Modeling Interleaved Hidden Processes

Hidden Markov models assume that obser-vations in time series data stem from some hidden process that can be compactly repre-sented as a Markov chain. We generalize this model by assuming that the observed data stems from multiple hidden processes, whose outputs interleave to form the sequence of ob-servations. Exact inference in this model is NP-hard. However, a tractable and effective inference algorithm is obtained by extend-ing structured approximate inference meth-ods used in factorial hidden Markov mod-els. The proposed model is evaluated in an activity recognition domain, where multiple activities interleave and together generate a stream of sensor observations. It is shown to be more accurate than a standard hidden Markov model in this domain. 1