FREQUENCY LINE TRACKING USING HIDDEN MARKOV-MODELS

This paper demonstrates how the problem of frequency line tracking can be formulated in terms of hidden Markov models (HMM's). Frequency cells comprising a subset, or gate, of the spectral bins from FFT processing are identified with the states of the hidden Markov chain. An additional zero state is included to allow for the possibility of track initiation and termination. Analytic expressions are obtained for the basic parameters of the HMM in terms of physically meaningful quantities, and optimization of the HMM tracker is carefully discussed. A measurement sequence based on a simple threshold detector forms the input to the tracker. The outputs of the HMM tracker are a discrete Viterbi track, a gate occupancy probability function, and a continuous mean cell occupancy track. The latter provides an estimate of the mean signal frequency as a function of time. The performance of the HMM tracker is evaluated for two sets of simulated data and is found to be remarkably good, comparing favorably to results from inspection of the signal spectrograms. A comparison of the HMM tracker to earlier, related trackers is presented, and possible extensions are discussed. © 1990 IEEE