A COMPARISON OF APPROXIMATE VITERBI TECHNIQUES AND PARTICLE FILTERING FOR DATA ESTIMATION IN DIGITAL COMMUNICATIONS

We consider trellis-based algorithms for data estimation in digital communication systems. We present a general framework which includes approximate Viterbi algorithms like the M-algorithm and the T-algorithm as well as particle filtering algorithms. The algorithmic concepts are very close, since the difference is simply the choice of the norm in the weights calculation. The general framework yields hence a new interpretation of these algorithms and may give rise to a series of new algorithms by using general selection schemes or a different choice for the norm. We show that the (approximate) expectation maximization Viterbi algorithm (EMVA) profits from using Chi Squared optimal selection compared to the standard EMVA.