Method of decoding a correcting code, for example a turbo-code, by analysis of the extended spectrum of the words of the code

Patent metadata

Field	Value
Publication number	US-9391644-B2
Application number	US-201414285160-A
Country	US
Kind code	B2
Filing date	May 22, 2014
Priority date	May 24, 2013
Publication date	Jul 12, 2016
Grant date	Jul 12, 2016

Abstract

An improved decoding method is provided making it possible to solve the problem of the error floor of a turbo-code or of an LDPC code, or more generally of a correcting code from the family of “turbo-like codes”, while preserving the same spectral efficiency, without any decrease in the useful throughput of the encoded stream. This result is obtained by an identification, on input to the decoder, of the bits on which an error has a strong impact, and a modification of the likelihoods corresponding to these bits so as to improve the convergence of the decoder.

First claim

The invention claimed is:

1. A computer-implemented iterative method for decoding a received code word from a correcting code, the code word being represented by a local likelihood value for each bit of the code word, the iterative method comprising the following steps executed by a processor and a memory:
analysing an extended spectrum of words of the correcting code to generate a table containing, for each bit of a code word, an information regarding distance between the code word and the code word obtained by decoding and then re-encoding the code word for which a value of a bit is erroneous, and
for each bit of index i of a subset of bits of the code word received, executing in an iterative manner the following steps:
decoding the received code word,
calculating an overall likelihood value IM(i) for the decoded code word,
comparing, onwards of a second iteration (i=2), the overall likelihood value IM(i) calculated at a current iteration with an overall likelihood value IM(i−1) calculated at a previous iteration,
modifying an amplitude and a sign of a local likelihood value LLR(i) for a bit of index i of the received code word, the selected index i corresponding to that index of the table which is associated with a lowest distance in the table, from among indices not selected at a previous iteration,
onwards of the second iteration (i=2), if the overall likelihood value IM(i) calculated at the current iteration is less than the overall likelihood value IM(i−1) calculated at the previous iteration, reinitializing the local likelihood value LLR(i−1) of the bit of index i−1 of the received code word that was modified at the previous iteration.

2. The iterative method of claim 1 , further comprising a stopping test.

3. The iterative method of claim 2 , in which the decoding step is iterative and the stopping test is a test on the number of iterations of the decoding step.

4. The iterative method of claim 2 , in which the stopping test is a test of convergence of the decoding step.

5. The iterative method of claim 4 , in which the convergence test comprises a detection of an error in a result of the decoding step by application of an error-detecting code.

6. The iterative method of claim 4 , in which the convergence test comprises a comparison of the overall likelihood value IM(i) with a predetermined threshold value.

7. The iterative method of claim 1 , in which the overall likelihood value IM(i) is a mutual information.

8. The iterative method of claim 1 , in which the overall likelihood value IM(i) is equal to an inverse of an Euclidian distance between the decoded and then re-encoded code word and the received code word.

9. The iterative method of claim 1 , in which the subset of the bits of the received code word is equal to the entirety of the bits of the received code word or to a subset consisting of bits with a local likelihood value LLR that is below a predetermined threshold.

10. The iterative method of claim 1 , in which analysing the extended spectrum comprises:
selecting a code word,
for each bit of the code word, executing in an iterative manner the following:
modifying the value of the bit,
decoding the modified code word,
encoding the result of the decoding, and
calculating a distance between the result of the encoding and the modified code word.

11. The iterative method of claim 1 , in which the correcting code is a turbo-code or an LDPC code.

12. A computer program comprising instructions stored on a tan non-transitory storage medium for executing on a processor an iterative method for decoding a received code word from a correcting code, the code word being represented by a local likelihood value for each bit of the code word, the iterative method comprising:
analysing an extended spectrum of words of the correcting code to generate a table containing, for each bit of a code word, an information regarding distance between the code word and the code word obtained by decoding and then re-encoding the code word for which the value of a bit is erroneous, and
for each bit of index i of a subset of bits of the code word received, executing in an iterative manner the following steps:
decoding the received code word,
calculating an overall likelihood value IM(i) for the decoded code word,
comparing, onwards of a second iteration (i=2), the overall likelihood value IM(i) calculated at a current iteration with an overall likelihood value IM(i−1) calculated at a previous iteration,
modifying an amplitude and a sign of a local likelihood value LLR(i) for a bit of index i of the received code word, the selected index i corresponding to that index of the table which is associated with a lowest distance in the table, from among indices not selected at a previous iteration,
onwards of the second iteration (i=2), if the overall likelihood value IM(i) calculated at the current iteration is less than the overall likelihood value IM(i−1) calculated at the previous iteration, reinitializing the local likelihood value LLR(i−1) of the bit of index i−1 of the received code word that was modified at the previous iteration.

13. A tangible non-transitory processor-readable recording medium on which is recorded a program comprising instructions for executing an iterative method for decoding a received code word from correcting code, the code word being represented by a local likelihood value for each bit of the code word, the iterative method comprising:
analysing an extended spectrum of words of the correcting code to generate a table containing, for each bit of code word, an information regarding distance between the code word and the code word obtained by decoding and then re-encoding the code word for which the value of a bit is erroneous, and
for each bit of index i of a subset of bits of the code word received, executing in an iterative manner the following steps:
decoding the received code word,
calculating an overall likelihood value IM(i) for the decoded code word,
comparing, onwards of a second iteration (i=2), the overall likelihood value IM(i) calculated at a current iteration with an overall likelihood value IM(i−1) calculated at a previous iteration,
modifying an amplitude and a sign of a local likelihood value LLR(i) for a bit of index i of the received code word, the selected index i corresponding to that index of the table which is associated with a lowest distance in the table from a on indices not selected at a previous iteration,
onwards of the second iteration (i=2), if the overall likelihood value IM(i) calculated at the current iteration is less than the overall likelihood value IM(i−1) calculated at the previous iteration, reinitializing the local likelihood value LLR(i−1) of the bit of index i−1 of the received code word that was modified at the previous iteration.

14. A device comprising a processor and a memory configured for executing an iterative method for decoding a received code word from a correcting code, the code word being represented by a local likelihood value for each bit of the code word, the iterative method comprising:
analysing an extended spectrum of words of the correcting code to generate a table containing, for each bit of a code word, an information regarding distance between the code word and the code word obtained by decoding and then re-encoding the code word for which the value of a bit is erroneous, and
for each bit of index i of a subset of bits of the code word received, executing in an iterative manner the following steps:
decoding the received code word,
calculating an overall likelihood value IM(i) for the decoded code word,
comparing, onwards of a second i

Assignees

• Thales Sa

Inventors

• Gadat Benjamin
• Van Wambeke Nicolas
• Ries Lionel

Classifications

• H03M13/3746

  with iterative decoding · CPC title

• H03M13/3753 Primary

  using iteration stopping criteria · CPC title

• H03M13/2957 Primary

  Turbo codes and decoding · CPC title

• H03M13/3776

  using a re-encoding step during the decoding process · CPC title