No description available.

Regression tasks belong to the set of core problems faced in statistics and machine learning and promising approaches can often be generalized to also deal with classification, interpolation or denoising problems. Whereas the most widely used classical statistical techniques place severe a priori constraints on the type of function that can be approximated (e.g. only lines, in the case of linear regression), the successes of sparse kernel learners, such as the SVM (support vector machine) demonstrate that good results may be obtained in a quite general framework by enforcing sparsity. Similarly, even very simple sparsity-based denoising techniques, such as classical wavelet shrinkage, can produce surprisingly good results on a wide variety of different signals, because, unlike noise, most signals of practical interest share vital characteristics (such as smoothness, or the ability to be well approximated by piece-wise linear polynomials of a low order) that allow a sparse representation in wavelet space. On the other hand results obtained from SVMs (and classical wavelet-shrinkage) suffer from a certain lack of interpretability, since one cannot straightforwardly attach probabilities to them. By contrast regression, and even more importantly classification, in a Bayesian context always entails a probabilistic measure of confidence in the results, which, provided the model assumptions are reasonably accurate, forms a basis for principled decision-making. The relevance vector machine (RVM) combines these strengths by explicitly encoding the criterion of model sparsity as a (Bayesian) prior over the model weights and offers a single, unified paradigm to efficiently deal with regression as well as classification tasks. However the lack of an explicit prior structure over the weight variances means that the degree of sparsity is to a large extent controlled by the choice of kernel (and kernel parameters). This can lead to severe overfitting or oversmoothing -- possibly even both at the same time (e.g. for the multiscale Doppler data). This thesis details an efficient scheme to control sparsity in Bayesian regression by incorporating a flexible noise-dependent smoothness prior into the RVM. The resultant smooth RVM (sRVM) encompasses the original RVM as a special case, but empirical results with a variety of popular data sets show that it can surpass RVM performance in terms of goodness of fit and achieved sparsity as well as computational performance in many cases. As the smoothness prior effectively makes it possible to use (highly efficient) wavelet kernels in an RVM setting this work also unveils a strong connection between Bayesian wavelet shrinkage and RVM regression and effectively further extends the applicability of the RVM to denoising tasks for up to millions of datapoints. We further discuss its applicability to classification tasks.

No description available.

O presente trabalho apresenta um resumo da fisiologia de produção da voz humana, das patologias mais comuns da laringe e seus principais efeitos sobre o som fonado, e apresenta diversos métodos quantitativos de avaliação do som de vozes patológicas. É então proposto um novo método de avaliação da soprosidade da voz, baseado na Transformada Wavelet Discreta (DWT) através da análise multi-resolução, usando como base ortogonal de decomposição a base Haar. São feitas duas aquisições por dois diferentes procedimentos, dos sinais de voz de 64 pacientes. É mostrado que é possível se obter um índice acústico para a característica soprosidade da voz por intermédio da DWT. Este índice é estatisticamente correlacionado com dois outros índices existentes para soprosidade, para ambos os procedimentos de aquisição O método serve tanto para auxílio ao diagnóstico como acompanhamento dos resultados obtido por um tratamento. / This work shows a brief review of human voice production physiology, including the most common larynx pathologies and its effects in the voice quality, and the maio methods of pathologic quantitative vocal fold assessment. lt is proposed a novel method of breathiness of voice characteristic evaluation, based on the Discrete Wavelet Transform, using the orthonormal Haar basis as a reconstruction basis. Two procedures of data acquisition were used for the 64 subjects voice signals. Results are obtained and statistically compared with the ones obtained by classical methods, for both acquisition procedures. This method can be an auxiliary tool for the diagnosis as well as an assessment of a specific treatment.

O presente trabalho de pesquisa trata da elaboração de um esquema classificador automático para massas nodulares identificadas em imagens mamográficas digitalizadas, com base na técnica da transformada wavelet. Esse classificador é parte integrante de um esquema computadorizado para auxílio ao diagnóstico (CAD, de \"computer-aided diagnosis\") em mamografia, que utiliza técnicas de processamento de imagens digitais para identificar, realçar e classificar estruturas de interesse clínico. Utilizou-se também um classificador de distâncias mínimas para distribuir as imagens em suas respectivas classes. Os resultados mostraram que o classificador é capaz de diferenciar com mais de 90% de acerto entre nódulos suspeitos e não suspeito. / This work performs an automatic classifier scheme addressed to nodular masses detected in digitalized mammographic images, based on the wavelet transform technique. This classifier is part of a computer-aided diagnosis (CAD) scheme in mammography, wich uses digital image processing techniques in order to detect, enchance and classify structures of clinical interest. Also a minimum distances classifier was used in order to distribute the images to their respective classes. Results show that this classifier is capable of differentiating suspect from non-suspect nodules with more than 90% of accuracy.

Made available in DSpace on 2014-06-12T17:39:24Z (GMT). No. of bitstreams: 2 arquivo6866_1.pdf: 1749281 bytes, checksum: 54338dde5012e86da31feb93e4e2892e (MD5) license.txt: 1748 bytes, checksum: 8a4605be74aa9ea9d79846c1fba20a33 (MD5) Previous issue date: 2009 / A arte de ocultar uma mensagem dentro de outro objeto é conhecida como Esteganografia. Detalham-se técnicas convencionais para ocultação de mensagens e propõese uma nova abordagem. Este novo método de esteganografia em dois passos combina a cifragem do texto-pleno através de um criptossistema padrão, seguido pela imersão dos dados cifrados no arquivo de áudio. O trabalho enfoca-se na inserção de textos curtos em arquivos com formato wav a entrada dos dados é realizada nas componentes que resultam da transformação do sinal mediante as transformadas de wavelet. O objetivo é introduzir dados de forma quase transparente, de tal maneira que a detecção por terceiros seja pouco provável, como também para garantir a recuperação praticamente inalterável dos dados. O áudio é decomposto em doze níveis mediante a escolha de uma wavelet-mãe, os dados são codificados e ocultados nos diferentes níveis segundo o critério do usuário. Para um melhor espalhamento dos dados em cada nível são utilizadas senhas alfanuméricas de tamanho proporcional à quantidade de caracteres ingressados em cada um dos níveis. A implementação computacional foi realizada no Matlab® e simulações com arquivos de áudio de diferentes tamanhos foram realizadas. Mudanças nos arquivos de áudio após a inserção dos dados foram medidas. Baseadas no esquema da Esteganografia, aplicações comerciais podem ser desenvolvidas para garantir a autenticidade dos arquivos, assim como a proteção de direitos autorais em arquivos digitais

Dissertação (Mestrado em Engenharia Elétrica) - Centro Universitário FEI, São Bernardo do Campo, 2016.

The magnetosphere-ionosphere contains a number of current systems. These currents vary on a wide range of spatial and temporal scales and physically couple with each other. To study the complicated behaviors of these coupled current systems, the ground-based magnetometer has been a useful tool, but the recorded magnetometer data are always multi-scaled and intermittent due to the nature of these current systems. To distinguish these geomagnetic effects with multiple temporal and frequency scales, the wavelet analysis technique is especially suitable because of its special abilities of presenting information in both temporal and frequency domains. In this dissertation, the geomagnetic disturbances and the ring current variability during storm and quiet times are studied by using wavelet analysis and ground-based magnetic data from multiple stations. The first part of this dis- sertation investigates the strengths of applying the wavelet procedure to geomagnetic data for ring current study during storm and quiet periods. The second part of this dissertation characterizes the geomagnetic effects caused by symmetric and asymmetric components of ring currents during storm and quiet times by applying wavelet analysis to geomagnetic data from multiple stations. The third part of this dissertation studies the spatial variabil- ity of the symmetric ring current by applying the wavelet analysis technique to multiple components of magnetic data from multiple stations. The results show the unique strengths of the wavelet method allow us to quantitatively distinguish the geomagnetic effects on ring current variations from other M-I current systems. The unique strengths of wavelet method also allow us to separate the magnetic effects of the symmetric ring current from those caused by the asymmetric ring current. Quantitative information of the spatial variability of the ring currents is essential for understanding the dynamics of the ring currents, as well as the magnetic storm processes. The techniques developed in this dissertation have potential values as space weather monitoring tools for satellite controls, power grids, com- munication systems, oil pipelines, and other high-tech systems that are vulnerable to the negative impacts of disruptive geomagnetic events.

The construction of the generalized continuous wavelet transform (GCWT) on Hilbert spaces is a special case of the coherent state transform construction, where the coherent state system arises as an orbit of an admissible vector under a strongly continuous unitary representation of a locally compact group. In this thesis we extend this construction to the setting of Hilbert C*-modules. In particular, we define a coherent state transform and a GCWT on Hilbert modules. This construction gives a reconstruction formula and a resolution of the identity formula analogous to those found in the Hilbert space setting. Moreover, the existing theory of standard normalized tight frames in finite countably generated Hilbert modules can be viewed as a discrete case of this construction We also show that the image space of the coherent state transform on Hilbert module is a reproducing kernel Hilbert module. We discuss the kernel and the intertwining property of the group coherent state transform.

Circuit breaker is an important interrupting device in power system network. It usually has a lifetime about 20 to 40 years. During breaker's service time, maintenance and inspection are imperative duties to achieve its reliable operation. To automate the diagnostic practice for circuit breaker operation and reduce the utility company's workload, Wavelet based analysis software of circuit breaker operation is developed here. Combined with circuit breaker monitoring system, the analysis software processes the original circuit breaker information, speeds up the analysis time and provides stable and consistent evaluation for the circuit breaker operation.