Aníbal R. Figueiras-Vidal

An adaptive beamforming technique based on cyclostationary signal properties

A new approach to adaptive beamforming is presented. The method is based on the property of cyclostationary signals to generate spectral lines when they pass through certain nonlinear transformations. The beamformer coefficients are selected according to a new optimization objective, which consists on minimizing the mean square error between the array output after the nonlinearity and a complex exponential. This approach optimally extracts any signal that generates a spectral line at the same frequency as the reference complex exponential. A gradient-based algorithm is derived to compute the optimum weights. Since the proposed cost function is a nonconvex function of the array coefficients, minima are analyzed for the three most common types of perturbations found in communications: Gaussian noise, multiple interferences, and multipath propagation. It is demonstrated via analysis and simulations that minima correspond to points where output noise power is minimized, interferences are canceled, and intersymbol interference is removed, i.e., the beamformer eliminates the distortion introduced by the radiocommunication channel. >

Adaptive iterative algorithms for spiky deconvolution

Considering the joint detection-estimation character that spiky deconvolution problems have, an adaptively contracted (projection) selection operator is introduced to detect the nonzero values of the solution, which can be combined with iterative algorithms to offer very efficient schemes for solving these problems. A number of gradient-type algorithms based on this principle are described, and their performance is illustrated through simulation examples. The approach is based on the idea of reducing the noise and defining the signal in an iterative form. Another possibility is to define the signal and reduce the overall energy outside its domain, also in an iterative form. This algorithm is more expensive from a computational point of view; however, simulations indicate that it has somewhat different properties. It seems to be slightly less robust against the noise, while it offers better resolution and even more accurate amplitude estimates. >

Automatic modelling and simulation of daily global solar radiation series

Abstract This paper presents a new model to generate simulated daily global solar radiation (DGSR) sequences. A DGSR value is the product of two factors: a seasonal low-frequency component, principally due to the suns periodic movement, and a random component due to rapid fluctuations of the atmospheric environment. Methods are provided to automatically separate and estimate both components from the available (generally short) DGSR records. Hence, other series describing the sun-position evolution or any kind of atmospheric conditions are not necessary. An illustrative example is included which results in a good global agreement between simulated and original sequences.

Recurrent radial basis function networks for optimal blind equalization

A recurrent version of a radial basis function (RBF) network can compute optimal symbol-by-symbol decisions for equalizing Gaussian channels in digital communication systems, but the (linear or not) channel response and the noise variance must be known. Starting from theoretical considerations, a novel technique for learning the channel parameters in a non-supervised, non-decision directed way is proposed. This technique provides a simple and fast algorithm that can be used for tracking in time variant environments or for blind equalization purposes.<<ETX>>

Pao networks for data transmission equalization

The authors demonstrate the advantage of using a Pao network in symbol by symbol data equalization. It is a simple and fast scheme that can be used even for nonlinear channel cases. One of the open problems is the selection of the order of the inputs. A cascade-correlation scheme offers the possibility of accomplishing this task in a particularly efficient manner. It is necessary to extend the obtained results to practical modulation constellations and models of transmission channels, especially to nonlinear models. Computer simulation is used to evaluate the advantages of the particular data transmission scheme.<<ETX>>

Channel Equalization with Neural Networks

Recently, several authors have explored the application of Neural Networks to compensate the channel effects in digital communication systems, with the goal of reducing the limitations of the conventional schemes: the suboptimal performance of the Linear Equalizer (LE) and the Decision Feedback Equalizer (DFE), or the complexity and the model dependence of Viterbi-based detectors.

Digital Equalization Using Modular Neural Networks: an Overview

Neural networks have a high potential to improve the performance of linear equalizers, but they are limited by a series of practical difficulties: the design of the adequate architecture, the application of algorithms for a fast training, and the selection of an appropriate objective function are the most important.

Competitive local linear modeling

This paper describes a novel approach for nonlinear signal modeling and prediction. We propose a nonlinear extension of the conventional AR model by using several linear models, each covering a subset of the whole signal data. Considering that AR modeling is conducted by associating a delay vector with the future value to be predicted, we interpret the signal as a codebook of patterns having the desired predicted value as the associated output. The pattern groups associated with each AR model are obtained by competition among the linear models as they are trained: the competition gives us the AR models as well as a classification of the data patterns. The obtained data groups are the basis for estimating a segmentation model, i.e., a classifier that, given a new data pattern, indicates what linear model should be used. The combination of the classifier and the linear models constitutes the final system. Several practical applications show the advantages of our approach.

Sampling Bandpass Signals

In a recent paper [1], Brown examined the sampling of a real finiteenergy bandpass signal having an (angular) bandwidth ¿ (in radians per second) at the theoretically minimum (average) rate of ¿/¿ samples per second. Following Grace and Pitts [2] quadrature sampling, a particular case of Kohlenbergs second-order sampling [3, 4], Brown has proved the feasibility of a separate interpolation of the in-phase and quadrature components of the signal when ¿o = k¿/2 (Browns condition), where ¿o is the center (angular) frequency of the signal and k is an arbitrary positive integer. Here the problem is reconsidered from a general point of view, introducing, under Browns condition, an interpolation formula which includes that of Grace and Pitt and which extends a theorem of Populis [5-7]. We indicate the necessary and sufficient conditions to obtain a separate interpolation, offering closed formulas to obtain the interpolation functions. We also discuss the minimum oversampling rate needed whenBrowns condition is not verified.

A linear transform for spectral estimation

The authors propose the introduction of a previous linear transform for rational spectral estimation by using a linear predictor. The method is based on works of Oppenheim [1], [2] on discrete signal representation. The connection with zero-pole (ARMA) spectral estimators is discussed, and the main potential advantages of the approach are indicated: reduction in estimator order, low implementation complexity, and the possibility of increasing frequency resolution. The use of this known linear transform seems to be useful in cases in which the presence of a strong periodicity in the signal to be analyzed is a priori known. Some promising results obtained in processing signals from hydrographic measurements are shown.