Multinomial HMMs for Intent Recognition in Maritime Domains

The need for increased maritime security has prompted research focus on intent recognition solutions for the naval domain. We consider the problem of early classification of the hostile behavior of agents in a dynamic maritime domain and propose a solution using multinomial Hidden Markov Models (HMMs). To enable early detection of hostile behaviors, the proposed approach encodes as observable symbols the rate of change (instead of static values) for parameters relevant to the task. We discuss our implementation of a one-versus-all intent classifier using multinomial HMMs and present the results of our system on three types of hostile behaviors (ram, herd, block) and a benign behavior.