Hidden Markov mixtures of experts with an application to EEG recordings from sleep

Stefan Liehr, Klaus Pawelzik, Jens Kohlmorgen, Klaus Muller

Research output: Contribution to journalArticle

16 Citations (Scopus)

Abstract

We present a framework for the analysis of time series from nonstationary dynamical systems that operate in multiple modes. The method detects mode changes and identifies the underlying subdynamics. It unifies the mixtures of experts approach and a generalized hidden Markov model with an input-dependent transition matrix. The adaptation of the individual experts and of the hidden Markov model is performed simultaneously. We illustrate the capabilities of our algorithm for chaotic time series and EEG recordings from human subjects during afternoon naps.

Original languageEnglish
Pages (from-to)246-260
Number of pages15
JournalTheory in Biosciences
Volume118
Issue number3-4
Publication statusPublished - 1999 Dec 1
Externally publishedYes

    Fingerprint

Keywords

  • Dynamical mode detection
  • EEG
  • Hidden Markov models
  • Nonstationarity
  • Segmentation
  • Sleep
  • Time series

ASJC Scopus subject areas

  • Agricultural and Biological Sciences(all)

Cite this

Liehr, S., Pawelzik, K., Kohlmorgen, J., & Muller, K. (1999). Hidden Markov mixtures of experts with an application to EEG recordings from sleep. Theory in Biosciences, 118(3-4), 246-260.