István Kollár

Statistical theory of quantization

The effect of uniform quantization can often be modeled by an additive noise that is uniformly distributed, uncorrelated with the input signal, and has a white spectrum. This paper surveys the theory behind this model, and discusses the conditions of its validity. The application of the model to floating-point quantization is demonstrated.

Four-parameter fitting of sine wave testing result: Iteration and convergence

Small improvements to the iteration procedure of the IEEE Standard 1241 -2001 are suggested, and extension of the standard MATLAB program implementing the sine wave test i s discussed. The program is compatible with the LabView program already announced, and in o ther working modes offers extensions, too.

Bias of mean value and mean square value measurements based on quantized data

This paper investigates the imperfect fulfilment of the validity conditions of the noise model quantization. The general expressions of the deviations of the moments from Sheppards corrections are derived. Approximate upper and lower bounds of the bias are given for the measurement of first- and second-order moments of sinusoidal, uniformly distributed, and Gaussian signals. It is shown that because of the uncontrollable mean value at the input of the ADC (offset, drift), the worst-case values have to be investigated; it is illustrated how a simple-form envelope function of the errors can be used as an upper bound. Since the worst-case relative positions of the signal and the quantization characteristics are taken into account, the results are valid for both midtread and midrise quantizers, while in the literature results are given for a selected quantizer type only. >

On frequency-domain identification of linear systems

The maximum-likelihood estimation of the parameters of linear systems and the properties of the estimator (Estimator for Linear Systems, ELiS) have been described by R. Pintelson and J. Schoukens (see ibid., vol. 39, no.4, p.565-573, Aug. 1990). The mathematics used in the development of the method and the proofs is rather involved, although several statements can be understood in heuristic terms. The present author discusses the complex-domain description of the method, which results in much simpler expressions. The method is compared to other formulations, giving more insight into the properties of the estimate. It turns out that robustness is at least partly due to the least-squares formulation. Derivations are avoided where possible, and intuitive explanations are given instead. >

Multiparameter optimization of inverse filtering algorithms

This paper investigates inverse filtering of transient signals. The problem is ill-conditioned, which means that a small uncertainty in the measurement causes large deviations in the reconstructed signal. This amplified noise has to be suppressed at the price of bias in the estimation. The most difficult task is to find the optimal degree of noise reduction. Deconvolution algorithms are usually controlled by one or a few parameters. Several algorithms can be found in the literature to find the best setting of inverse filtering methods; however, usually methods with only one free parameter are handled. In this paper, an algorithm is proposed to optimize several parameters, on the basis of a spectral model. Multiparameter inverse filtering methods have the advantage that they can be better adapted to the measurement system, and to the noise and signal to be measured. The superiority of the proposed optimization method is demonstrated both on simulated and on experimental data.

Statistical analysis of nonparametric transfer function estimates

The Empirical Transfer Function Estimate (ETFE) is the ratio of the Fourier transforms of the output and input signals of a system. It works well when the input signal is deterministic and exactly known. However, when the input signal is random, or it can only be observed with an observation error, the quality of the ETFE deteriorates. Its variance can be infinite even for large signal-to-noise ratios. This is not well known. This paper establishes and analyzes a mathematical model of the ETFE with noisy input signals. It explains the cause of the large variance and suggests modifications which eliminate the above problems.

Frequency Domain System Identification Toolbox for MATLAB

Abstract A frequency domain system identification package is described, written in MATLAB. The whole experiment design and evaluation procedure is supported: excitation signal optimization (binary and arbitrary waveforms), data preprocessing and variance analysis, parameter estimation via nonlinear least squares fitting in the frequency domain, model validation, transfer function and pole/zero plots with uncertainties, simulations. The results can be matched to those of the time domain SYSTEM IDENTIFICATION TOOLBOX, also available for MATLAB.

Optimal FIR and IIR Hilbert transformer design via LS and minimax fitting

A novel method for the design of digital finite impulse response (FIR) and infinite impulse response (IIR) Hilbert transformers using the least squares (LS) and the minimax criteria is presented. The LS approximation is performed in the complex domain. Also presented is an iterative extension of the algorithm, which results in a minimax (Chebyshev) approximation, also in the complex domain. For FIR filters the results are the same as those of the optimal methods known from the literature. For the same task, stable IIR filters have also been successfully designed. The procedures proposed are usable for the design of digital filters other than Hilbert transformers, since the desired frequency response can be given point by point. >

Identification of Volterra kernels using interpolation

This paper presents a new method for the identification of frequency-domain Volterra kernels. Since the nonlinear kernels often play a secondary role compared to the dominant, linear component of the system, it is worth establishing a balance between the degree of liberty of these components and their effect on the overall accuracy of the model. This is necessary in order to reduce the model complexity, hence the required measurement length. Based on the assumption that frequency-domain kernels are locally smooth, the kernel surfaces can be approximated by interpolation techniques, thus reducing the complexity of the model. Similarly to the unreduced (Volterra) model, this smaller model is also i) linear in the unknowns; ii) only locally sensitive to its parameters; and iii) free of structural assumptions about the system. The parameter estimation boils down to solving a linear system of equations in the least-squares (LS) sense. The design of the interpolation scheme is described and the performance of the approximation is analyzed and illustrated by simulation. The algorithm allows a significant saving in measurement time compared to other kernel estimation methods.

Automatic testing of graphical user interfaces

A m Graphical User Interfaces me very difficuh io lesi, since resting requires simulation of fhe activity of a person. The paper presenfs an approach where “guided” random selection and aclivafion of fhe controls is perfomed Guidance is implemenfed on fhe bcrsis of aprobabilify tuble. The fechnical means to perform the lest is an acfion recorder (event recorder). Besides testing, fhis is a useful tool fo perform demonsfrations and selflguided infroduction to the CUI. The recorder has been implemented in MATLAB, and it is available on the WEB.