Ground vehicle classification based on Hierarchical Hidden Markov Model and Gaussian Mixture Model using wireless sensor networks

In this paper, multiple ground vehicles passing through a region that are observed by audio sensor arrays are efficiently classified using a Hierarchical Hidden Markov Model (HHMM). The states in the HHMM contain another HMM which represents a time sequence of the vehicle acoustic signals. The HMM represents the distribution of the output of the HHMM, where The HMM models the features of the continuous acoustic emissions. The output of the states of this HMM is modeled as Gaussian Mixture Model (GMM), where the number of states and the number of Gaussians are experimentally determined, while the other parameters are estimated using Expectation Maximization (EM). The HHMM is used to model the sequence of the local decisions which are based on multiple hypothesis testing with maximum likelihood approach. The states in the HHMM represent various combinations of vehicles of different types. Simulation results demonstrate the efficiency of this scheme.