Hidden Markov independent components for biosignal analysis
W.D. Penny · 2000
A generative model is proposed for the analysis of biomedical signals which are known to be highly non-stationary multivariate time series. The model combines three existing statistical models, hidden Markov models (HMM), independent component analysis (ICA) and generalized autoregressive models (GAR), into a single overall model. The model can learn the dynamics of underlying sources and the parameters of mixing process. It can detect discrete changes in either the source dynamics or the mixing or both.