CN101841338B - Method for ending iterative decoding in advance by using low-density parity-check code - Google Patents

Abstract

The invention discloses a method for early termination of iterative decoding of a low-density parity-check code. The method includes: during each iteration of the iterative decoding of the low-density parity-check code, calculating the number of non-zero elements in the check matrix Number N _i ; Calculate the number of non-zero elements in the check matrix according to the number of non-zero elements in the check matrix in this and the previous iteration process. n _i =N _i-1 N _i ; record In the decoding process, the number of negative values of ni _appears , denoted as num, stop iterating and outputting codewords when the number of iterations is greater than zero and num=2 or 3, or when the number of iterations is greater than half of the maximum number of iterations and num=2 or 3, stop iteratively outputting codewords. The invention effectively reduces the complexity of the low-density parity-check code ending iteration criterion in advance, and greatly reduces the average number of iterations under the condition of small performance loss, which is beneficial to hardware implementation.

Description

A method for early termination of iterative decoding of low-density parity-check codes

technical field

The invention relates to the technical field of channel coding and decoding in the communication field, in particular to a method for early termination of iterative decoding of a low-complexity low-density parity-check code (LDPC).

Background technique

In the communication system, in order to resist various interferences in the transmission process, it is necessary to increase a certain degree of redundancy so that the transmitted information has the ability of automatic error detection or error correction. Channel coding and decoding technology introduces redundant information into information to detect and correct errors to improve the reliability of the system.

Low Density Parity Check (LDPC) code is a kind of channel coding whose performance is closest to Shannon's threshold so far [1]. Now it has been widely used in communication systems such as UWB[2], DVB-S2[3] and IEEE 802.16e[4]. LDPC decoding uses an iterative decoding algorithm, so one decoding requires multiple iterations to complete. In each iteration, the information of variable nodes and check nodes is updated to determine whether the decoding is successful. The decoder will continue to iterate until the correct codeword is decoded or the number of iterations reaches the preset maximum number of iterations. The setting of the maximum number of iterations varies according to different standards. For example, in the DVB-S2 standard, the maximum number of iterations is set to 40, while in the CMMB standard, the maximum number of iterations is 30.

Generally speaking, when the signal-to-noise ratio is relatively high, the LDPC decoder only needs a small number of iterations to solve the correct codeword. However, when the SNR is relatively low, even if the number of iterations reaches the maximum number of iterations, the decoder still cannot solve the correct codeword, which consumes a lot of calculation and increases the power consumption of the decoder. Therefore, it is very important to end those unnecessary iterative operations in advance to reduce the power consumption of the decoder.

references:

[1] Xu Hua. Research on LDPC code related technologies [J]. Microelectronics and Computers, 2005, 22(12): 178.

[2] C.-C.Lin, K.-L.Lin, C.-C.Chung, and C.-Y.Lee, "A 3.33Gb/s(1200, 720) low-density parity check code decoder ,” in Proc. ESSCIRC, 2005, pp.211-214.

[3]European Telecommunications Standards Institude(ETSI)，“DigitalVideo Broadcasting(DVB)Second generation framing structure，channelcoding and modulation systems for Broadcasting，Interactive Services，NewsGathering and other broadband satellite applications；TM 2860r1DVBS2-74r8，”www.dvb.org .

[4] T.Brack, M.Alles, F.Kienle, and N.When, "A synthesizable IP core for WIMAX 802.16E LDPC code decoding," in Proc.IEEE 17th Int.Symp.Personal, Indoor and Mobile Radio Communications, Sep.2006, pp.1-5.

Contents of the invention

(1) Technical problems to be solved

In view of this, the main purpose of the present invention is to provide a method for LDPC early termination of iterative decoding, so as to reduce the complexity of LDPC early termination of iteration criteria, and reduce the average number of iterations with less performance loss.

(2) Technical solutions

In order to achieve the above object, the present invention provides a method for early termination of iterative decoding of low-density parity-check codes, the method comprising:

When iteratively decoding the low-density parity-check code, calculate the number N _i of non-zero elements in the check matrix during each iteration;

According to the number of non-zero elements in the check matrix in this and the last iteration, calculate the reduction of the number of non-zero elements in the check matrix n _i =N _i-1 - _{N i} for each iteration;

Record the number of negative values of _ni in the decoding process, denoted as num, stop iterating and outputting codewords when the number of iterations is greater than zero and num=2 or 3, or when the number of iterations is greater than half of the maximum number of iterations and hum=2 or 3, stop iteratively outputting codewords.

In the above solution, when the number of iterations is greater than zero and num=2 or 3, the stopping of iteratively outputting codewords is performed when the system can bear relatively large performance loss.

In the above solution, the situation that the system can withstand relatively large performance loss means that the signal-to-noise ratio noise threshold value is increased by 0.5~ 1dB.

In the above solution, when the number of iterations is greater than half of the maximum number of iterations and num=2 or 3, stopping the iterative output of codewords is performed when the system can bear relatively small performance loss.

In the above solution, the situation that the system can withstand relatively small performance loss refers to: the signal-to-noise ratio noise threshold value is increased by 0~ 0.5dB.

(3) Beneficial effects

As can be seen from the foregoing technical solutions, the present invention has the following beneficial effects:

1. The method for ending iterative decoding of this LDPC in advance provided by the present invention can easily determine the convergence of the iterative decoding of the LDPC decoder by observing the number of non-zero elements in the parity check matrix, which is different from various current LDPC Compared with the criterion of ending the iteration ahead of time, the convergence of the iterative decoding of the LDPC decoder can be determined without complex calculations.

2. The LDPC method for ending iterative decoding in advance provided by the present invention has the characteristics of low complexity, and can greatly reduce the average number of iterations with less performance loss.

3. The LDPC method of the present invention ends the iterative decoding in advance, according to the actual needs of the system, the compromise between the performance and the average number of iterations is achieved by adjusting the number of iterations when starting to calculate num and the value of num.

Description of drawings

Below in conjunction with accompanying drawing and embodiment the present invention is further described:

Fig. 1 is the method flow chart of LDPC early end iterative decoding provided by the present invention;

Fig. 2 is according to an embodiment of the present invention under the CMMB standard 800 * 9216 bit error rate curves of LDPC codewords under the given Case1～5 situation, simulation condition is no RS sign indicating number, 6bit-LDPC, code rate R= 1/2, the mode of modulation and demodulation is 16QAM soft decision;

Fig. 3 is according to an embodiment of the present invention under the CMMB standard 800 * 9216 LDPC codewords curve of the average number of iterations under given Case1～5 situation, simulation condition is no RS sign indicating number, 6bit-LDPC, code rate R= 1/2, the mode of modulation and demodulation is 16QAM soft decision.

Case1: Do not use early end iteration (regular)

Case2: There are two sign inversions (num=2)

Case3: There are three sign inversions (num=3)

Case4: There are two sign inversions after the number of iterations is greater than 15 (iter>15&&num=2)

Case5: Three sign inversions occur after the number of iterations is greater than 15 (iter>15&&num=3)

Detailed ways

The purpose of the present invention is to improve the condition of iteration stop in various current LDPC early termination iterative decoding methods, and propose a low-complexity LDPC early termination iterative decoding method. Compared with the current LDPC early termination iteration criteria, the method can determine the convergence of the iterative decoding of the LDPC decoder without complex calculations.

As shown in Figure 1, Figure 1 is a flowchart of the realization of the method of LDPC early termination iterative decoding provided by the present invention, the method includes:

Step 101: During LDPC decoding, calculate the number N _i of non-zero elements in the parity check matrix during each iteration;

Step 102: According to the number of non-zero elements in the check matrix in the current and previous iterations, calculate the amount of reduction n _i =N _i-1 - _{N i} for each iteration to make the number of non-zero elements in the check matrix;

Step 103: record the number of negative values of n _i in the decoding process, which is recorded as num, if the system can bear a relatively large performance loss, then when the number of iterations iter is greater than zero, in the case of num=2 or 3 Next, stop iteratively outputting codewords; if the system can bear a relatively small performance loss, then stop iteratively outputting codewords when num=2 or 3 when the number of iterations iter is greater than half of the maximum number of iterations.

Wherein, the situation that the system can withstand relatively large performance loss means that the SNR threshold value is increased by 0.5-1 dB compared with the SNR threshold value when the iteration is not terminated early under the same system condition. For example, if the system can withstand the increase of the SNR threshold of 0.6dB, then iterative output codewords can be stopped when hum=2 or 3 when the iteration number iter is greater than zero.

The situation that the system can withstand relatively small performance loss means that the SNR threshold value is increased by 0-0.5dB compared with the SNR threshold value when the iteration is not terminated early under the same system condition. For example, if the system can only bear the increase of the SNR threshold of 0.3dB at most, then the iterative output codeword can be stopped when num=2 or 3 when the number of iterations iter is greater than half of the maximum number of iterations.

Fig. 2 has shown according to an embodiment of the present invention under the CMMB standard 800 * 9216 bit error rate curves of LDPC codewords under the given Case1～5 situation, simulation condition is no RS sign indicating number, 6bit-LDPC, code rate R=1/2, the mode of modulation and demodulation is 16QAM soft decision.

Case1: Do not use early end iteration (regular)

Case2: There are two sign inversions (num=2)

Case3: There are three sign inversions (num=3)

Case4: There are two sign inversions after the number of iterations is greater than 15 (iter>15&&num=2)

Case5: Three sign inversions occur after the number of iterations is greater than 15 (iter>15&&num=3)

Assuming that the lowest SNR that can obtain the bit error rate is the SNR threshold of correct decoding, it can be seen from Figure 2 that Case2 and Case3 have a large loss in performance, and the SNR threshold of correct decoding The value has changed greatly, around 0.5dB. In terms of performance, Case4 and Case5 are closest to the case where Case1 does not use early end generation. There is only a small difference when the SNR is between 6 and 7, and the noise ratio threshold of correct decoding has a fluctuation of 0 to 0.1dB. .

Fig. 3 is according to an embodiment of the present invention under the CMMB standard 800 * 9216 LDPC codewords curve of the average number of iterations under given Case1～5 situation, simulation condition is no RS sign indicating number, 6bit-LDPC, code rate R= 1/2, the mode of modulation and demodulation is 16QAM soft decision.

Case1: Do not use early end iteration (regular)

Case2: There are two sign inversions (num=2)

Case3: There are three sign inversions (hum=3)

Case4: There are two sign inversions after the number of iterations is greater than 15 (iter>15&&num=2)

Case5: Three sign inversions occur after the number of iterations is greater than 15 (iter>15&&num=3)

Assuming that the lowest SNR that can obtain the bit error rate is the SNR threshold of correct decoding, it can be seen from Figure 3 that Case2 and Case3 greatly reduce the average iteration of decoding when the SNR is relatively low The number of times, the average number of iterations is reduced by about 25 times when the SNR is 1-6dB. In Case4 and Case5, the average number of iterations is also greatly reduced when the signal-to-noise ratio is relatively low, and the average number of iterations is reduced by about 10 times when the SNR is 1-6dB. In the case of a relatively high signal-to-noise ratio, such as when the SNR is 7-10 dB, the average number of iterations of Case2-Case5 and Case1 without early termination of the iteration is basically the same.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (1)

1. A method for ending iterative decoding of a low-density parity-check code ahead of schedule, characterized in that the method comprises: When iteratively decoding the low-density parity-check code, calculate the number N _i of non-zero elements in the check matrix during each iteration; According to the number of non-zero elements in the parity check matrix in this and the last iteration, calculate the amount of reduction n _i =N _i-1 - _{N i} in the number of non-zero elements in the parity check matrix for each iteration; and Record the number of negative values of _ni in the decoding process, which is recorded as num, stop iterating and outputting codewords when the number of iterations is greater than zero and num=2 or 3, or when the number of iterations is greater than half of the maximum number of iterations and num=2 or 3, stop iteratively outputting codewords; Wherein, in the step of stopping the iterative output codeword when the number of iterations is greater than zero and num=2 or 3, or stopping the iterative output codeword when the number of iterations is greater than half of the maximum number of iterations and num=2 or 3 , if the system can bear a relatively large performance loss, when the number of iterations iter is greater than zero, stop iteratively outputting codewords when num=2 or 3; if the system can bear a relatively small performance loss, then When the number of iterations iter is greater than half of the maximum number of iterations, stop the iterative output codeword under the situation of num=2 or 3; The situation where the system can withstand relatively large performance loss refers to: the signal-to-noise ratio noise threshold value is increased by 0.5-1dB compared with the signal-to-noise ratio-noise ratio threshold value when the iteration is not terminated early in the case of the same system; The situation that the system can withstand relatively small performance loss means that the SNR threshold value is increased by 0-0.5dB compared with the SNR threshold value when the iteration is not terminated in advance under the same system condition.

Images