Generalized viterbi decoding algorithms

• US 5,208,816 A
• Filed: 03/11/1992
• Issued: 05/04/1993
• Est. Priority Date: 08/18/1989
• Status: Expired due to Term

• Current Assignee: AT&T, Inc.
• Original Assignee: AT&T, Inc.
• Inventors: Seshardi, Nambirajan, Sundberg, Carl-Erik W.
• Primary Examiner(s): Beausoliel, Jr., Robert W.
• Assistant Examiner(s): LO, ALLEN
• Application Number:
  US07/850,239
• Time in Patent Office: 419 Days
• Field of Search: 371/43, 371/44, 371/45, 371/37.1
• US Class Current: 714/795
• CPC Class Codes: H03M 13/39 Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes
• CPC Class Codes: H03M 13/4115 list output Viterbi decoding

Abstract

A data transmission system and method for processing speech, image and other data disclosed which embodies parallel- and serial-generalized Viterbi decoding algorithms (GVA) that produce a rank ordered list of the L best candidates after a trellis search. Error detection is performed by identifying unreliable sequences through comparison of the likelihood metrics of the two or more most likely sequences. Unreliable sequences are re-estimated using inter-frame redundancy or retransmission.

Claims

1. A method of processing a coded information sequence received from a transmitter comprising,

determining the value for a metric indicative of the relative quality for each of the L (>

1) best candidates to be selected as the sequence actually transmitted,

wherein said candidates each correspond to paths through a trellis structure having a plurality of stages and a plurality of states at each stage,

wherein each path entering each state at each stage has a partial metric associated with it,

wherein said determining comprises finding the L paths through said trellis which have the highest sum of partial metrics,

said determining further comprises recognizing at each stage of the trellis the L paths entering each state that have the highest cumulative metric at that stage, and

retaining such recognized paths as partial paths leading to the L paths through the trellis having the largest sum of partial metrics.

2. The method of claim 1 further comprising

identifying as unreliable the candidate having the highest value for said metric whenever said highest value differs from the value for said metric for another of said L sequences by less than a predetermined amount.

3. The method of claim 2 further comprising

accepting the candidate having the highest value for said metric as the sequence transmitted when it is not identified as being unreliable, and

re-estimating a transmitted sequence when the candidate having the highest value for said metric is identified as being unreliable.

4. The method of claim 3 wherein said re-estimating is accomplished using redundant information in said sequences.

5. The method of claim 4 wherein said re-estimating is accomplished by using information about one or more previously received sequences.

6. The method of claim 4 wherein said coded information contains speech information.

7. The method of claim 4 wherein said coded information contains visual image information.

8. The method of claim 3 wherein

said information sequences are encoded using at least an outer code to introduce redundant information,

said processing includes the further step of decoding said outer code, and

wherein said re-estimating is accomplished by selecting as the transmitted sequence that one of said L candidates which most nearly reflects an error-free decoding of said outer code.

9. The method of claim 8 wherein

said outer code is a block code, and wherein

said re-estimating is further accomplished by selecting that one of said L candidates which upon decoding of said outer code yields a self-consistent set of parity signals.

10. The method of claim 1 wherein

said step of determining comprises finding the L candidates having the L highest likelihoods of being the sequence transmitted by said transmitter.

11. The method of claim 4 wherein

said coded information has been coded by a convolutional coder, and

said identifying is accomplished by comparing metrics for each of said L paths in said trellis characterizing said convolutional code.

12. The method of claim 11 wherein

said determining comprises

selecting the L paths entering each state for each stage of said trellis which have the highest partial metric.

13. The method of claim 1 further comprising

requesting a repetition of a transmitted sequence when a processed sequence is found to be unreliable.

14. The method of claim 1 further comprising iteratively evaluating said L paths having the highest values for said metric.

15. The method of claim 14 wherein

the ith best path, i=2, . . . , L, is determined by selecting, for all candidate paths at stage j, and all j=0 to j=N+M the candidate path which first merges at stage with the path having the highest value for said metric, and has the ith largest partial metric of the paths entering stage j.

16. The method of claim 14 wherein said step of iteratively evaluating comprises

determining the path having the highest metric,

determining, for each l=2, 3, . . . , L, the trellis stage at which the lth best path remerges with the best path.

17. The method of claim 14 wherein L=2.

18. The method of claim 1 wherein L=2.

19. A method of processing received coded information sequences from a transmitter which received sequences correspond to respective paths through a trellis structure having a plurality of stages and a plurality of states at each stage comprising

determining the value for a metric indicative of the relative quality for the best candidate to be selected as the sequence actually transmitted, and

on L-1 subsequent evaluations of said metric for paths traversing the complete trellis associated with said sequences received from said transmitter, determining the metric value for the L-1 next best candidates,

said subsequent evaluations comprising evaluating metrics for partial paths that remerge for the first time at a stage j of said trellis with the path for the best candidate after having diverged from said path for the best candidate, and retaining the partial paths having the best metric at stage j and discarding all others, and

at the final stage selecting the path with the second best metric.

20. The method of claim 19 further comprising

generating an indication that said best candidate is unreliable whenever said metric for said best path differs by less than a predetermined amount from any of said (L-1) next best candidates.

21. The method of claim 20 further comprising

requesting that the transmitted sequence be retransmitted.

22. The method of claim 20 wherein L=2.

23. The method of claim 20 further comprising

re-estimating the transmitted sequence whenever the best candidate is indicated to be unreliable.