Source-Channel Approach to Channel Coding with Side Information

Abstract

Code designs for channel coding with side information (CCSI) based on combined source-channel coding are disclosed. These code designs combine trellis-coded quantization (TCQ) with irregular repeat accumulate (IRA) codes. The EXIT chart technique is used for IRA channel code design (and especially for capacity-approaching IRA channel code design). We emphasize the role of strong source coding and endeavor to achieve as much granular gain as possible by using TCQ. These code designs synergistically combine TCQ with IRA codes. By bringing together TCQ and EXIT chart-based IRA code designs, we are able to approach the theoretical limit of dirty-paper coding.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

The following detailed description makes reference to the accompanying drawings, which are now briefly described.

FIG. 1 illustrates the general framework of channel coding with side information (CCSI).

FIG. 2 illustrates one embodiment of a method M_NE for the encoding of message information. The method M_NE involves both trellis-coded quantization (TCQ) and irregular repeat accumulate (IRA) encoding.

FIG. 3 illustrates one embodiment of a system S_NE for encoding message information that involves both trellis-coded quantization (TCQ) and irregular repeat accumulate (IRA) encoding.

FIG. 4 illustrates one embodiment of an iterative method M_ND for recovering message information from received signal information.

FIG. 5 illustrates one embodiment of a system S_ND for recovering message information from received signal information.

FIG. 6 illustrates one embodiment of a method M_SE for encoding message information that involves both a systematic portion and a parity portion.

FIG. 7 illustrates one embodiment of a system S_SE for encoding message information that involves both a systematic portion and a parity portion.

FIG. 8 illustrates one embodiment of a method M_SD for recovering message information from received signal information that involves both a systematic portion and a parity portion.

FIG. 9 illustrates one embodiment of a system S_SD for recovering message information from received signal information that involves both a systematic portion and a parity portion.

FIG. 10 is a table illustrating the modulo loss SNR_M for TCQ of different numbers of states and the predicted total performance loss ΔSNR in embodiments of our dirty-paper code designs, assuming the packing loss SNR_P from IRA codes is 0.34 dB and a target rate of C*=0.25b/s.

FIG. 11 is block diagram of one embodiment of a dirty-paper encoder based on TCQ and the non-systematic IRA code.

FIG. 12 is a block diagram of one embodiment of a decoder with TCQ and the non-systematic IRA code.

FIG. 13 illustrates VND (variable node decode) EXIT charts with different variable node degrees and with 256-state TCQ.

FIG. 14 illustrates EXIT charts of the non-systematic IRA code at SNR=−2.844 dB with K=60,000 and N=240,000 and with 256-state TCQ.

FIG. 15 illustrates EXIT charts of the non-systematic IRA code at SNR=−2.993 dB with K=22,500 and N=90,000 bits and with 1024-state TCQ.

FIG. 16 is a block diagram of one embodiment of a dirty-paper encoder based on TCQ and the systematic IRA code.

FIG. 17 is a block diagram of one embodiment of a decoder with TCQ and the systematic IRA code.

FIG. 18 illustrates EXIT charts of the systematic IRA code at SNR=−2.844 dB with K=60,000 and N=180,000. Both TCQ₁ and TCQ₂ have 256 states.

FIG. 19 illustrates EXIT charts of the systematic IRA code at SNR=−2.945 dB with K=30,000 and N=90,000 bits. TCQ₁ has 512 states and TCQ₂ has 1024 states.

FIG. 20 is a PDF of quantization error X with the 256-state TCQ in embodiments of our dirty-paper code designs, together with a Gaussian PDF.

Claims

1. A method for encoding data, the method comprising: performing IRA-based channel coding on a set of message bits in order to generate a stream of expanded parity words;performing source coding on the stream of expanded parity words to generate a modulated signal, wherein said performing source coding includes modifying the expanded parity words using output from a trellis coded quantization;wherein the modulated signal is usable, in conjunction with a channel state signal, to generate an output signal X.

2. The method of claim 1, wherein the output from the trellis-coded quantization is generated in response to an input vector b supplied to the trellis-coded quantization, wherein the input vector b has been selected so that the output signal X satisfies a power constraint.

3. The method of claim 1, wherein said performing source coding also includes performing pulse amplitude modulation based on the modified expanded parity words to obtain the modulated signal.

4. The method of claim 1 further comprising subtracting a scaled and dithered version of the channel state signal from the modulated signal to obtain the output signal X.

5. The method of claim 1 further comprising transmitting the output signal over a channel to one or more destinations.

6. The method of claim 1 further comprising storing the output signal in a memory medium.

7. A method for recovering message bits from an input signal, the method comprising: inner decoding an input signal in order to generate first feedforward information, wherein the inner decoding includes (a) performing trellis decoding on the input signal over a joint trellis corresponding at least to an accumulator portion of an irregular repeat accumulate (IRA) encoder and a trellis coded quantization and (b) performing first check node decoding on data including a modified version of output from the trellis decoding, wherein the first feedforward information is a modified version of output from the first check node decoding;outer decoding a permuted version of the first feedforward information in order to generate first feedback information, wherein said outer decoding includes performing variable node decoding, corresponding to variable nodes of the IRA encoder, on the permuted version of the first feedforward information, wherein the first feedback information is a modified version of output from the variable node decoding;wherein the output of the variable node decoding is usable to determine an estimate for the message bits.

8. The method of claim 7, wherein the input signal is a scaled and dithered version of a received signal.

9. The method of claim 7, wherein said inner decoding also includes performing second check node decoding on a permuted version of the first feedback information, wherein the trellis decoding uses output from the second check node decoding.

10. The method of claim 7 further comprising determining the estimate for the message bits from the output of the variable node decoding.

11. The method of claim 10 further comprising storing the estimate for the message bits in a memory medium.

12. A method for encoding data, the method comprising: performing IRA-based channel coding on message bits in order to generate a stream of expanded parity words;performing first source coding on the stream of expanded parity words to generate a first modulated signal, wherein said performing first source coding includes modifying the expanded parity words using output from a first trellis coded quantization;expanding the message bits to obtain expanded message words;performing second source coding on the stream of expanded message words to generate a second modulated signal, wherein said performing second source coding includes modifying the expanded message words using output from a second trellis coded quantization;generating a signal u from the first modulated signal and the second modulated signal, wherein the signal u is usable, in conjunction with a channel state signal, to generate an output signal X.

13. The method of claim 12, wherein the output from the first trellis-coded quantization is generated in response to an input vector bP supplied to the first trellis-coded quantization, wherein the output from the second trellis-coded quantization is generated in response to an input vector bS supplied to the second trellis-coded quantization, wherein the input vectors bP and bS have been selected so that the output signal X satisfies a power constraint.

14. The method of claim 12, wherein said performing first source coding also includes performing pulse amplitude modulation based on the modified expanded parity words to obtain the first modulated signal.

15. The method of claim 12, wherein said performing second source coding also includes performing pulse amplitude modulation based on the modified expanded message words to obtain the second modulated signal.

16. The method of claim 12 further comprising subtracting a scaled and dithered version of the channel state signal from the signal u to obtain the output signal X.

17. The method of claim 12 further comprising transmitting the output signal X over a channel to one or more destinations.

18. The method of claim 12 further comprising storing the output signal X in a memory medium.

19. A method for recovering message bits from an input signal, the method comprising: performing first decoding operations on a parity portion of the input signal in order to generate first feedforward information, wherein said performing first decoding operations includes (a) performing a first trellis decoding on the parity portion over a joint trellis corresponding at least to an accumulator portion of an irregular repeat accumulate (IRA) encoder and a first trellis coded quantization and (b) performing first check node decoding on data including a modified version of output from the first trellis decoding, wherein the first feedforward information is a modified version of output from the first check node decoding;performing second decoding operations on a systematic portion of the input signal and a permuted version of the first feedforward information in order to generate first feedback information and second feedback information, wherein said performing second decoding operations includes (c) performing a second trellis decoding on the systematic portion over a trellis corresponding to a second trellis coded quantization and (d) performing variable node decoding on the permuted version of the first feedforward information and a modified version of output from the second trellis decoding, wherein the first feedback information and the second feedback information are outputs of the variable node decoding;wherein the second feedback information is usable to determine an estimate for the message bits.

20. The method of claim 19, wherein the input signal is a scaled and dithered version of a received signal.

21. The method of claim 19, wherein said performing the second decoding operations also includes subtracting the permuted version of the first feedforward information from the second feedback information to obtain third feedback information.

22. The method of claim 21, wherein said performing first decoding operations also includes performing second check node decoding on a permuted version of the third feedback information, wherein the first trellis decoding uses output from the second check node decoding.

23. The method of claim 19, wherein said second trellis decoding uses the first feedback information.

24. The method of claim 19 further comprising determining the estimate for the message bits from the second feedback information.

25. A computer system for encoding data, the computer system comprising: a memory configured to store program instructions;a processor configured to access the program instructions from the memory and to execute the program instructions, wherein the program instructions are executable to: perform IRA-based channel coding on a set of message bits in order to generate a stream of expanded parity words;perform source coding on the stream of expanded parity words to generate a modulated signal, wherein said performing source coding includes modifying the expanded parity words using output from a trellis coded quantization;wherein the modulated signal is usable, in conjunction with a channel state signal, to generate an output signal X.

26. The computer system of claim 25, wherein the output from the trellis-coded quantization is generated in response to an input vector b supplied to the trellis-coded quantization, wherein the input vector b has been selected so that the output signal X satisfies a power constraint.

27. The computer system of claim 25, wherein said performing source coding also includes performing pulse amplitude modulation based on the modified expanded parity words to obtain the modulated signal.

28. The computer system of claim 25, wherein the program instructions are also executable to subtract a scaled and dithered version of the channel state signal from the modulated signal to obtain the output signal X.

29. The computer system of claim 25, wherein the program instructions are also executable to transmit the output signal over a channel to one or more destinations.

30. The computer system of claim 25, wherein the program instructions are also executable to store the output signal in a memory medium.

31. A computer-accessible memory medium configured to store program instructions, wherein the program instructions are executable to: perform IRA-based channel coding on a set of message bits in order to generate a stream of expanded parity words;perform source coding on the stream of expanded parity words to generate a modulated signal, wherein said performing source coding includes modifying the expanded parity words using output from a trellis coded quantization;wherein the modulated signal is usable, in conjunction with a channel state signal, to generate an output signal X.

32. The memory medium of claim 31, wherein the output from the trellis-coded quantization is generated in response to an input vector b supplied to the trellis-coded quantization, wherein the input vector b has been selected so that the output signal X satisfies a power constraint.

33. The memory medium of claim 31, wherein said performing source coding also includes performing pulse amplitude modulation based on the modified expanded parity words to obtain the modulated signal.

34. The memory medium of claim 31, wherein the program instructions are also executable to subtract a scaled and dithered version of the channel state signal from the modulated signal to obtain the output signal X.

35. The memory medium of claim 31, wherein the program instructions are also executable to transmit the output signal over a channel to one or more destinations.

37. The memory medium of claim 31, wherein the program instructions are also executable to store the output signal in a memory medium.