HomeChair: Ananthram Swami, Army Research Laboratory, USA
Home
Stephen Bates, Massana, Dublin (Ireland)
Steve McLaughlin, University of Edinburgh, Scotland (U.K.)
In this paper the effective bandwidths of stable distributions are studied. Effective bandwidths are being heavily promoted as the most appropriate method for call admission control (CAC) and resource allocation within ATM networks. Recent work in teletraffic modelling has suggested that models based on stable distributions provide an efficient mechanism forcapturing the long range dependence and infinite variance associated with teletraffic data (the Joseph and Noah effects.This has potentially serious implications for effective bandwidths and we show how the effective bandwidth of such data is theoretically infinite. We then present two approximate methods for estimating the effective bandwidth of data based on stable distribution.
ic981284.pdf (From Postscript)
TOPLuc Pronzato, CNRS (France)
We consider the situation where n items have to be selected among a series of N presented sequentially, the information contained in each item being random. The problem is to get a collection of n items with maximal information. We consider the case where the information is additive, and thus need to maximize the sum of n independently identically distributed random variables x(k) observed sequentially in a sequence of length N. This is a stochastic dynamic-programming problem, the optimal solution of which is derived when the distribution of the x(k)'s is known. The asymptotic behaviour of this optimal solution (when N tends to infinity with n fixed) is considered. A (forced) certainty--equivalence policy is proposed for the case where the distribution is unknown and estimated on--line.
ic981396.pdf (From Postscript) TOP
Peter J Sherman, Iowa State University (U.S.A.)
Soon-Seng Lau, Iowa State University (U.S.A.)
This work addresses the influence of point spectrum on large sample statistics of the autoregressive spectral estimator. In particular, the asymptotic distributions of the AR coefficients, the innovations variance, and the spectral density estimator of a finite order AR(p) model for a mixed spectrum process are presented. Numerical simulations are performed to verify the analytical results.
ic981624.pdf (From Postscript) TOP
Er-wei Bai, University of Iowa (U.S.A.)
Minyue Fu, University of Newcastle (Australia)
Roberto Tempo, Politecnico di Torino (Italy)
Yinyu Ye, University of Iowa (U.S.A.)
The so-called analytic center approach to parameter estimation has been proposed recently as an alternative to the wel-known least squares approach. This new approach offers a parameter estimate that is consistent with the past data observations, has a simple geometric interpretation, and is computable using linear programming algorithms. In this paper, we study the asymptotic performance of the analytic center approach and show that the resulting estimate converges to the true parameter asymptotically, provided some mild conditions are satisfied. These conditions involve some weak persistent excitation and independence between noise and regressor, similar to the least squares case. This result is usedto derive a new parameter estimation approach which offers both good transient and asymptotic performances.
ic981628.pdf (From Postscript) TOP
Joel Le Roux, University of Nice, CNRS (France)
Cécile Huet, University of Nice, CNRS (France)
This paper presents a necessary and sufficient condition for the factorizability of higher order spectra of complex signals. Such a factorizability condition can be used to test if a complex signal can modelize the output of a linear and time invariant system driven by a stationary non gaussian white input. The condition developped here is based on the symmetries of higher order spectra and on an extension of a formula proposed by Marron et al. to unwrap third order spectrum phases. It is an identity between products of six higher order spectra values (which reduces to four values if only phases are considered). Our factorizability test requires no phase unwrapping, unlike existing methods developped in the cepstral domain. Moreover its extension to the N-th order case is direct. Simulations illustrate the deviation to this factorizability condition in a factorizable case (linear system) and a non factorizable case (non linear system).
ic981715.pdf (From Postscript) TOP
David Declercq, ETIS CNRS (France)
Patrick Duvaut, ETIS CNRS (France)
We present, in this contribution, some aspects of nongaussian H-ARMA models. After recalling that an H-ARMA process is obtained by passing an ARMA process through a Hermite polynomial nonlinearity, we describe the theoretical analysis of their cumulants and cumulant spectra. The main advantage of this kind of model is that the cumulant structure of the output can be deduced directly from the input covariance sequence. We give the analytic forms of these cumulants, together with some comments on their estimation. Then, we present the problems we are facing concerning the identification of the model's parameters, and give a first (and naive) method for their estimation. We give some results obtained on synthetic data and finally conclude with some remarks on this class of processes.
ic982016.pdf (Scanned) TOP
Robin L Murray, University of Rhode Island (U.S.A.)
Antonia Papandreou-Suppappola, University of Rhode Island (U.S.A.)
G. Faye Boudreaux-Bartels, University of Rhode Island (U.S.A.)
For analysis of signals with arbitrary dispersive phase laws, we extend the concept of higher order moment functions and define their associated higher order spectra. We propose a new higher order time-frequency representation(TFR), the higher order generalized warped Wigner distribution (HOG-WD). The HOG-WD is obtained by warping the previously proposed higher order Wigner distribution, and is important for analyzing signals with arbitrary time-dependent instantaneous frequency. We discuss links to prior higher order techniques and investigate properties of the HOG-WD. We extend the HOG-WD to a class of higher order, alternating sign, frequency-shift covariant TFRs. Finally, we demonstrate the advantage of using the generalized higher order spectra to detect phase coupled signals with dispersive instantaneous frequency characteristics.
ic982091.pdf (From Postscript) TOP
Ananthram Swami, Army Research Laboratory (U.S.A.)
Mounir Ghogho, University of Strathclyde, Scotland (U.K.)
The problem of interest is the estimation of the parameters of harmonics in the presence of additive and multiplicative noise. Expressions for the asymptotic performance of the cyclic-variance (CV) based method are derived when the multiplicative noise has non-zero mean. We show that the CV-based method may yield more accurate results than methods based onthe cyclic mean (CM), depending upon the color of the noise and the intrinsic and local SNRs. Performance is analyzed in detail for several special cases of the multiplicative noise, such as white Gaussian, AR and generalized-Gaussian noise.
ic982098.pdf (From Postscript) TOP
Anne Ferreol, Thomson-CSF (France)
Pascal Chevalier, Thomson-CSF (France)
Most of the HO blind source separation methods developed this last decade aim at blindly separating statistically independent sources, assumed stationary and ergodic. Nevertheless, in many situations such as in radiocommunications, the sources are non stationary and very often (quasi)-cyclostationary (digital modulations). In these contexts, it is important to wonder if the performance of these HO blind source separation methods may be affected by the potential non stationarity of the sources. The purpose of this paper is to bring some answers to this question through the behaviour analysis of the classical fourth-order cumulant estimators in the presence of (quasi)-cyclostationary sources.
ic982578.pdf (From Postscript) TOPConstantinos B. Papadias, Lucent Technologies/Bell Laboratories (U.S.A.)
We consider the problem of separating adaptively p synchronous user signals that are received by an m-element antenna array without the use of training sequences. We establish a set of necessary and sufficient conditions for perfect recovery of all the transmitted signals. Based on these conditions we propose optimization criteria that lead to adaptive algorithms for efficient blind source separation of non-Gaussian signals. Convergence analysis shows important global convergence properties of the proposed techniques. Combined with their low computational complexity, these features make the proposed algorithms good candidates for adaptive source separation.
ic981609.pdf (From Postscript) TOP