Hidden Markov model topology estimation to characterize the dynamic structure of repetitive lifting data

Hidden Markov modeling (HMM) provides a probabilistic framework for modeling time series of multivariate observations. An HMM describes the dynamic behavior of the observations in terms of movement among the states of a finite-state machine. We have developed an algorithm that estimates the topology of an HMM for a given set of time series data. Our algorithm iteratively removes states and state transitions from a large general HMM topology and selects the topology estimate based on a likelihood criterion and a heuristic evaluation of complexity. The goal of our approach is to allow the data to reveal their own dynamic structure without external assumptions concerning the number of states or the pattern of transitions. In this paper, we describe the algorithm and apply it to estimate the dynamic structure of human body motion data from a repetitive lifting task. The estimated topology for low back pain patients was different from the topology for a control subject group. The body motions of patients tend not to change over the task, but the body motions of control subjects change systematically.