Forskning ved Københavns Universitet - Københavns Universitet


An Efficient Algorithm for Modelling Duration in Hidden Markov Models, with a Dramatic Application

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningfagfællebedømt

  • Søren Hauberg
  • Jakob Sloth
For many years, the hidden Markov model (HMM) has been one of the most popular tools for analysing sequential data. One frequently used special case is the left-right model, in which the order of the hidden states is known. If knowledge of the duration of a state is available it is not possible to represent it explicitly with an HMM. Methods for modelling duration with HMM's do exist (Rabiner in Proc. IEEE 77(2):257---286, [1989]), but they come at the price of increased computational complexity. Here we present an efficient and robust algorithm for modelling duration in HMM's, and this algorithm is successfully used to control autonomous computer actors in a theatrical play.
Udgivelsesdato: 2008
TidsskriftJournal of Mathematical Imaging and Vision
Udgave nummer2-3
Sider (fra-til)165-170
Antal sider6
StatusUdgivet - 2008

ID: 9197652