# 32 for k 4 receiver observations r1 r2 r3 r4

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

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: p · M L−1 paths at stage n to stage n + 1 by adding the ˆ new branch cost to the previous path cost; 3. at stage n + 1, for each state, compare all path cost into the state and keep only the one with lowest cost (i.e. prune the rest of the paths). Example 4 - Coherent DPSK: Consider the coherent DPSK MLSE problem discussed last week in Subsection 2.3.2. For K = 4 receiver observations, {r1 , r2 , r3 , r4 } = {1.3, −.7, 1.1, −0.2}, the MLSE cost was K min 4 S1 D (r 4 , S 4 ) = 1 1 n=1 ||rn − Sn ||2 , (18) where Sn is the state (and also the symbol) at time n for the hypothesized sequence S 4 . For this problem, the state can be either “1” or “-1”. The trellis diagram, 1 ﬁrst shown in Figure 2 from Lecture 4, is redrawn here in Figure 2 to show the stage-by-stage implementation of the Viterbi algorithm. As before, each branch label with its cost ||rn − Sn ||2. Figure 2(a) shows processing up to stage 2, where the ﬁrst Viterbi pruning takes place. At each state, costs for all incoming paths...
View Full Document

## This document was uploaded on 10/12/2009.

Ask a homework question - tutors are online