Smart-phone based electrocardiogram wavelet decomposition and neural network classification

This paper discusses ECG classification after parametrizing the ECG waveforms in the wavelet domain. The aim of the work is to develop an accurate classification algorithm that can be used to diagnose cardiac beat abnormalities detected using a mobile platform such as smart-phones. Continuous time recurrent neural network classifiers are considered for this task. Records from the European ST-T Database are decomposed in the wavelet domain using discrete wavelet transform (DWT) filter banks and the resulting DWT coefficients are filtered and used as inputs for training the neural network classifier. Advantages of the proposed methodology are the reduced memory requirement for the signals which is of relevance to mobile applications as well as an improvement in the ability of the neural network in its generalization ability due to the more parsimonious representation of the signal to its inputs.

On the application of optimal wavelet filter banks for ECG signal classification

This paper discusses ECG signal classification after parametrizing the ECG waveforms in the wavelet domain. Signal decomposition using perfect reconstruction quadrature mirror filter banks can provide a very parsimonious representation of ECG signals. In the current work, the filter parameters are adjusted by a numerical optimization algorithm in order to minimize a cost function associated to the filter cut-off sharpness. The goal consists of achieving a better compromise between frequency selectivity and time resolution at each decomposition level than standard orthogonal filter banks such as those of the Daubechies and Coiflet families. Our aim is to optimally decompose the signals in the wavelet domain so that they can be subsequently used as inputs for training to a neural network classifier.

Application of two-dimensional wavelet transform in near-shore X-band radar images

Among existing remote sensing applications, land-based X-band radar is an effective technique to monitor the wave fields, and spatial wave information could be obtained from the radar images. Two-dimensional Fourier Transform (2-D FT) is the common algorithm to derive the spectra of radar images. However, the wave field in the nearshore area is highly non-homogeneous due to wave refraction, shoaling, and other coastal mechanisms. When applied in nearshore radar images, 2-D FT would lead to ambiguity of wave characteristics in wave number domain. In this article, we introduce two-dimensional Wavelet Transform (2-D WT) to capture the non-homogeneity of wave fields from nearshore radar images. The results show that wave number spectra by 2-D WT at six parallel space locations in the given image clearly present the shoaling of nearshore waves. Wave number of the peak wave energy is increasing along the inshore direction, and dominant direction of the spectra changes from South South West (SSW) to West South West (WSW). To verify the results of 2-D WT, wave shoaling in radar images is calculated based on dispersion relation. The theoretical calculation results agree with the results of 2-D WT on the whole. The encouraging performance of 2-D WT indicates its strong capability of revealing the non-homogeneity of wave fields in nearshore X-band radar images.

VELOCITY FIELD OF COMPRESSIBLE MAGNETOHYDRODYNAMIC TURBULENCE: WAVELET DECOMPOSITION AND MODE SCALINGS

We study compressible magnetohydrodynamic turbulence, which holds the key to many astrophysical processes, including star formation and cosmic-ray propagation. To account for the variations of the magnetic field in the strongly turbulent fluid, we use wavelet decomposition of the turbulent velocity field into Alfven, slow, and fast modes, which presents an extension of the Cho & Lazarian decomposition approach based on Fourier transforms. The wavelets allow us to follow the variations of the local direction of the magnetic field and therefore improve the quality of the decomposition compared to the Fourier transforms, which are done in the mean field reference frame. For each resulting component, we calculate the spectra and two-point statistics such as longitudinal and transverse structure functions as well as higher order intermittency statistics. In addition, we perform a Helmholtz-Hodge decomposition of the velocity field into incompressible and compressible parts and analyze these components. We find that the turbulence intermittency is different for different components, and we show that the intermittency statistics depend on whether the phenomenon was studied in the global reference frame related to the mean magnetic field or in the frame defined by the local magnetic field. The dependencies of the measures we obtained are different for different components of the velocity; for instance, we show that while the Alfven mode intermittency changes marginally with the Mach number, the intermittency of the fast mode is substantially affected by the change.

A WAVELET-BASED SPEAKER VERIFICATION ALGORITHM

This paper presents a study on wavelets and their characteristics for the specific purpose of serving as a feature extraction tool for speaker verification (SV), considering a Radial Basis Function (RBF) classifier, which is a particular type of Artificial Neural Network (ANN). Examining characteristics such as support-size, frequency and phase responses, amongst others, we show how Discrete Wavelet Transforms (DWTs), particularly the ones which derive from Finite Impulse Response (FIR) filters, can be used to extract important features from a speech signal which are useful for SV. Lastly, an SV algorithm based on the concepts presented is described.

Improved voice activity detection algorithm using wavelet and support vector machine

This paper proposes an improved voice activity detection (VAD) algorithm using wavelet and support vector machine (SVM) for European Telecommunication Standards Institution (ETS1) adaptive multi-rate (AMR) narrow-band (NB) and wide-band (WB) speech codecs. First, based on the wavelet transform, the original IIR filter bank and pitch/tone detector are implemented, respectively, via the wavelet filter bank and the wavelet-based pitch/tone detection algorithm. The wavelet filter bank can divide input speech signal into several frequency bands so that the signal power level at each sub-band can be calculated. In addition, the background noise level can be estimated in each sub-band by using the wavelet de-noising method. The wavelet filter bank is also derived to detect correlated complex signals like music. Then the proposed algorithm can apply SVM to train an optimized non-linear VAD decision rule involving the sub-band power, noise level, pitch period, tone flag, and complex signals warning flag of input speech signals. By the use of the trained SVM, the proposed VAD algorithm can produce more accurate detection results. Various experimental results carried out from the Aurora speech database with different noise conditions show that the proposed algorithm gives considerable VAD performances superior to the AMR-NB VAD Options 1 and 2, and AMR-WB VAD. (C) 2009 Elsevier Ltd. All rights reserved.

A note on a practical relationship between filter coefficients and scaling and wavelet functions of Discrete Wavelet Transforms

In this paper, the relationship between the filter coefficients and the scaling and wavelet functions of the Discrete Wavelet Transform is presented and exemplified from a practical point-of-view. The explanations complement the wavelet theory, that is well documented in the literature, being important for researchers who work with this tool for time-frequency analysis. (c) 2011 Elsevier Ltd. All rights reserved.

Wavelet-based dynamic time warping

Dynamic Time Warping (DTW), a pattern matching technique traditionally used for restricted vocabulary speech recognition, is based on a temporal alignment of the input signal with the template models. The principal drawback of DTW is its high computational cost as the lengths of the signals increase. This paper shows extended results over our previously published conference paper, which introduces an optimized version of the DTW I hat is based on the Discrete Wavelet Transform (DWT). (C) 2008 Elsevier B.V. All rights reserved.

Wavelet-based cepstrum calculation

In this paper we present a new wavelet-based algorithm for low-cost computation of the cepstrum. It can be used for real time precise pitch determination in automatic speech and speaker recognition systems. Many wavelet families are examined to determine the one that works best. The results confirm the efficacy and accuracy of the proposed technique for pitch extraction. (C) 2008 Elsevier B.V. All rights reserved.

Wavelet regression with correlated errors on a piecewise Holder class

This paper generalizes the methodology of Cat and Brown [Cai, T., Brown, L.D., 1998. Wavelet shrinkage for nonequispaced samples. The Annals of Statistics 26, 1783-1799] for wavelet shrinkage for nonequispaced samples, but in the presence of correlated stationary Gaussian errors. If the true function is a member of a piecewise Holder class, it is shown that, even for long memory errors, the rate of convergence of the procedure is almost-minimax relative to the independent and identically distributed errors case. (c) 2008 Elsevier B.V. All rights reserved.

Computational aspects of harmonic wavelet Galerkin methods and an application to a precipitation front propagation model

This article is dedicated to harmonic wavelet Galerkin methods for the solution of partial differential equations. Several variants of the method are proposed and analyzed, using the Burgers equation as a test model. The computational complexity can be reduced when the localization properties of the wavelets and restricted interactions between different scales are exploited. The resulting variants of the method have computational complexities ranging from O(N(3)) to O(N) (N being the space dimension) per time step. A pseudo-spectral wavelet scheme is also described and compared to the methods based on connection coefficients. The harmonic wavelet Galerkin scheme is applied to a nonlinear model for the propagation of precipitation fronts, with the front locations being exposed in the sizes of the localized wavelet coefficients. (C) 2011 Elsevier Ltd. All rights reserved.

Aplicação da transformada Wavelet Packet na análise e classificação de sinais de vozes patológicas

o exame para o diagnóstico de doenças da laringe é usualmente realizado através da videolaringoscopia e videoestroboscopia. A maioria das doenças na laringe provoca mudanças na voz do paciente. Diversos índices têm sido propostos para avaliar quantitativamente a qualidade da voz. Também foram propostos vários métodos para classificação automática de patologias da laringe utilizando apenas a voz do paciente. Este trabalho apresenta a aplicação da Transformada Wavelet Packet e do algoritmo Best Basis [COI92] para a classificação automática de vozes em patológicas ou normais. Os resultados obtidos mostraram que é possível classificar a voz utilizando esta Transformada. Tem-se como principal conclusão que um classificador linear pode ser obtido ao se empregar a Transformada Wavelet Packet como extrator de características. O classificador é linear baseado na existência ou não de nós na decomposição da Transformada Wavelet Packet. A função Wavelet que apresentou os melhores resultados foi a sym1et5 e a melhor função custo foi a entropia. Este classificador linear separa vozes normais de vozes patológicas com um erro de classificação de 23,07% para falsos positivos e de 14,58%para falsos negativos.

Estudo comparativo da transformada Karhunen-Loève na compressão de imagens

A representação de funções através da utilização de bases (KERNEL) de representação tem sido fundamental no processamento digital de sinais. A Transformada KARHUNEN-LOÈVE (KLT), também conhecida como Transformada HOTELLING, permite a representação de funções utilizando funções-base formadas pelos autovetores da matriz de correlação do sinal considerado. Nesse aspecto essa transformada fornece uma base ótima, isto é, aquela que proporciona o menor valor de Erro Quadrático Médio entre o sinal reconstruído e o original, para um determinado número de coeficientes. A dificuldade na utilização da KLT está no tempo adicional para calcular os autovetores (base) da matriz de correlação, o que muitas vezes inviabiliza a sua utilização nas aplicações em tempo real. Em muitas aplicações a KLT é utilizada em conjunto com outras transformadas melhorando os resultados destas aplicações. Sendo considerada a transformada ótima no sentido do Erro Quadrático Médio, este trabalho apresenta um estudo da Transformada KARHUNEN-LOÈVE nas aplicações de compressão de imagens bidimensionais estáticas e em tons de cinza, realizando também a comparação desta técnica com outras técnicas (DCT e WAVELET) buscando avaliar os pontos fortes e fracos da utilização da KLT para este tipo de aplicação. Duas técnicas importantes para solucionar o problema de cálculo dos autovalores e autovetores da matriz de correlação (Método de JACOBI e Método QL) são também apresentadas neste trabalho. Os resultados são comparados utilizando a Razão Sinal/Ruído de Pico (PSNR), a Razão de Compressão (CR) e os tempos de processamento (em segundos) para geração dos arquivos compactados.

Uma ferramenta para geração de filtros FIR paralelos otimizados com coeficientes constantes

Esta dissertação trata da elaboração de uma ferrramenta para a geração de filtros FIR otimizados paralelos com coeficientes constantes. A ferramenta desenvolvida é capaz de gerar uma descrição VHDL de um filtro FIR paralelo com coeficientes constantes a partir das especificações do filtro. São exploradas técnicas de otimização de coeficientes e de otimização arquitetural. As técnicas empregadas são baseadas no uso de representações ternárias e redução do número de digitos não-zero dos coeficientes, uso de fatores de escala e eliminação de sub-expressões comuns. No texto, uma breve introdução sobre os filtros digitais é apresentada seguida por uma série de trabalhos encontrados na literatura relacionados às técnicas mencionadas e que são apresentados como base para o desenvolvimento da ferramenta implementada nesta dissertação. O funcionamento da ferramenta é detalhado tanto nos seus aspectos de algoritmo quanto em nível de implementação. São apresentados resultados de síntese em alguns de filtros hipotéticos projetados utilizando a ferramenta desenvolvida. Uma análise detalhada dos resultados obtidos é realizada. Os apêndices deste trabalho apresentam o código fonte da ferramenta de síntese de filtros desenvolvida.