MHT Viterbi Algorithm for Number-of-Track Estimation

In this section we describe an approach to controlling the computational load for the MHT methods described above. This approach starts with a trellis diagram representation on hypothesized track sets, and uses a generalized Viterbi algorithm approach to hypothesis pruning and merging. A principal advantage of this approach to hypothesis reduction is that at any time n it systematically keeps some less likely hypotheses, which reduces the probability of dropping hypotheses that may become most likely at a later time.

In the first two subsections we assume that K is known and constant over the processing time. In Subsection 4.3 we expand that approach to unknown but constant K, and in Subsection 4.4 we discuss time varying unknown K.