In soft output viterbi algorithm (SOVA) there are two decoding stages. In the first stage and at each state we compute the reliability information about the survivor paths based on the difference of ML values (or any measure) between the survivor and the competing paths. After that the first stage should output a decoded sequence plus a reliability measure about each value on that sequence. suppose that the reliability value is the distance between the survivor and the competing path and this information along with the sequence will enter as an apriori information to the second SOVA, how this data can be used to build the trellis in the second stage to increase decision reliability and to decode the sequence?
Thanks in advance
Dear Kassem, the correct use of soft reliability for successive decoding in serial or parallel concatenated decoding is a crucial issue and in general it is not easy to say what is the best solution. When the soft reliability is an LLR, i.e., log(P(0|r)/P(1|r)) it is easy to disentangle two different contribution, i.e., LLR = log(P(r|0)/P(r|1)) + log(P(0)/P(1)), where the first contribution is denoted by extrinsic LLR and is due to the observations while the second is due to a priori information. Hence, in cascade decoding, the output of a decoder is directly fed at the input of the other decoder as a priori information
Please see best written articles
http://www.ece.lehigh.edu/~jingli/teach/S2003turbo/notes/SOVA.pdf
https://theses.lib.vt.edu/theses/available/etd-71897-15815/unrestricted/chap4.pdf
it may help you lot
- Badges
- Science topic