Kernel-based deterministic blue-noise sampling of arbitrary probability density functions

This paper provides an efficient method for approximating a given continuous probability density function (pdf) by a Dirac mixture density. Optimal parameters are determined by systematically minimizing a distance measure. As standard distance measures are typically not well defined for discrete densities on continuous domains, we focus on shifting the mass distribution of the approximating density as close to the true density as possible. Instead of globally comparing the masses as in a previous paper, the key idea is to characterize individual Dirac components by kernel functions representing the spread of probability mass that is appropriate at a given location. A distance measure is then obtained by comparing the deviation between the true density and the induced kernel density. This new method for Dirac mixture approximation provides high-quality approximation results, can handle arbitrary pdfs, allows considering constraints for, e.g., maintaining certain moments, and is fast enough for online processing.

[1]  Uwe D. Hanebeck,et al.  Localized Cumulative Distributions and a multivariate generalization of the Cramér-von Mises distance , 2008, 2008 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems.

[2]  Uwe D. Hanebeck,et al.  Dirac mixture approximation of multivariate Gaussian densities , 2009, Proceedings of the 48h IEEE Conference on Decision and Control (CDC) held jointly with 2009 28th Chinese Control Conference.

[3]  Uwe D. Hanebeck,et al.  PGF 42: Progressive Gaussian filtering with a twist , 2013, Proceedings of the 16th International Conference on Information Fusion.

[4]  O.C. Schrempf,et al.  Density Approximation Based on Dirac Mixtures with Regard to Nonlinear Estimation and Filtering , 2006, Proceedings of the 45th IEEE Conference on Decision and Control.

[5]  Hugh F. Durrant-Whyte,et al.  A new method for the nonlinear transformation of means and covariances in filters and estimators , 2000, IEEE Trans. Autom. Control..

[6]  Uwe D. Hanebeck,et al.  S2KF: The Smart Sampling Kalman Filter , 2013, Proceedings of the 16th International Conference on Information Fusion.

[7]  Uwe D. Hanebeck,et al.  Greedy algorithms for dirac mixture approximation of arbitrary probability density functions , 2007, 2007 46th IEEE Conference on Decision and Control.

[8]  T. Singh,et al.  The higher order unscented filter , 2003, Proceedings of the 2003 American Control Conference, 2003..

[9]  Robert Ulichney,et al.  Dithering with blue noise , 1988, Proc. IEEE.

[10]  Joris De Schutter,et al.  A The Linear Regression Kalman Filter , 2005 .

[11]  Gerhard Kurz,et al.  Bearings-only sensor scheduling using circular statistics , 2013, Proceedings of the 16th International Conference on Information Fusion.

[12]  Gerhard Kurz,et al.  Constrained object tracking on compact one-dimensional manifolds based on directional statistics , 2013, International Conference on Indoor Positioning and Indoor Navigation.

[13]  Raanan Fattal.  Blue-noise point sampling using kernel density model , 2011, SIGGRAPH 2011.

[14]  Uwe D. Hanebeck,et al.  Efficient deterministic dirac mixture approximation of Gaussian distributions , 2013, 2013 American Control Conference.

[15]  U.D. Hanebeck,et al.  Recursive Prediction of Stochastic Nonlinear Systems Based on Optimal Dirac Mixture Approximations , 2007, 2007 American Control Conference.

[16]  Simon J. Julier,et al.  The scaled unscented transformation , 2002, Proceedings of the 2002 American Control Conference (IEEE Cat. No.CH37301).

[17]  Gerhard Kurz,et al.  Recursive nonlinear filtering for angular data based on circular distributions , 2013, 2013 American Control Conference.

[18]  Uwe D. Hanebeck,et al.  Dirac Mixture Density Approximation Based on Minimization of the Weighted Cramer-von Mises Distance , 2006, 2006 IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems.

[19]  Marco F. Huber,et al.  Gaussian Filter based on Deterministic Sampling for High Quality Nonlinear Estimation , 2008 .

[20]  Oliver Deussen,et al.  Farthest-point optimized point sets with maximized minimum distance , 2011, HPG '11.

[21]  Ondrej Straka,et al.  Randomized unscented Kalman filter in target tracking , 2012, 2012 15th International Conference on Information Fusion.

[22]  E. Allgower,et al.  Introduction to Numerical Continuation Methods , 1987 .