Using Hidden Markov Models to analyse time series data
Ragnhild C. Noven · 2004
Hidden Markov models describe the data by assuming that it depends in a random way on an underlying Markov process which cannot be observed. The Markov process transitions between different states, and each state gives a different probability of observing a given value. This allows us to decompose the variation in the data as coming from two sources, the current state of the Markov process and its probability distribution.