Unformatted text preview: ...., s (k+D) )=
(k+D) =A1 ) Pk (s =A 1 , .... , s (k) =A 1) =A 2) ... P (s =A 1 , .... , s
k−1 (k) (k+D) =A2 ) Pk (s =A 1 , .... , s (k) (k+D) ... ..... ..... P (s =A 2 , .... , s
k−1 (k) (k+D) =A 1) Pk (s =A 2 , .... , s (k) (k+D) =A 1) ... .....
... ..... P (s =A M , .... , s
k−1 (k) (k+D) =AM )
D+1 Pk (s =A M , .... , s
D+1 (k) (k+D) =AM ) M states M states Figure 4: Illustration of computation of state probabilities required for the forward-only symbol-by-symbol MAP algorithm. We now describe the BCJR algorithm (Bahl, Cocke, Jelinek, Raviv; IEEE Trans. on IT; March 1974) which eﬃciently solves Eq (??) using a trellis formulation. This is, the BCJR algorithm is a forward/backward time-recursive block processing algorithm used to compute these probabilities for the block symbol-by-symbol MAP problem. The BCJR algorithm forms the basis for turbo (and more generally iterative) decoders. As is typical for a description or application of the BCJR algorithm, we will assume the binary symbol case (i.e. M = 2). Adopting a n...
View Full Document
- Spring '09