Hidden Markov Models in radar target classification

A classification technology is presented that uses Hidden Markov Models (HMMs) to classify simulated and real radar signals from five classes of targets: personnel, tracked vehicles, wheeled vehicles, helicopters and propeller aircrafts. Similar to techniques that have been well proven in speech recognition, the time-varying nature of radar Doppler data is exploited. The method classifies the different targets by their different Doppler characteristics. The purpose of this paper is to make a comparison between three kinds of HMM Methods: 1. HMM with continuous outputs (CHMM) 2. HMM with discrete outputs (DHMM) 3. Semi-continuous Hidden Markov Models (SCHMM)