Digital data decoder that derives codeword estimates from soft data

• US 6,145,110 A
• Filed: 06/22/1998
• Issued: 11/07/2000
• Est. Priority Date: 06/22/1998
• Status: Expired due to Term

• Current Assignee: Ericsson, Inc. (Telefonaktiebolaget LM Ericsson)
• Original Assignee: Ericsson, Inc. (Telefonaktiebolaget LM Ericsson)
• Inventors: Khayrallah, Ali S.
• Primary Examiner(s): MOISE, EMMANUEL LIONEL
• Application Number:
  US09/102,291
• Time in Patent Office: 869 Days
• Field of Search: 714/751, 714/752, 714/755, 714/780, 714/782, 714/785, 714/761
• US Class Current: 714/752
• CPC Class Codes: H04L 1/0057 Block codes H04L1/0061, H04L1/0064 take precedence

Abstract

A forward error correction decoder receives a received word comprising a plurality of characters. The decoder also receives a reliability value for each character. The decoder calculates a syndrome for each member of the set of possible received words by transforming each member with the parity check matrix. Members of the set having the same syndrome are grouped together into cosets. A weight based on the reliability value is then assigned to each member in each coset. The weight is defined to be the number of low reliability, non-zero characters in each member of the set of possible received words. For each coset, the member that has the lowest weight is selected as a coset leader. A syndrome is computed for the soft codeword by transforming it with the parity check matrix. The coset leader with the same syndrome as the received word is subtracted from the received word resulting in a codeword estimate.

Claims

1. A method for use in a data decoder for deriving a codeword estimate from an actual received word, said codeword estimate comprising restored encoded data, said actual received word comprising one of a predetermined set of possible received words, said actual received word comprising a plurality of characters, each of said characters having a reliability value, said method comprising the steps of:

accepting an actual received word;

computing a syndrome for said actual received word;

computing a syndrome for each member of the predetermined set of possible received words;

classifying each member of the predetermined set of possible received words into a plurality of cosets according to each member'"'"'s syndrome;

determining a weight for each member of the predetermined value of possible received words, said weight comprising a function of said reliability values;

selecting a coset leader from each of said cosets, said coset leader having a predefined weight;

summing said plurality of reliability values and selecting a coset leader table corresponding to said sum from a plurality of coset leader tables;

selecting a coset leader with the same syndrome as said received word syndrome; and

subtracting said selected coset leader from said received word to derive said codeword estimate.

2. A method in accordance with claim 1 wherein said step of computing a syndrome for said actual received word comprises manipulating said received word with a decoding matrix.

3. A method in accordance with claim 1 wherein said predefined weight is the lowest of said determined weights.

4. A method in accordance with claim 1 wherein said predefined weight comprises the fewest number of non-zero characters in said received word.

5. A method in accordance with claim 1 wherein said predefined weight comprises the fewest number of non-zero characters with low reliability scores.

6. A linear decoder that is capable of using hard information comprising definite data and soft information comprising indefinite information, said method comprising:

a memory storing a table including a plurality of syndromes and an equal plurality of coset leaders, each of said plurality of syndromes being paired with one of said plurality of coset leaders;

means for computing a syndrome in response to accepting a received word;

means for using said soft information to select one of said plurality of coset leader tables; and

means for determining an estimate of said codeword using said received code word and said selected coset leader.

7. A linear decoder in accordance with claim 6 wherein said means for determining an estimate of said codeword comprises means for selecting a coset leader responsive to said syndrome.

8. A linear decoder in accordance with claim 6 wherein said means for determining an estimate comprises means for subtracting said coset leader from said received word.

9. A method for deriving a codeword estimate from a codeword comprising a plurality of characters, each of said characters having been assigned a reliability value, said method comprising:

determining whether said received word is in error; and

deriving a codeword estimate by correcting one or more of said characters with low reliability values, comprising determining a limited set of possible received words, determining a syndrome for each member of said set, classifying each member of said set into a coset based on its syndrome, weighting each member of said received word based on said characters with low reliability, selecting a coset leader for each coset based on a predetermined weight, determining a syndrome for said received word, and subtracting said coset leader with the same syndrome as said received word from said received word.

10. A method in accordance with claim 9 wherein said step of determining whether said received word is in error comprises computing a syndrome for said received word, wherein one or more of the resulting syndromes indicate an error.

11. A method in accordance with claim 10 wherein said codeword is encoded by manipulating said codeword with an encoding matrix, and wherein computing a syndrome for said codeword comprises manipulating said codeword with a decoding matrix.

12. A method in accordance with claim 11 wherein manipulating said encoding matrix and said decoding matrix yields zero, and wherein an error is detected when said syndrome of said received word is non-zero.

13. A method for developing a table of coset leaders corresponding to a syndrome for use in a decoder that derives a codeword estimate from a received word by subtracting a selected one of said coset leaders according to its syndrome, said codeword comprising a plurality of characters, each of said characters having a reliability value, said method comprising the steps of:

(a) determining all members of the set of possible received words;

(b) determining a syndrome for each member of said set;

(c) classifying each member of said set into a coset based on its syndrome;

(d) weighting each member of said set based on one or more of said characters having low reliability values;

(e) selecting a coset leader from each coset having a predetermined weight;

(f) repeating steps (d) and (e) for each member of said set'"'"'s reliability values; and

(g) storing said table in a memory device.

14. A method for use in a data decoder for deriving a codeword estimate from a received word, said codeword estimate comprising restored encoded data, said received word comprising one of a predetermined set of received words, said received word comprising a plurality of characters, each of said characters having a reliability value, said data decoder including a memory wherein each member of said predetermined set of received words is classified into one of a plurality of cosets according to the word'"'"'s syndrome, said method comprising the steps of:

accepting a received word and a reliability value for each of said plurality of characters in said received word;

computing a syndrome for said received word;

selecting one of said plurality of cosets corresponding to said computed syndrome of said received word;

selecting a coset leader from each of said cosets according to said reliability values for each of said plurality of characters in said received word; and

subtracting said selected coset leader from said received word to derive said codeword estimate.