la réponse impulsionnelle doivent être estimées à partir de la seule image observée qui résulte d’une convolution et est affectée d’erreurs de mesure. Ce problème est notoirement difficile et pour pallier les ambiguïtés et les instabilités numériques inhérentes à ce type d’inversions, il faut recourir à des informations ou contraintes supplémentaires, telles que la positivité qui s’est avérée un levier de stabilisation puissant dans les problèmes d’imagerie non aveugle. La thèse propose de nouveaux algorithmes d’inversion aveugle dans un cadre discret ou discrétisé, en supposant que l’image inconnue, la matrice à inverser et les données sont positives. Le problème est formulé comme un problème d’optimisation (non convexe) où le terme d’attache aux données à minimiser, modélisant soit le cas de données de type Poisson (divergence de Kullback-Leibler) ou affectées de bruit gaussien (moindres carrés), est augmenté par des termes de pénalité sur les inconnues du problème. La stratégie d’optimisation consiste en des ajustements alternés de l’image à reconstruire et de la matrice à inverser qui sont de type multiplicatif et résultent de la minimisation de fonctions coût “surrogées” valables dans le cas positif. Le cadre assez général permet d’utiliser plusieurs types de pénalités, y compris sur la variation totale (lissée) de l’image. Une normalisation éventuelle de la réponse impulsionnelle ou de la matrice est également prévue à chaque itération. Des résultats de convergence pour ces algorithmes sont établis dans la thèse, tant en ce qui concerne la décroissance des fonctions coût que la convergence de la suite des itérés vers un point stationnaire. La méthodologie proposée est validée avec succès par des simulations numériques relatives à différentes applications telle que la déconvolution aveugle d'images en astronomie, la factorisation en matrices positives pour l’imagerie hyperspectrale et la déconvolution de densités en statistique. / Doctorat en Sciences / info:eu-repo/semantics/nonPublished