CN112615629B

CN112615629B - Decoding method, system, medium, equipment and application of multi-element LDPC code

Info

Publication number: CN112615629B
Application number: CN202011342379.6A
Authority: CN
Inventors: 侯典浩; 万飞; 白宝明; 朱敏; 刘震
Original assignee: Xidian University
Current assignee: Xidian University
Priority date: 2020-11-26
Filing date: 2020-11-26
Publication date: 2023-09-26
Anticipated expiration: 2040-11-26
Also published as: CN112615629A

Abstract

The application belongs to the technical field of wireless communication, and discloses a decoding method, a system, a medium, equipment and application of a multi-element LDPC code, wherein the iteration number of an IJDD decoder is set to be l, the iteration number of a neural network and an IJDD algorithm is set to be l, and a receiving end sends a received sequence with the length of n after channel noise addition to the IJDD decoder; after hard judgment, checking the obtained correction result, if the correction result passes the check, ending the decoding, otherwise, correcting the vector with the length of n by using a neural network, and finally outputting a complex vector with the length of n; after hard judgment, checking the complex vector with the length of n, if the complex vector passes the check, decoding is finished, and outputting a decoding result, otherwise, retransmitting the complex vector into the IJDD decoder, and decoding again until the maximum iteration number is reached; and outputting a decoding result after decoding is finished. The application can assist the decoding algorithm of the existing IJDD and can achieve the purpose of improving the noise resistance.

Description

Decoding method, system, medium, equipment and application of multi-element LDPC code

Technical Field

The application belongs to the technical field of wireless communication, and particularly relates to a decoding method, a system, a medium, equipment and application of a multi-element LDPC code.

Background

At present: compared with binary LDPC codes, the multi-element LDPC codes have more obvious advantages and higher coding gain, but due to the transition from binary to multi-element, the very high decoding complexity of the multi-element LDPC codes always prevents the multi-element LDPC codes from being popularized in practice. It is therefore important to reduce the decoding complexity of multi-LDPC codes with as little or no loss of their error resistance. The neural network constructed by deep learning can easily capture the related information between data and is similar to error correction codes in mathematics and other contexts. The neural network has better performance and fault tolerance capability, and obviously reduces the computational complexity, so the neural network is gradually applied to the field of coding and decoding, and successfully reduces the computational complexity of a decoding algorithm and improves the decoding performance.

An iterative joint detection decoding (Iterative Joint Detection Decoding, IJDD) algorithm designed for multi-LDPC codes combines multi-LDPC decoding with hard message transfer based signal detection, greatly reducing decoding complexity, but with the reduction in complexity, certain performance losses are incurred. The IJDD algorithm of the multi-element LDPC code based on deep learning is perfected based on the problem, the iteration of the hard message in the algorithm is suitable for a neural network, the assistance of the neural network is also needed, and the performance is improved while the complexity is reduced.

For example, the authors MinZhu, quan Guo, baoming Bai, and Xiao Ma published in journal IEEE Transactions on Communications, search No. 000368353700002, entitled Reliability-based joint detection-decoding algorithm for nonbinary LDPC-coded modulation systems, disclose an iterative joint detection decoding method for multi-LDPC codes, which comprises the following steps:

(1) Performing maximum likelihood judgment on each signal in the received signal vector y to obtain an independent detection signal

(2) For detection signalsDemapping to obtain the input z, { z of the decoder _j J=0, 1,..n-1 } is each element in the vector.

(3) Obtaining the checksum of the check symbol according to the input z of the decoder and the check matrix H

(4) Variable node v for corresponding position _j Making an estimation so as to obtain a check characterThe checksum of the number is equal to 0, v _j The estimate of (2) can be described asThis process is called message update criteria for check nodes.

(5) For variable node v _j Estimated values received from all check nodes adjacent theretoVoting to obtain z _j Reliability measure translated into symbol a +.>I.e. the number of votes obtained, and recording the element a having the greatest reliability _max Update result as variable node +.>The highest number of votes f obtained _max And the next highest vote count f _sub And their difference Δf _j Will->And transmitting the variable nodes to a detector to finish updating the variable nodes.

(6) The task of the detector is based on messages sent from variable nodesUpdating the received vector y ^(k) Generating y ^(k+1) The method comprises the steps of carrying out a first treatment on the surface of the Then delivers the updated hard decision vector z ^(k+1) And giving variable nodes, wherein k represents the iteration times.

(7) Detecting node update criteria: for j=0, 1..n-1, ifMapping of->ThenWherein->

Otherwise

Although the method uses iterative joint detection decoding, the loss in decoding performance caused by the reduction of decoding complexity is not improved, so that the performance of the conventional IJDD decoding scheme needs to be improved.

Through the above analysis, the problems and defects existing in the prior art are as follows: the communication frame error rate in the prior art is high.

The difficulty of solving the problems and the defects is as follows: the prior art fails to improve performance without reducing complexity.

The meaning of solving the problems and the defects is as follows: in that the complexity is reduced so that it has the potential for practical use, not just for algorithm research.

Disclosure of Invention

Aiming at the problems existing in the prior art, the application provides a decoding method, a system, a medium, a device and an application of a multi-element LDPC code.

The application is realized in such a way that a decoding method of a multi-element LDPC code comprises:

setting the iteration number of the IJDD decoder as l and the iteration number of the neural network and the IJDD algorithm as l, and sending the received sequence with the length of n after channel noise addition to the IJDD decoder by the receiving end, wherein the noise added sequence can be used as obtained data to be sent to the decoder for decoding or fed to the neural network for training; the method comprises the steps of carrying out a first treatment on the surface of the

The obtained correction result is subjected to hard judgment and then is checked, if the correction result passes the check, decoding is finished, and if the result output by decoding passes the check, the following steps are not needed, so that the iteration times are reduced, and the calculation complexity of an algorithm is reduced;

otherwise, correcting the vector with the length of n by using a neural network, and reducing the number of error bits by using a convolutional neural network mode in deep learning;

performing hard judgment on the complex vector with the length of n after the neural network correction, and performing verification, if the complex vector passes the verification, finishing decoding, outputting a decoding result, otherwise, re-transmitting the complex vector into an IJDD decoder, and performing decoding again until the maximum iteration number is reached;

and outputting a decoding result after decoding is finished.

Further, the decoding method of the multi-element LDPC code sets the iteration number of the IJDD decoder as l and the iteration number of the neural network and the IJDD algorithm as l, and the receiving end sends the received sequence y with the length of n after channel noise addition to the IJDD decoder; and (3) performing hard judgment on y, and continuously correcting by using an IJDD decoding algorithm on the basis of y to obtain a vector y' with the length of n.

Further, the decoding method of the multi-element LDPC code is to carry out verification after hard judgment on the obtained correction result, and if the correction result passes the verification, the decoding is ended: hard decision is carried out on y' to obtain a hard decision vector with the length of nFor->And checking, and outputting a decoding result if the direct jump translation code is finished through checking.

Further, the decoding method of the multi-element LDPC code modifies y' by using a neural network otherwise: since y 'is a complex vector and the neural network input is a real vector with a length of 2n, each element in y' needs to be sequentially unfolded into a real part and an imaginary part to obtain a vectorWill->Feeding a trained neural network, network pair +.>Correction is carried out to obtain a real vector +.>Every two elements in the real vector with the length of 2n of the real vector are respectively taken as the real part and the imaginary part of the complex vector to be recovered, and the complex vector with the length of n is obtained>

Further, the decoding method of the multi-LDPC code corrects the complex vector with the length of n after the neural network is correctedMaking hard decision to obtain a hard decision vector of length n->For example, cross->Outputting the decoding result by checking the end of the entering decoding, otherwise +.>The re-transmission to the IJDD decoder enters a sequence with the length of n to be transmitted to the IJDD decoder until the maximum iteration number is ended, then enters the decoding end, and outputs a decoding result.

Further, the decoding method of the multi-LDPC code is finished, and a decoding result is output

It is a further object of the present application to provide a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of:

setting the iteration number of the IJDD decoder as l and the iteration number of the neural network and the IJDD algorithm as l, and sending the received sequence with the length of n after channel noise addition to the IJDD decoder by the receiving end;

performing verification after hard judgment on the obtained correction result, and ending decoding if the correction result passes the verification;

otherwise, correcting the vector with the length of n by using a neural network;

and outputting a decoding result after decoding is finished.

Another object of the present application is to provide a computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:

and outputting a decoding result after decoding is finished.

Another object of the present application is to provide a wireless communication information data processing terminal for implementing the decoding method of the multi-LDPC code.

Another object of the present application is to provide a decoding system of a multi-LDPC code implementing the decoding method of a multi-LDPC code, the decoding system of a multi-LDPC code comprising:

the iteration number setting module is used for setting the iteration number of the IJDD decoder as l and the iteration number of the neural network and the IJDD algorithm as l;

the IJDD decoding algorithm module is used for sending the sequence with the length of n to the IJDD decoder;

the neural network correction module is used for correcting the vector with the length of n obtained by correction by using the neural network;

the verification module is used for verifying the vector result obtained after each correction and hard judgment, terminating other decoding steps in advance through verification, and if not, continuing the decoding operation;

and the decoding result output module is used for outputting a decoding result after decoding is finished.

By combining all the technical schemes, the application has the advantages and positive effects that: in the decoding process of the multi-element LDPC code, the neural network structure is utilized to learn the correlation of the code words through training so as to assist the existing IJDD decoding algorithm, and the aim of improving the noise resistance can be achieved. By further optimizing the convolutional neural network, better performance can be obtained and the complexity is lower.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments of the present application will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flowchart of a decoding method of a multi-element LDPC code according to an embodiment of the present application;

fig. 2 is a schematic structural diagram of a decoding system of a multi-element LDPC code according to an embodiment of the present application;

in fig. 2: 1. an iteration number setting module; 2. IJDD decoding algorithm module; 3. a neural network correction module; 4. a verification module; 5. and the decoding result output module.

Fig. 3 is a flowchart of a decoding method of a multi-element LDPC code according to an embodiment of the present application.

Fig. 4 is a simulated comparison diagram of bit error rates of communication with the IJDD decoding method of the present application for the conventional multi-LDPC code.

Detailed Description

The present application will be described in further detail with reference to the following examples in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

Aiming at the problems existing in the prior art, the application provides a decoding method, a system, a medium, equipment and application of a multi-element LDPC code, and the application is described in detail below with reference to the accompanying drawings.

As shown in fig. 1, the decoding method of the multi-element LDPC code provided by the present application includes the following steps:

s101: setting the iteration number of the IJDD decoder as l and the iteration number of the neural network and the IJDD algorithm as l, wherein the receiving end receives a sequence with the length n of 42 after channel noise addition;

s102: a sequence of length n of 42 is fed to the IJDD decoder: performing hard judgment on a sequence with the length n of 42, continuously correcting by using an IJDD decoding algorithm on the basis of the sequence with the length n of 42 to obtain a vector with the length n of 42, performing hard judgment on the corrected vector with the length n of 42 to obtain a hard judgment vector with the length n of 42, checking the obtained hard judgment vector with the length n of 42, and directly jumping to S105 if checking is performed;

s103: correcting the vector with the length n of 42 by using a neural network: since the vector with length n of 42 is a complex vector and the neural network input is a real vector with length 2n of 84, each element in the real vector with length 2n of 84 needs to be sequentially unfolded into a real part and an imaginary part to obtain a vector; feeding the vector obtained by expanding the real part and the imaginary part into a trained neural network, correcting the vector obtained by expanding the real part and the imaginary part by the network to obtain a real number vector with the length of 2n being 84, and entering S104;

s104: taking each two elements in the real number vector with the length of 2n being 84 as the real part and the imaginary part of the complex number vector to be recovered respectively, obtaining the complex number vector with the length of 42, carrying out hard decision on the complex number vector with the length of 42 to obtain the hard decision vector with the length of 42, if the hard decision vector with the length of 42 passes through the check, entering S105, otherwise, retransmitting the hard decision vector with the length of 42 into the IJDD decoder again to enter S102 until the maximum iteration number is reached, and entering S105;

s105: and outputting a decoding result after decoding is finished.

Those skilled in the art may also implement other steps in the decoding method of the multi-LDPC code provided by the present application, and the decoding method of the multi-LDPC code provided by the present application of fig. 1 is merely a specific embodiment.

As shown in fig. 2, the decoding system of the multi-element LDPC code provided by the present application includes:

the iteration number setting module 1 is used for setting the iteration number of the IJDD decoder as l and the iteration number of the neural network and the IJDD algorithm as l;

the IJDD decoding algorithm module 2 is used for sending the sequence with the length of 42 to an IJDD decoder;

the neural network correction module 3 is used for correcting the vector with the length of 42 obtained by correction by using a neural network;

the verification module 4 is used for verifying the vector result obtained after each correction and hard judgment, terminating other decoding steps in advance through verification, and if not, continuing the decoding operation;

and the decoding result output module 5 is used for outputting a decoding result after decoding is finished.

The technical scheme of the application is further described below with reference to the accompanying drawings.

As shown in fig. 3, a specific embodiment of the decoding method of a multi-element LDPC code provided by the present application includes the following steps:

the first step, initializing, namely setting the iteration times of the IJDD decoder as l and setting the iteration times of the neural network and the IJDD algorithm as l;

step two, receiving the signal vector subjected to channel noise addition, wherein the receiving end receives a sequence y with the length n of 42 after channel noise addition;

thirdly, sending the signal vector to an IJDD decoder for decoding, and sending y to the IJDD decoder: performing hard judgment on y, and continuously correcting by using an IJDD decoding algorithm on the basis of y to obtain a vector y' with the length n of 42;

fourth, hard decision is carried out on the decoded result, and hard decision is carried out on y' to obtain a hard decision vector with the length n of 42For->Checking, namely judging whether the hard judgment result passes a check equation, if so, directly jumping to the seventh step, otherwise, entering the fifth step;

fifthly, sending the decoding result into a convolutional neural network, and correcting y' by using the neural network: since y 'is a complex vector and the neural network input is a real vector with length 2n of 84, each element in y' needs to be sequentially expanded by real part and imaginary part to obtain a vectorWill->Feeding a trained neural network, network pair +.>Correction is performed to obtain a real vector +.>For real number vector->Each two elements of the complex vector are respectively used as a real part and an imaginary part of the complex vector to be recovered, and the complex vector with the length of n being 42 is obtained>

Sixth, hard decision is made on the complex vector recovered from the network, and thenMaking hard decision to obtain a hard decision vector with length n of 42->Judging whether the hard judgment result meets the check equation, such as +.>Through checking, the seventh step is carried out, otherwise, whether the maximum iteration times are reached or not is judged, if so, the seventh step is carried out, otherwise, the method is carried out>Re-transmitting the data to the IJDD decoder to enter a third step;

seventh, decoding is finished, and decoding result is output

The number of nodes in each layer of the convolutional neural network, the number of convolutional kernels, the size and other parameters are shown in table 1.

TABLE 1

Wherein ReLU activation function is denoted by f (x) =max (0, x).

The technical effects of the present application will be described in detail with reference to simulation.

1. Simulation conditions and content:

the simulation environment refers to table 2, and according to the simulation parameters of table 3, the comparison simulation of the transmission bit error rate is performed on the existing decoding scheme and the present application, and the result is shown in fig. 4.

TABLE 2 simulation Environment

Parameters (parameters)	Description of the application
		Simulation environment	Windows7 python3.6
Network building tool	TensorFlow2.0
		Display card information	1080ti

TABLE 3 simulation parameters

2. Simulation result analysis:

as shown in fig. 4, the horizontal axis represents the signal-to-noise ratio of the transmission channel, and the vertical axis represents the bit error rate per transmission. The three curves in the figure are the performance of the IJDD decoding algorithm iterated once under the relevant noise, the performance of the IJDD algorithm assisted by the CNN network and the IJDD decoding after the CNN network. Through the performance curve of the CNN network auxiliary IJDD algorithm in the figure, it can be found that the performance of the IJDD algorithm can be improved by about 0.5dB in the network redesigned by taking the characteristics of input data into consideration, and the soft information received from a channel by a receiving end is optimized by using a neural network, and then the IJDD algorithm is performed, so that the performance seems to be better, and the performance of the IJDD algorithm has about 0.7dB gain compared with that of an IJDD decoding algorithm under the condition of medium and low signal to noise ratio. The decoding method of the present application is described as being capable of reducing the frame error rate as compared with the existing decoding method.

It should be noted that the embodiments of the present application can be realized in hardware, software, or a combination of software and hardware. The hardware portion may be implemented using dedicated logic; the software portions may be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or special purpose design hardware. Those of ordinary skill in the art will appreciate that the apparatus and methods described above may be implemented using computer executable instructions and/or embodied in processor control code, such as provided on a carrier medium such as a magnetic disk, CD or DVD-ROM, a programmable memory such as read only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The device of the present application and its modules may be implemented by hardware circuitry, such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., as well as software executed by various types of processors, or by a combination of the above hardware circuitry and software, such as firmware.

The foregoing is merely illustrative of specific embodiments of the present application, and the scope of the application is not limited thereto, but any modifications, equivalents, improvements and alternatives falling within the spirit and principles of the present application will be apparent to those skilled in the art within the scope of the present application.

Claims

1. A method for decoding a multi-LDPC code, the method comprising:

outputting a decoding result after decoding is finished;

the decoding method of the multi-element LDPC code is characterized in that the obtained correction result is hard judged and then checked, and if the correction result passes the check, the decoding is finished: hard decision is carried out on y' to obtain a hard decision vector with the length of nFor->Checking, if the direct jump translation code is finished through checking, outputting a decoding result;

the decoding method of the multi-element LDPC code is characterized in that the iteration number of the IJDD decoder is set to be l, the iteration number of the neural network and the IJDD algorithm is set to be l, and a receiving end sends a received sequence y with the length of n after channel noise addition to the IJDD decoder; carrying out hard judgment on y, and continuously correcting by using an IJDD decoding algorithm on the basis of y to obtain a vector y' with the length of n;

the decoding method of the multi-LDPC code uses a neural network to correct y': since y 'is a complex vector and the neural network input is a real vector of length 2n, each element in y' needs to be sequentially subjected to real part and imaginary partExpanding to obtain vectorWill->Feeding a trained neural network, network pair +.>Correction is carried out to obtain a real vector +.>Every two elements in the real vector with the length of 2n of the real vector are respectively taken as the real part and the imaginary part of the complex vector to be recovered, and the complex vector with the length of n is obtained>

2. The decoding method of multi-LDPC code according to claim 1, wherein the decoding method of multi-LDPC code corrects a complex vector of length n by a neural networkMaking hard decision to obtain a hard decision vector of length n->For example, cross->Outputting the decoding result by checking the end of the entering decoding, otherwise +.>Re-entering the IJDD decoder enters the sequence of length n into the IJDD decoder,and after the maximum iteration times are reached, the decoding is finished, and a decoding result is output.

3. The decoding method of the multi-LDPC code of claim 1, wherein the decoding method of the multi-LDPC code ends decoding and outputs a decoding result

4. A computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the method of any of claims 1 to 3.

5. A computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of the method of any one of claims 1 to 3.

6. A wireless communication data processing terminal for implementing the decoding method of the multi-LDPC code as claimed in any one of claims 1 to 3.

7. A decoding system of a multi-LDPC code that implements the decoding method of a multi-LDPC code according to any one of claims 1 to 3, characterized in that the decoding system of a multi-LDPC code comprises:

the IJDD decoding algorithm module is used for sending a sequence with the length of n into IJDD decoding;