Quadratic Predictor based Differential Encoding and Decoding of Speech Signals

Modeling nonlinear systems using Volterra series is a century old method but practical realizations were hampered by inadequate hardware to handle the increased computational complexity stemming from its use. But interest is renewed recently, in designing and implementing filters which can model much of the polynomial nonlinearities inherent in practical systems. The key advantage in resorting to Volterra power series for this purpose is that nonlinear filters so designed can be made to work in parallel with the existing LTI systems, yielding improved performance. This paper describes the inclusion of a quadratic predictor (with nonlinearity order 2) with a linear predictor in an analog source coding system. Analog coding schemes generally ignore the source generation mechanisms but focuses on high fidelity reconstruction at the receiver. The widely used method of differential pnlse code modulation (DPCM) for speech transmission uses a linear predictor to estimate the next possible value of the input speech signal. But this linear system do not account for the inherent nonlinearities in speech signals arising out of multiple reflections in the vocal tract. So a quadratic predictor is designed and implemented in parallel with the linear predictor to yield improved mean square error performance. The augmented speech coder is tested on speech signals transmitted over an additive white gaussian noise (AWGN) channel.

Context-aware visual exploration of molecular databases

Facilitating the visual exploration of scientific data has received increasing attention in the past decade or so. Especially in life science related application areas the amount of available data has grown at a breath taking pace. In this paper we describe an approach that allows for visual inspection of large collections of molecular compounds. In contrast to classical visualizations of such spaces we incorporate a specific focus of analysis, for example the outcome of a biological experiment such as high throughout screening results. The presented method uses this experimental data to select molecular fragments of the underlying molecules that have interesting properties and uses the resulting space to generate a two dimensional map based on a singular value decomposition algorithm and a self organizing map. Experiments on real datasets show that the resulting visual landscape groups molecules of similar chemical properties in densely connected regions.

Very large inverse problems in atmosphere and ocean modelling

For the very large nonlinear dynamical systems that arise in a wide range of physical, biological and environmental problems, the data needed to initialize a numerical forecasting model are seldom available. To generate accurate estimates of the expected states of the system, both current and future, the technique of ‘data assimilation’ is used to combine the numerical model predictions with observations of the system measured over time. Assimilation of data is an inverse problem that for very large-scale systems is generally ill-posed. In four-dimensional variational assimilation schemes, the dynamical model equations provide constraints that act to spread information into data sparse regions, enabling the state of the system to be reconstructed accurately. The mechanism for this is not well understood. Singular value decomposition techniques are applied here to the observability matrix of the system in order to analyse the critical features in this process. Simplified models are used to demonstrate how information is propagated from observed regions into unobserved areas. The impact of the size of the observational noise and the temporal position of the observations is examined. The best signal-to-noise ratio needed to extract the most information from the observations is estimated using Tikhonov regularization theory. Copyright © 2005 John Wiley & Sons, Ltd.

Modified radial basis function neural network using output transformation

A modified radial basis function (RBF) neural network and its identification algorithm based on observational data with heterogeneous noise are introduced. The transformed system output of Box-Cox is represented by the RBF neural network. To identify the model from observational data, the singular value decomposition of the full regression matrix consisting of basis functions formed by system input data is initially carried out and a new fast identification method is then developed using Gauss-Newton algorithm to derive the required Box-Cox transformation, based on a maximum likelihood estimator (MLE) for a model base spanned by the largest eigenvectors. Finally, the Box-Cox transformation-based RBF neural network, with good generalisation and sparsity, is identified based on the derived optimal Box-Cox transformation and an orthogonal forward regression algorithm using a pseudo-PRESS statistic to select a sparse RBF model with good generalisation. The proposed algorithm and its efficacy are demonstrated with numerical examples.

The system identification and control of Hammerstein system using non-uniform rational B-spline neural network and particle swarm optimization

In this paper a new system identification algorithm is introduced for Hammerstein systems based on observational input/output data. The nonlinear static function in the Hammerstein system is modelled using a non-uniform rational B-spline (NURB) neural network. The proposed system identification algorithm for this NURB network based Hammerstein system consists of two successive stages. First the shaping parameters in NURB network are estimated using a particle swarm optimization (PSO) procedure. Then the remaining parameters are estimated by the method of the singular value decomposition (SVD). Numerical examples including a model based controller are utilized to demonstrate the efficacy of the proposed approach. The controller consists of computing the inverse of the nonlinear static function approximated by NURB network, followed by a linear pole assignment controller.

PSO assisted NURB Neural Network Identification

A system identification algorithm is introduced for Hammerstein systems that are modelled using a non-uniform rational B-spline (NURB) neural network. The proposed algorithm consists of two successive stages. First the shaping parameters in NURB network are estimated using a particle swarm optimization (PSO) procedure. Then the remaining parameters are estimated by the method of the singular value decomposition (SVD). Numerical examples are utilized to demonstrate the efficacy of the proposed approach.

Elastic-net prefiltering for two-class classification

A two-stage linear-in-the-parameter model construction algorithm is proposed aimed at noisy two-class classification problems. The purpose of the first stage is to produce a prefiltered signal that is used as the desired output for the second stage which constructs a sparse linear-in-the-parameter classifier. The prefiltering stage is a two-level process aimed at maximizing a model's generalization capability, in which a new elastic-net model identification algorithm using singular value decomposition is employed at the lower level, and then, two regularization parameters are optimized using a particle-swarm-optimization algorithm at the upper level by minimizing the leave-one-out (LOO) misclassification rate. It is shown that the LOO misclassification rate based on the resultant prefiltered signal can be analytically computed without splitting the data set, and the associated computational cost is minimal due to orthogonality. The second stage of sparse classifier construction is based on orthogonal forward regression with the D-optimality algorithm. Extensive simulations of this approach for noisy data sets illustrate the competitiveness of this approach to classification of noisy data problems.

Particle swarm optimisation assisted classification using elastic net prefiltering

A novel two-stage construction algorithm for linear-in-the-parameters classifier is proposed, aiming at noisy two-class classification problems. The purpose of the first stage is to produce a prefiltered signal that is used as the desired output for the second stage to construct a sparse linear-in-the-parameters classifier. For the first stage learning of generating the prefiltered signal, a two-level algorithm is introduced to maximise the model's generalisation capability, in which an elastic net model identification algorithm using singular value decomposition is employed at the lower level while the two regularisation parameters are selected by maximising the Bayesian evidence using a particle swarm optimization algorithm. Analysis is provided to demonstrate how “Occam's razor” is embodied in this approach. The second stage of sparse classifier construction is based on an orthogonal forward regression with the D-optimality algorithm. Extensive experimental results demonstrate that the proposed approach is effective and yields competitive results for noisy data sets.

Determinantes do fluxo de fundos de investimento no Brasil

Nesse trabalho, procuramos identificar fatores sistemáticos que expliquem uma variação significativa nos fluxos destinados às diversas categorias de fundos de investimento brasileiros, a partir de análises de uma amostra de dados agregados de captações e resgates nesses produtos. O estudo buscou avaliar a existência de padrões de comportamento comuns aos investidores de fundos locais através da análise da migração de fluxos entre as diversas classes de fundos. Foram inicialmente tratados os fatores não comportamentais conhecidos que impactam o fluxo dos fundos, a variável dependente. Esses fatores conhecidos foram apurados através de uma revisão dos trabalhos acadêmicos dos mercados internacional e local. Após esse tratamento foi aplicado o método de decomposição de valores singulares (SVD - Singular Value Decomposition), com o objetivo de avaliarmos os efeitos comportamentais agrupados dos investidores. A decomposição em valores singulares sugere como principais fatores comuns comportamentos de entrada e saída de fundos em massa e migrações entre as classes de fundos de menor e as de maior risco, o que Baker e Wurgler (2007) chamaram de demanda especulativa, e que, segundo esses e outros autores pesquisados, poderia ser interpretada como uma proxy do sentimento dos investidores. Guercio e Tkac (2002) e Edelen et al. (2010), encontraram em suas pesquisas evidências da diferença de comportamento entre investidores de atacado e de varejo, o que foi detectado para a classes de fundos de Renda Variável no caso do presente estudo sobre o mercado brasileiro. O entendimento das variações na tolerância a risco dos investidores de fundos de investimento pode auxiliar na oferta de produtos mais compatíveis com a demanda. Isso permitiria projetar captações para os produtos com base nas características dessa oferta, o que também desenvolvemos nessa pesquisa para o caso das categorias de fundos Multimercado e Renda variável, através de um modelo de espaço de estados com sazonalidade determinística e inicialização SVD. O modelo proposto nesse trabalho parece ter conseguido capturar, na amostra avaliada (2005-2008), um comportamento que se manteve fora da amostra (2009-2011), validando, ao menos na amostra considerada, a proposta de extração dos componentes principais agregados do comportamento dos investidores de fundos brasileiros.

Filtro kl/dvs para atenuação do ruído "ground roll"

In geophysics there are several steps in the study of the Earth, one of them is the processing of seismic records. These records are obtained through observations made on the earth surface and are useful for information about the structure and composition of the inaccessible parts in great depths. Most of the tools and techniques developed for such studies has been applied in academic projects. The big problem is that the seismic processing power unwanted, recorded by receivers that do not bring any kind of information related to the reflectors can mask the information and/or generate erroneous information from the subsurface. This energy is known as unwanted seismic noise. To reduce the noise and improve a signal indicating a reflection, without losing desirable signals is sometimes a problem of difficult solution. The project aims to get rid of the ground roll noise, which shows a pattern characterized by low frequency, low rate of decay, low velocity and high amplituds. The Karhunen-Loève Transform is a great tool for identification of patterns based on the eigenvalues and eigenvectors. Together with the Karhunen-Loève Transform we will be using the Singular Value Decomposition, since it is a great mathematical technique for manipulating data

Parâmetro de regularização em problemas inversos: estudo numérico com a transformada de Radon

In general, an inverse problem corresponds to find a value of an element x in a suitable vector space, given a vector y measuring it, in some sense. When we discretize the problem, it usually boils down to solve an equation system f(x) = y, where f : U Rm ! Rn represents the step function in any domain U of the appropriate Rm. As a general rule, we arrive to an ill-posed problem. The resolution of inverse problems has been widely researched along the last decades, because many problems in science and industry consist in determining unknowns that we try to know, by observing its effects under certain indirect measures. Our general subject of this dissertation is the choice of Tykhonov´s regulaziration parameter of a poorly conditioned linear problem, as we are going to discuss on chapter 1 of this dissertation, focusing on the three most popular methods in nowadays literature of the area. Our more specific focus in this dissertation consists in the simulations reported on chapter 2, aiming to compare the performance of the three methods in the recuperation of images measured with the Radon transform, perturbed by the addition of gaussian i.i.d. noise. We choosed a difference operator as regularizer of the problem. The contribution we try to make, in this dissertation, mainly consists on the discussion of numerical simulations we execute, as is exposed in Chapter 2. We understand that the meaning of this dissertation lays much more on the questions which it raises than on saying something definitive about the subject. Partly, for beeing based on numerical experiments with no new mathematical results associated to it, partly for being about numerical experiments made with a single operator. On the other hand, we got some observations which seemed to us interesting on the simulations performed, considered the literature of the area. In special, we highlight observations we resume, at the conclusion of this work, about the different vocations of methods like GCV and L-curve and, also, about the optimal parameters tendency observed in the L-curve method of grouping themselves in a small gap, strongly correlated with the behavior of the generalized singular value decomposition curve of the involved operators, under reasonably broad regularity conditions in the images to be recovered

Identificação de parâmetros modais utilizando apenas as respostas da estrutura: identificação estocástica de subespaço e decomposição no domínio da frequência

Pós-graduação em Engenharia Mecânica - FEIS

Identificação dos parâmetros modais utilizando apenas as respostas da estrutura: identificação no domínio do tempo

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Análise de resolução em tomografia interpoços em 3-D

A aplicação de métodos de inversão aos dados de múltiplos levantamentos sísmicos interpoços permite a reconstrução de modelos de vagarosidade em 3-D de alta resolução adequados para monitoramento de processos de recuperação avançada de petróleo e caracterização de reservatórios. Entretanto, a falta de cobertura volumétrica uniforme de raios de levantamentos interpoços exige informação adicional ao sistema tomográfico para obtenção de soluções estáveis. A discretização do modelo em uma malha 3-D com células prismáticas triangulares e a decomposição em valores singulares são utilizadas para avaliar a reconstrução tomográfica em 3-D. O ângulo da projeção de modelos-alvo no subespaço ortogonal ao espaço nulo efetivo da matriz tomográfica é um critério adequado para se otimizar a malha de discretização do modelo interpretativo e a geometria de aquisição dos dados de modo a melhorar o condicionamento da reconstrução tomográfica. Esta abordagem pode ser utilizada durante as iterações lineares para redefinir a malha ou avaliar a necessidade de informação a priori adicional ao sistema tomográfico.

Decomposição em valores singulares aplicada a dados de campo magnético

Neste trabalho, a decomposição em valores singulares (DVS) de uma matriz A, n x m, que representa a anomalia magnética, é vista como um método de filtragem bidimensional de coerência que separa informações correlacionáveis e não correlacionáveis contidas na matriz de dados magnéticos A. O filtro DVS é definido através da expansão da matriz A em autoimagens e valores singulares. Cada autoimagem é dada pelo produto escalar dos vetores de base, autovetores, associados aos problemas de autovalor e autovetor das matrizes de covariância A^TA e AA^T. Este método de filtragem se baseia no fato de que as autoimagens associadas a grandes valores singulares concentram a maior parte da informação correlacionável presente nos dados, enquanto que a parte não correlacionada, presumidamente constituída de ruídos causados por fontes magnéticas externas, ruídos introduzidos pelo processo de medida, estão concentrados nas autoimagens restantes. Utilizamos este método em diferentes exemplos de dados magnéticos sintéticos. Posteriormente, o método foi aplicado a dados do aerolevantamento feito pela PETROBRÁS no Projeto Carauari-Norte (Bacia do Solimões), para analisarmos a potencialidade deste na identificação, eliminação ou atenuação de ruídos e como um possível método de realçar feições particulares da anomalia geradas por fontes profundas e rasas. Este trabalho apresenta também a possibilidade de introduzir um deslocamento estático ou dinâmico nos perfis magnéticos, com a finalidade de aumentar a correlação (coerência) entre eles, permitindo assim concentrar o máximo possível do sinal correlacionável nas poucas primeiras autoimagens. Outro aspecto muito importante desta expansão da matriz de dados em autoimagens e valores singulares foi o de mostrar, sob o ponto de vista computacional, que a armazenagem dos dados contidos na matriz, que exige uma quantidade n x m de endereços de memória, pode ser diminuída consideravelmente utilizando p autoimagens. Assim o número de endereços de memória cai para p x (n + m + 1), sem alterar a anomalia, na reprodução praticamente perfeita. Dessa forma, concluímos que uma escolha apropriada do número e dos índices das autoimagens usadas na decomposição mostra potencialidade do método no processamento de dados magnéticos.