Computational Modelling Group

Seminar  29th April 2010 3:45 p.m.  54/10B

Sequential MCMC for tracking of dynamically evolving groups and musical audio score estimation

Simon Godsill
University of Cambridge

Categories
Acoustics, Complex Systems, Monte Carlo, Scientific Computing
Submitter
Stefanie Biedermann

In this talk I will describe recent developments in sequential MCMC approaches to Bayesian tracking. Probabilities are propagated through time using a Monte Carlo representation, as in standard particle filters, but updating of states is carried out using specially constructed MCMC schemes. Applications are presented from tracking of dynamically evolving groups of objects and for transcription of musical audio scores.