A suboptimal estimator of the sampling jitter variance using the bispectrum

نویسندگان

  • Ilan Sharfer
  • Hagit Messer
چکیده

We consider the problem of estimating parameters of an irregular sampling process defined as a uniform sampling process in which the deviations from the nominal sampling times constitute a random IID process (jitter). Emphasis is placed on estimating the variance of the jitter, based on observation of samples taken from a continuous band-limited third-order stationary process. We derive an estimation procedure which uses the bispectrum estimates of a process with a priori known bispectrum. Derivation of the generalized likelihood ratio in the bispectral domain, leads to a statistic with which a bispectrum-based maximum likelihood estimation can be done. We propose a suboptimal estimator, and show that it is asymptotically unbiased and consistent. The dependence of the estimator's performance on the data length and the skewness is studied for a specific example. The estimator's variance is compared to the bispectrum-based Cramer-Rao bound (BCRB), and is shown to approach it for sufficiently large data length or skewness. Computer simulations verify the effectiveness of the proposed estimation method for small jitter. Zusammenfassung Wir betrachten die Sch/itzung von Parametern eines unregelm/iBigen Abtastprozesses. Dieser ist definiert als regelm/iBiger AbtastprozeB mit statistisch unabh/ingigen und identisch verteilten zuf/illigen Abweichungen von den nominellen Abtastzeitpunkten (Jitter). Im Vordergrund steht die Schfitzung der Varianz des Jitters basierend auf beobachteten Abtastwerten eines stetigen, bandbegrenzten und bis zur dritten Ordnung station/iren Prozesses. Wir leiten eine Sch/itzmethode ab, die die Bispektrum-Sch~itzwerte eines Prozesses mit bekanntem Bispektrum verwendet. Die Ableitung des verallgemeinerten Likelihood-Verh/iltnisses im Bispektrum-Bereich ffihrt zu einer Statistik, die eine auf dem Bispektrum beruhende Maximum-Likelihood-Sch/itzung erm6glicht. Wir schlagen einen suboptimalen Sch~itzer vor und zeigen dessen asymptotische Erwartungstreue und Konsistenz. Die Abh/ingigkeit der Leistungsf~ihigkeit des Sch~itzers vom Umfang der Daten und yon der Asymmetrie wird anhand eines konkreten Beispiels untersucht. Die Varianz des Sch/itzers wird mit der auf dem Bispektrum beruhenden Cramer-Rao-Schranke (BCRB) verglichen, und es wird gezeigt, dab sich die Varianz der BCRB bei hinreichend groBem Datenumfang oder hinreichend groBer Asymmetrie ann~ihert. Computersimulationen best/itigen die Effektivit/it der vorgeschlagenen Sch/itzmethode im Fall kleinen Jitters. *Corresponding author. E-maih [email protected]. 0165-1684/94/$7.00 ,~ 1994 Elsevier Science B.V. All rights reserved SSDI OI 6 5 1 6 8 4 [ 9 4 ) O O O 2 6 V 170 1. Shar[er, H. Messer / Signal Processing 38 (1994) 169 186 R~ume Nous consid~rons dans cet article le probl~me de l'estimation des param~tres d'un processus fi ~chantillonnage irr6gulier d6fini comme processus ~ 6chantillonnage uniforme pour lequel les d6viations vis-fi-vis des instants d'6chantillonnage uniforme constituent un processus al6atoire ind6pendant fi distribution constante (gigue). L'emphase est raise sur l'estimation de la variance de la gigue, sur la base d'observation d'6chantillons obtenus fi partir d'un processus continu stationnaire d'ordre trois fi bande limit+. Nous d6rivons une proc6dure d'estimation qui utilise les estim~es du bispectre d'un processus dont le bispectre est connu a priori. La d6rivation du rapport de vraisemblance g6n6ralis6 dans le domaine bispectral conduit fi une statistique avec laquelle une estimation aux maximum de vraisemblance bas+e sur le bispectre peut ~tre effectu6e. Nous proposons un estimateur sous-optimal, et montrons qu'il est asymptotiquement non biais6 et consistant. La d6pendance de la performance de I'estimateur vis-~t-vis de la longueur des donn6es et de l'asymm6trie de la distribution est 6tudi6e sur un exemple sp6cifique. La variance de l'estimateur est compar6e fi la borne de Cramer-Rao sur le bispectre, et il est montr6 qu'elle s'en approche pour une longueur de donn~es ou une asymm6trie de distribution suffisamment grande. Des simulations sur ordinateur permettent de v6rifier l'efficacit6 de la m6thode d'estimation propos6e dans le cas d'une gigue faible.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

On reducing sampling variance in covariate shift using control variates

Covariate shift classification problems can in principle be tackled by importanceweighting training samples. However, the sampling variance of the risk estimator is often scaled up dramatically by the weights. This means that during cross-validation when the importance-weighted risk is repeatedly evaluated suboptimal hyperparameter estimates are produced. We study the sampling variances of the ...

متن کامل

CMB bispectrum, trispectrum, non-Gaussianity, and the Cramer-Rao bound

Minimum-variance estimators for the parameter fnl that quantifies local-model non-Gaussianity can be constructed from the cosmic microwave background (CMB) bispectrum (three-point function) and also from the trispectrum (four-point function). Some have suggested that a comparison between the estimates for the values of fnl from the bispectrum and trispectrum allow a consistency test for the mod...

متن کامل

Estimation of Variance of Normal Distribution using Ranked Set Sampling

Introduction     In some biological, environmental or ecological studies, there are situations in which obtaining exact measurements of sample units are much harder than ranking them in a set of small size without referring to their precise values. In these situations, ranked set sampling (RSS), proposed by McIntyre (1952), can be regarded as an alternative to the usual simple random sampling ...

متن کامل

Stratified Median Ranked Set Sampling: Optimum and Proportional Allocations

In this paper, for the Stratified Median Ranked Set Sampling (SMRSS), proposed by Ibrahim et al. (2010), we examine the proportional and optimum sample allocations that are two well-known methods for sample allocation in stratified sampling. We show that the variances of the mean estimators of a symmetric population in SMRSS using optimum and proportional allocations to strata are smaller than ...

متن کامل

Importance Sampling in Path Space for Diffusion Processes

Importance sampling is a widely used technique to reduce the variance of the Monte Carlo method. It uses the idea of change of measure to design efficient Monte Carlo estimators. In this work, we study the importance sampling method in the framework of diffusion process and consider the change of measures which can be realized by adding a control force to the original dynamics. For certain expo...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:
  • Signal Processing

دوره 38  شماره 

صفحات  -

تاریخ انتشار 1994