For m 8 and l 8 there would be over 1 106 states and

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ay S states, there are at most S 2 branch costs to compute and S path costs to store at each symbol time. Because of this the Viterbi algorithm is broadly used in digital communications as well as in many other signal processing applications. Even so, for many MLSE problems, S will be prohibitively large. For example, for CPM, the number of trellis states is proportional to M L−1 . For M = 8 and L = 8, there would be over 1 × 106 states, and thus on the order of 1012 branches. For a problem of this size, even the Viterbi algorithm is impractical. For problems where the number of trellis states is too large, there are suboptimum alternatives to the Viterbi algorithm. These are often based on alternative sequence pruning strategies. The depth-first class of pruning algorithms were commonly used prior to popularization of the Viterbi algorithm in the 1970’s, and are still used today. One of these is the Stack algorithm, which is discussed in Subsection 8.2.7 of the Course Text within the context of convolution co...
View Full Document

This document was uploaded on 10/12/2009.

Ask a homework question - tutors are online