We consider the relationship between modelling dynamics with a nonlinear function approximation and the application of a noise reduction algorithm. Filtering the data with such an algorithm is then shown to provide a ...
详细信息
ISBN:
(纸本)0819416207
We consider the relationship between modelling dynamics with a nonlinear function approximation and the application of a noise reduction algorithm. Filtering the data with such an algorithm is then shown to provide a better deterministic model for the data than an ordinary least squares estimate.
A two-sided (or complete) orthogonal decomposition of an m × n matrix A is a product of an orthogonal matrix, a triangular matrix, and another orthogonal matrix. Two examples are the URV and ULV decompositions. I...
详细信息
ISBN:
(纸本)0819416207
A two-sided (or complete) orthogonal decomposition of an m × n matrix A is a product of an orthogonal matrix, a triangular matrix, and another orthogonal matrix. Two examples are the URV and ULV decompositions. In this paper we present and analyze URV and ULV algorithms that are efficient whenever the numerical rank k of the matrix is much less than min(m,n). We also prove that good estimates of the singular vectors, needed in the algorithms, lead to good approximations of the singular subspaces of A.
A number of adaptive condition number estimators have been proposed in the past to dynamically estimate the sensitivity of the coefficient matrix of a linear systems of equations. Applications of these techniques ofte...
详细信息
ISBN:
(纸本)0819416207
A number of adaptive condition number estimators have been proposed in the past to dynamically estimate the sensitivity of the coefficient matrix of a linear systems of equations. Applications of these techniques often arise in the context of signalprocessing, where the information matrix is being updated with rank-one modifications. Various schemes, such as ACE, ALE and ICE, were proposed to cope with this problem. In this paper, we will briefly review the past work, and show how the small-sample condition estimator can be used in an adaptive manner.
The concern here is of retrieving damped harmonics and polynomial phase signals in the presence of additive noise. The damping function is not limited to the exponential model, and in certain cases, the additive noise...
详细信息
ISBN:
(纸本)0819416207
The concern here is of retrieving damped harmonics and polynomial phase signals in the presence of additive noise. The damping function is not limited to the exponential model, and in certain cases, the additive noise does not have to be white. Three classes of algorithms are presented, namely DFT based, Kumaresan-Tufts type extensions, and subspace variants including the MUSIC algorithm. Preference should be based on the available data length and frequency separations. In addition, retrieval of self coupled damped harmonics, which may be present when nonlinearities exist in physical systems, is investigated. Simulation examples illustrate main points of the paper.
The Rapid Prototyping of Application Specific signal Processors (RASSP) program has as its goal the dramatic improvement of the process by which signal processors are specified, designed, documented, manufactured, and...
详细信息
ISBN:
(纸本)0819416207
The Rapid Prototyping of Application Specific signal Processors (RASSP) program has as its goal the dramatic improvement of the process by which signal processors are specified, designed, documented, manufactured, and supported. A key part of the RASSP process is the method of architecture selection. An architecture overview is presented, together with a discussion of architecture classification, architecture evaluation factors, and signalprocessing flow. Two important architecture support aspects are briefly discussed: open systems architecture and integrated diagnostics. Recommendations for preferred approaches and rules of thumb for architecture selection are given. Finally, the RASSP architecture selection process is outlined, including a discussion of a standard hardware/software module interface.
We present an algorithm-based fault tolerant scheme for QR decomposition, that extends the well-known Gentleman-Kung-McWhirter triangular systolic array architecture. Assuming that the array is subject to transient fa...
详细信息
ISBN:
(纸本)0819416207
We present an algorithm-based fault tolerant scheme for QR decomposition, that extends the well-known Gentleman-Kung-McWhirter triangular systolic array architecture. Assuming that the array is subject to transient faults, widely separated in time and each affecting a single processor, we give an algorithm that corrects the full triangular array with computational overhead equivalent, on average, to the interpolation of a single extra vector into the data stream.
A massively parallel signal and image processing architecture is considered. The architecture is comprised of 2D arrays of cells that simulate the response of retina neurons. The results of simulations are compared to...
详细信息
ISBN:
(纸本)0819416207
A massively parallel signal and image processing architecture is considered. The architecture is comprised of 2D arrays of cells that simulate the response of retina neurons. The results of simulations are compared to previously published experimental results and the system is applied to detection of spatio-temporal features in sequences of images representative of pulse- doppler radar images. By arranging the output layer so that the cells respond to various key input features an array of feature extraction cells can be obtained. The system is characterized by developing an image space to feature space mapping.
Much research has focused on estimating the parameters of a sum of exponentially damped sinusoids. In some applications, such as nuclear magnetic resonance signalprocessing, only a few of the sinusoids are of interes...
详细信息
ISBN:
(纸本)0819416207
Much research has focused on estimating the parameters of a sum of exponentially damped sinusoids. In some applications, such as nuclear magnetic resonance signalprocessing, only a few of the sinusoids are of interest. This paper presents some new frequency-selective techniques for estimating the parameters of sinusoids within a specified frequency region using a subspace and SVD-based estimation algorithm and an FIR filter matrix. The applicable estimation algorithms in this technique are the linear prediction method, the matrix pencil method, Kung et al.'s method and its total-least-squares variant, called the HTLS method. The benefits of the frequency-selective technique combined with the HTLS estimation method are confirmed through simulations.
We propose a scale-limited signal model based on wavelet representation and study the reconstructability of scale-limited signals via extrapolation in this research. In analogy with the band-limited case, we define a ...
详细信息
ISBN:
(纸本)0819416207
We propose a scale-limited signal model based on wavelet representation and study the reconstructability of scale-limited signals via extrapolation in this research. In analogy with the band-limited case, we define a scale-limited time-concentrated operator, and examine various vector spaces associated with such an operator. It is proved that the scale-limited signal space can be decomposed into the direct sum of two subspaces and only the component in one subspace can be exactly reconstructed, where the reconstructable subspace can be interpreted as a space consisting of scale/time-limited signals. Due to the ill-posedness of scale-limited extrapolation, a regularization process is introduced for noisy data extrapolation.
In this paper, a highly scalable hardware accelerator design for digital signalprocessing is presented. The key features of this accelerator are minimum I/O operations, highly scalable massive parallelism, easy progr...
详细信息
ISBN:
(纸本)0819416207
In this paper, a highly scalable hardware accelerator design for digital signalprocessing is presented. The key features of this accelerator are minimum I/O operations, highly scalable massive parallelism, easy programming, and modularity and regularity. With a very large register file (> 1000) per processing element, the reuse factor per datum in this accelerator can be increased significantly (as compared to traditional DSP architectures). System performance is improved because the amount of data transfer between the on-chip cache and the off-chip cache/memory is reduced by the same factor. Since the basic building block of this accelerator is simply a VLSI chip with several processing elements, scalable massive parallelism can be achieved by connecting multiple chips together in a SIMD `vector- like' fashion. Finally, programming of this accelerometer is not difficult because it is operated under the SIMD `vector-like' mode. With the expected VLSI technology in the next few years, the throughput of one single accelerator chip can approach GFLOPs performance. Hence, the high computing power needed by digital signalprocessing applications can be provided by just connecting a small number of this chip together.
暂无评论