Training by ART-2 and classification of ballistic missiles using Hidden Markov Model

Abstract

This paper addresses the classification of different ranges of Ballistic Missiles (BM) for air defense applications using Adaptive Resonance Theory (ART-2) and Hidden Markov Model (HMM). ART-2 finds the initial clusters using unsupervised learning to be fed to HMM for classification using recursive method. The classification is based on derived parameters of specific energy, acceleration, altitude and velocity which in turn are acquired from measured data by radars. To meet the conflicting requirements of classifying short as well as long-range BM trajectories, we are proposing a formulation for partitioning the trajectory by using a moving window concept. Experimental results show that the HMM model is able to classify above 95% within time of the order of milliseconds once initial data is trained using ART2. © Springer-Verlag 2013.