Title:Lattice Particle Filters

Abstract:A standard approach to approximate inference in state-space models isto apply a particle filter, e.g., the Condensation this http URL, the performance of particle filters often varies significantlydue to their stochastic this http URL present a class of algorithms, called lattice particle filters, thatcircumvent this difficulty by placing the particles deterministicallyaccording to a Quasi-Monte Carlo integration this http URL describe a practical realization of this idea, discuss itstheoretical properties, and its this http URL results with a synthetic 2D tracking problem show that thelattice particle filter is equivalent to a conventional particle filterthat has between 10 and 60% more particles, depending ontheir "sparsity" in the this http URL also present results on inferring 3D human motion frommoving light displays.