WO2013117076A1 - 一种迭代译码方法及系统 - Google Patents

一种迭代译码方法及系统 Download PDF

Info

Publication number
WO2013117076A1
WO2013117076A1 PCT/CN2012/078686 CN2012078686W WO2013117076A1 WO 2013117076 A1 WO2013117076 A1 WO 2013117076A1 CN 2012078686 W CN2012078686 W CN 2012078686W WO 2013117076 A1 WO2013117076 A1 WO 2013117076A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
node
decoding
variable node
check
Prior art date
Application number
PCT/CN2012/078686
Other languages
English (en)
French (fr)
Inventor
耿敏明
陈为刚
董同昕
Original Assignee
中兴通讯股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Publication of WO2013117076A1 publication Critical patent/WO2013117076A1/zh

Links

Classifications

    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/03Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
    • H03M13/05Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
    • H03M13/11Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits using multiple parity bits
    • H03M13/1102Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
    • H03M13/1105Decoding
    • H03M13/1111Soft-decision decoding, e.g. by means of message passing or belief propagation algorithms
    • H03M13/1117Soft-decision decoding, e.g. by means of message passing or belief propagation algorithms using approximations for check node processing, e.g. an outgoing message is depending on the signs and the minimum over the magnitudes of all incoming messages according to the min-sum rule
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/03Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
    • H03M13/05Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits
    • H03M13/11Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words using block codes, i.e. a predetermined number of check bits joined to a predetermined number of information bits using multiple parity bits
    • H03M13/1102Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
    • H03M13/1105Decoding
    • H03M13/1111Soft-decision decoding, e.g. by means of message passing or belief propagation algorithms
    • H03M13/1117Soft-decision decoding, e.g. by means of message passing or belief propagation algorithms using approximations for check node processing, e.g. an outgoing message is depending on the signs and the minimum over the magnitudes of all incoming messages according to the min-sum rule
    • H03M13/112Soft-decision decoding, e.g. by means of message passing or belief propagation algorithms using approximations for check node processing, e.g. an outgoing message is depending on the signs and the minimum over the magnitudes of all incoming messages according to the min-sum rule with correction functions for the min-sum rule, e.g. using an offset or a scaling factor
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/65Purpose and implementation aspects
    • H03M13/6522Intended application, e.g. transmission or communication standard
    • H03M13/6544IEEE 802.16 (WIMAX and broadband wireless access)
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/65Purpose and implementation aspects
    • H03M13/6577Representation or format of variables, register sizes or word-lengths and quantization
    • H03M13/6594Non-linear quantization

Definitions

  • the present invention relates to digital signal transmission and storage systems in the field of communications, and more particularly to an iterative decoding method and system.
  • LDPC Low Density Parity Check Code
  • LDPC is a type of linear block code defined on a sparse matrix H.
  • H is the check matrix of the LDPC.
  • the check matrix H of the LDPC can be represented by a bipartite graph.
  • the soft decision decoding algorithm based on the check matrix of the bipartite graph is the commonly used decoding method of LDPC.
  • the soft decision decoding algorithm can be implemented by iterative decoding based on the Belief Propagation (BP) decoding algorithm, and good bit error rate performance can be obtained.
  • BP Belief Propagation
  • some simplified BP decoding algorithms are generated, such as the minimum sum (MS, Min-Sum).
  • the decoding algorithm the normalized minimum sum (NMS, Normalized Min-Sum) decoding algorithm and the offset minimum sum (OMS, Offset Min-Sun) decoding algorithm.
  • the main operations used in the minimum and decoding algorithms are the comparison operation and the summation operation.
  • the algorithm complexity is very low, but its performance has a large loss compared with the BP decoding algorithm.
  • the normalization minimum and decoding algorithm And the offset minimum and decoding algorithms are both improved decoding algorithms based on the minimum and decoding algorithms.
  • the normalized minimum and decoding algorithm introduces a correction factor ⁇ of less than 1, with minimal offset and the decoding algorithm introduces an offset ⁇ .
  • the offset minimum and decoding algorithms are used in the decoder in most cases.
  • the two parts corresponding to the N columns and the M rows LDPC in the check matrix H respectively contain two types of nodes: N check nodes c and M variable nodes V, and the edges in the bipartite graph correspond to the check matrix "1" in H, if the element h(m,n) in the check matrix H is 1, there is an edge between the check node cm and the variable node vn in the bipartite graph.
  • M(n) is the set of check nodes connected to the variable node vn in the bipartite graph
  • N(m) is the set of variable nodes participating in the check equation m.
  • M(n) ⁇ m denotes the removal of the element cm in the set M(n)
  • N(m) ⁇ n denotes the element vn removed from the set N(m).
  • the definition In represents the prior probability information of the variable node vn.
  • the external information transmitted by the check node cm to the variable node vn is Lmn
  • the external information transmitted by the variable node vn to the check node cm is znm
  • the variable node vn The posterior probability information is LQn.
  • Step 1 initializing a variable node to the risk of information ⁇ l n is initialized to:
  • variable node outer information Znm l n .
  • Step 2 Verify the node update, verify that the node outside information ⁇ is updated to:
  • is a small positive number, which is the minimum offset and the offset correction factor of the decoding algorithm.
  • the offset correction factor generally determines its optimal value according to the density evolution method or simulation method.
  • LDPC decoders in hardware requires consideration of hardware resource issues, especially on-chip memory (RAM, Random Access Memory) resources and reclaimed resources.
  • RAM Random Access Memory
  • reclaimed resources In terms of memory resources, since LDPC uses the soft information iterative decoding method, a large amount of storage resources are needed to store external information in the decoding process.
  • the RAM resources inside the FPGA are very limited.
  • ASICs a large amount of on-chip RAM occupies too much chip area, which in turn squeezes the space of on-chip logic resources, limits the logic complexity of the chip, and increases the power consumption of the chip.
  • a limited precision LDPC decoding algorithm is needed in the hardware implementation process of the decoder, that is, the channel information of the input decoder needs to be quantized, and used in the decoding process. Limited precision data operation.
  • the quantization of the channel information of the input decoder is a process of representing the channel information value of the input decoder with a predetermined finite number of integer values.
  • a predetermined finite number of integer values is called a quantized value.
  • the quantized value is represented by a q-bit signed binary code, and the data bit width of the binary code is called quantization of the quantized value. Precision.
  • the highest bit of the signed binary code is the sign bit. If the quantized value is negative, the highest bit is taken.
  • the remaining q-1 bits of the binary code represent the modulus of the quantized value.
  • This requires dividing the value interval of the channel information value of the input decoder into a plurality of quantization intervals corresponding to a limited number of quantized values.
  • the end point value of the quantization section corresponding to the quantized value i i e [- ⁇ ,...,-1,0,1,..., ⁇ ]) is called a magnitude value.
  • Definition AV ⁇ qrqw is the quantization interval.
  • a quantization method in which the quantization intervals are equal is called homo-height quantization, and a quantization method in which the quantization intervals are not equal is called non-homogeneous quantization.
  • Quantifying the channel information results in a loss of decoding performance of the decoding algorithm. Both the quantization accuracy and the magnitude value in the quantization scheme affect the decoding performance loss of the decoding algorithm. Therefore, for the design of limited precision LDPC code decoding algorithm, the selection of quantization precision and magnitude in the channel information quantization scheme is an important task.
  • the main work is to balance the decoding performance loss caused by quantization and the hardware resources of the decoder. If the quantization precision of the quantization method design is high, the data in the decoding operation process has a wide data bit width, which can improve the data operation precision in the decoding process, thereby reducing the decoding performance loss caused by the quantization. However, higher quantization accuracy requires both the data bus and the external information storage unit of the decoder to have a wider data bit width. Such a hardware implementation of the decoder requires a large amount of hardware logic resources and storage resources.
  • the quantization precision of the quantization method design is low, the purpose of saving the hardware logic resources and the storage resources of the decoder can be achieved, but the data operation precision in the decoding process is also reduced, which may result in the decoding algorithm. The decoding performance is greatly lost.
  • the main work is to minimize the loss of decoding performance caused by quantization by selecting appropriate magnitude values and quantization intervals.
  • the quantization accuracy is high, the averaging method can be used.
  • the quantization precision is low, it is difficult to simultaneously ensure the coverage of the quantized data and the accuracy of the quantized data by using the uniform quantization method, and the loss of the decoding performance caused by the quantization is large, so it is necessary to use the non-uniform quantization method.
  • the quantization between smaller magnitude values can be reduced Separate, while expanding the quantization interval between larger magnitude values. This ensures the accuracy and coverage of the quantized data.
  • the decoding performance loss caused by quantization can be further reduced by optimizing the design of the magnitude value.
  • the data operation in the decoding process is a finite precision data operation.
  • the data bit width may be expanded due to the addition carry, and the data bit width of the external information storage unit of the decoder cannot be expanded.
  • the data bit width expansion data is generated during the decoding process to perform data bit width limitation.
  • the use of different data bit width limiting methods for data that produces data bit width extensions also has a large impact on the decoding performance of the decoding algorithm. For different LDPC codes, it is necessary to optimize the data bit width limiting method of the bit width extension data in the decoding process to reduce the decoding performance loss of the decoding algorithm caused by quantization.
  • an object of embodiments of the present invention is to provide an iterative decoding method and system to reduce storage resource and hardware implementation complexity of an LDPC decoder.
  • the iterative decoding system of the embodiment of the present invention includes:
  • a quantization unit configured to perform non-homogeneous quantization processing on the received channel information
  • a decoding unit configured to perform initialization processing on the prior probability information of the variable node, and use the quantized channel information as a first risk of the variable node 3 ⁇ 4 ⁇ rate information; based on the first risk information of the variable node, and using the adaptive offset to check the update processing of the information outside the node;
  • the out-of-node information is subjected to hierarchically accumulated update processing; and it is judged whether or not decoding is ended.
  • system may further include: a channel information storage unit configured to store the quantized channel information.
  • the decoding unit may include:
  • An initialization unit configured to initialize the prior probability information of the variable node,
  • the quantized channel information is used as the first risk information of the variable node;
  • a check node update unit configured to perform update processing of the check node out-of-node information according to the prior probability information of the variable node, and using the adaptive offset;
  • variable node updating unit configured to perform an update process of hierarchically accumulating the information outside the variable node according to the updated information outside the check node
  • a judging unit is arranged to judge whether the decoding ends.
  • system may further include:
  • a first storage unit configured to store updated information outside the check node
  • a second storage unit configured to store updated variable node out-of-node information.
  • An embodiment of the present invention further provides an iterative decoding method, including:
  • the step of performing the non-uniformity dequantization processing on the channel information of the input decoding unit may include:
  • the received channel information for inputting the decoding unit is subjected to N-bit non-uniform quantization processing, and the N is 4 or 5 or 6; After channel information.
  • the step of performing N-bit non-uniform hook quantization processing on the received channel information for inputting the decoding unit may include: Performing N-bit non-uniform quantization processing on the received channel information for inputting the decoding unit, and quantizing the received channel information into:
  • i is the quantized value
  • is the received channel information, which is the quantized channel information.
  • the step of performing the update processing of the out-of-node information according to the prior probability information of the variable node and using the adaptive offset may include:
  • the offset is adaptively adjusted according to the number of decoding iterations
  • the information outside the variable node associated with the check node is grouped into two groups, and the minimum value in the modulus value of the information outside the variable node is obtained by hierarchically comparing the modulus values of the information outside the variable node, and then the minimum value is obtained.
  • the value is compared with the value of the offset, if the minimum value is greater than the value of the offset, the updated checkout node information is equal to the difference between the minimum value and the value of the offset; If the value is less than or equal to the value of the offset, the updated check node information is equal to 0;
  • the step of adaptively adjusting the offset according to the number of decoding iterations may include:
  • the offset used is 0; when the number of decoding iterations reaches the threshold of the number of iterations, if the check constraint relationship corresponding to the check node is in the previous iterative decoding process If the middle is satisfied, the offset used in the iterative decoding process is 0; if the check constraint relationship corresponding to the check node is not satisfied in the previous iterative decoding process, then in this iteration In the decoding process, the offset used is 1.
  • the step of performing an update process of hierarchically accumulating the information outside the variable node according to the updated information of the check node may include:
  • the information outside the check node is divided into groups of two values, and the N bits are non-uniformly quantized.
  • the variable node prior probability information and the information of each group of check nodes are subjected to hierarchical accumulation processing, and each time the level of accumulation is performed, the bit width of the sum value obtained by the addition is increased by one bit;
  • the updated information outside the variable node is 2 ⁇ -1, and if the update value obtained by accumulating the variable node is less than ⁇ -l, the updated variable is updated.
  • the out-of-node information is ⁇ -l. If the update value obtained by accumulating and updating the variable node is within the interval [ ⁇ -l ⁇ -l], the updated value obtained by accumulating the variable node is used as the value of the updated information outside the variable node; Store updated information outside the variable node.
  • the step of determining whether the decoding ends or not may include:
  • the decoding succeeds and the decoding process ends; if the product does not satisfy the even parity condition, it is determined whether the number of times of the stacked decoding generation exceeds The preset maximum iteration number iter max , if the number of decoding iterations does not exceed the preset maximum iteration number iter max , then the check node out-of-node information, the update processing of the variable node information is continuously performed, and it is judged whether the decoding ends; The number of code iterations exceeds the preset maximum number of iterations iter max , ending the decoding process and declaring the decoding failure.
  • the iterative decoding method and system provided by the embodiments of the present invention perform non-uniform quantization processing on the received channel information; perform initialization processing on the prior probability information of the variable node, and use the quantized channel information as the prior probability information of the variable node According to the prior probability information of the variable node, and using the adaptive offset to perform update processing of the check node out-of-node information; and updating and updating the information outside the variable node according to the updated information of the check node; Determining whether the decoding ends; therefore, using the non-uniform quantized channel information, which utilizes 4 or 5 or 6 bits, in the iterative decoding process can save hardware logic resources and storage resources; in addition, the update process of the information outside the variable node can Using the hierarchical accumulation to obtain high-precision updated information outside the variable node, and using the adaptive offset to check the update of the information outside the node, the update processing of the variable node and the check node can reduce the translation caused by quantization. Code performance loss; In summary,
  • FIG. 1 is a schematic structural diagram of an iterative decoding system implemented in an embodiment of the present invention
  • FIG. 2 is a schematic flowchart of implementing an iterative decoding method according to an embodiment of the present invention
  • FIG. 3 is a schematic diagram showing an update of hierarchical accumulation of information outside a 16-input variable node in an embodiment of the present invention
  • FIG. 4 is a schematic diagram of a 6-input comparison unit in a check node update unit according to an embodiment of the present invention
  • FIG. 5 is a schematic diagram of a 7-input comparison unit in a check node update unit according to an embodiment of the present invention
  • FIG. 7 is a schematic diagram of bit error rate performance of an iterative decoding method in accordance with an embodiment of the present invention.
  • the received channel information is subjected to non-uniform hook quantization processing; the prior probability information of the variable node is initialized, and the quantized channel information is used as the prior probability information of the variable node;
  • the prior probability information, and the update processing of the information outside the check node is performed by using the adaptive offset; and the update processing of the information outside the variable node is performed according to the updated information of the check node; End.
  • FIG. 1 is a schematic structural diagram of an iterative decoding system according to an embodiment of the present invention. As shown in FIG. 1, the system includes: a quantization unit 10, and a decoding unit 20;
  • the quantization unit 10 is configured to perform non-homogeneous quantization processing on the received channel information.
  • the decoding unit 20 is configured to perform initialization processing on the prior probability information of the variable node, and use the quantized channel information as the first risk of the variable node. ⁇ rate information; based on the first risk information of the variable node, and using the adaptive offset to verify the update of the information outside the node; The information outside the node is checked, and the update processing of the information outside the variable node is hierarchically accumulated; and the decoding is judged to be ended.
  • the system further includes: a channel information storage unit 30 configured to store the quantized channel information.
  • the decoding unit 20 includes:
  • the initializing unit 21 is configured to perform initialization processing on the prior probability information of the variable node, and use the quantized channel information as the first risk information of the variable node;
  • the check node update unit 22 is configured to perform update processing of the check node out-of-node information according to the prior probability information of the variable node, and using the adaptive offset;
  • variable node updating unit 23 is configured to perform an update process of hierarchically accumulating the information outside the variable node according to the updated information outside the check node;
  • the determining unit 24 is arranged to determine whether the decoding is ended.
  • the decoding unit 20 further includes:
  • the first storage unit 25 is configured to store the updated check node out-of-node information
  • the second storage unit 26 is configured to store the updated out-of-variable node information.
  • FIG. 2 is a schematic flowchart of an iterative decoding method according to an embodiment of the present invention. As shown in FIG. 2, the method includes the following steps: Step 201: Receive The obtained channel information is subjected to non-uniform hook quantization processing, and the quantized channel information is stored;
  • the quantization processing unit performs N-bit non-uniform quantization processing on the channel information received by the self-input decoding unit according to the channel information N-bit non-uniform hook quantization scheme obtained by the simulation optimization, where N is 4 or 5 or 6: storing the quantized channel information in the channel information storage unit; performing N-bit non-uniform hook quantization processing on the received channel information for inputting the decoding unit, and quantizing the received channel information into: Wherein, it is a magnitude value, i is a quantized value, and ⁇ is the received channel information, which is quantized channel information; here, the quantized channel information has the same sign bit as the received channel information y n .
  • Step 202 Initialize the prior probability information of the variable node, and use the quantized channel information as the first risk information of the variable node;
  • the initialization unit in the decoding unit initializes the prior probability information ⁇ l n of the variable node, and uses the quantized channel information as the first risk information of the variable node, namely:
  • Step 203 Calculate the first risk information of the variable node, and use the adaptive offset to perform update processing of the information outside the check node;
  • the check node update unit in the decoding unit updates the check node information to:
  • N(m) is the set of variable nodes of m; min denotes the minimum value, max denotes the maximum value; sign is the symbol function:
  • h m represents the mth row of the check matrix H, and each row of the H matrix is a constraint relationship that the decoded codeword needs to satisfy;
  • the offset is adaptively adjusted according to the number of decoding iterations, specifically: when the number of decoding iterations is less than the iterative number threshold iter ad , the offset used is 0; when the number of decoding iterations reaches the number of iterations When the threshold is iter ad , if the check constraint relationship corresponding to the check node is satisfied in the previous iterative decoding process, in the present iterative decoding process, the offset of the check node information is used. The shift amount is 0. If the check constraint relationship corresponding to the check node is not satisfied in the previous iterative decoding process, the offset used when updating the check node out information is updated during the iterative decoding process.
  • the quantity is 1; wherein, the check node outer information Z m ⁇ sign bit can be obtained by XORing the sign bit of the variable node outer information ⁇ m ( ne N ⁇ m) ⁇ n ) obtained by the previous iteration decoding;
  • the comparison unit in the check node update unit first divides the information of the variable node associated with the check node into two groups, and compares the information of the outer information of the variable node by hierarchically.
  • the value f obtains the minimum value of the modulus f of the information outside the variable node, and compares the obtained minimum value with the value of the offset. If the minimum value is greater than the value of the offset, the updated school
  • the outer information of the node is equal to the difference between the minimum value and the value of the offset; if the minimum value is less than or equal to the value of the offset, the updated check node outer information is equal to 0;
  • the first storage unit in the decoding unit stores the updated check node out-of-node information; in practical applications, the decoding unit may be a decoder, such as an LDPC decoder.
  • Step 204 Perform an update process of hierarchically accumulating information outside the variable node according to the updated information of the check node, wherein the data bandwidth is expanded step by step;
  • variable node update unit in the decoding unit updates the variable node information and the posterior probability information (3 ⁇ 4 to:
  • the checkout node information Z ( me M(n) ⁇ m ) is divided into groups of two values, and N bits are divided.
  • the non-homogeneously quantized variable node prior probability information / chest and each set of check node information is subjected to hierarchical accumulation processing, and each time one level of accumulation is performed, the bit width of the sum value obtained by the addition is increased by one bit; If the updated value obtained by accumulating the update of the variable node is greater than S ⁇ -l, the updated information outside the variable node is assigned S ⁇ -l. If the update value obtained by accumulating the variable node is less than ⁇ -l, the updated value will be updated.
  • the value of the parameter outside the variable node is N- 1 - 1, and if the variable node is accumulated and updated, the updated value is in the interval.
  • the second storage unit in the decoding unit stores the updated out-of-variable node information.
  • the input 4 bits of channel information and the check node outgoing information are grouped into two groups for hierarchical accumulation, and the data bit width of the operation data in the hierarchical accumulation process is expanded step by step with the cumulative number of stages.
  • the data bit width of the accumulated result of each level is extended by one bit on the basis of the data bit width of the previous level; if the node degree of the variable node is d v , the data to be accumulated in the process of updating the information outside the variable node includes one channel Information and check information outside the node, there are +1 accumulated data, and the accumulated sum value of d v + l data needs at least log 2 « + l) level accumulation operation, so the updated value obtained by the cumulative update of the variable node
  • the data bit width is extended by "log 2 ( + l bit) based on the input data bit width, so that the data calculation accuracy in the process of updating the information outside the variable node is higher than 4 bits, and the update of the information outside the variable node is completed.
  • the 16 pieces of information data of the accumulating unit of the input variable node update unit are all 4-bit data, and 16 are 4-bit data is divided into 8 groups in groups of two, and the packets are accumulated;
  • the accumulated result of the output of the eight accumulation units is increased by one bit of data width based on the input data, which is 5 bits of data;
  • the second stage accumulation, the third level accumulation and the fourth level accumulation manner The result is the same as the accumulation of the first stage;
  • the result of the output of the four accumulation units after the second stage is 6-bit data, and the result of the output of the two accumulation units after the third stage is 7-bit data, and the output result of the accumulation unit after the fourth-stage accumulation
  • the 8-bit accumulated update result is converted into 4-bit updated variable-node information by searching for the optimized design quantization precision conversion table.
  • Step 205 determining whether the decoding ends
  • the determining unit in the decoding unit determines that the decoding is successful and ends the decoding process; if the product does not satisfy the even state After the condition is met, the determining unit determines whether the number of times of the superposition of the decoding exceeds the preset maximum number of iterations iter max , and if the number of decoding iterations does not exceed the preset maximum number of iterations iter max , repeating steps 203 to 205; The number of code iterations exceeds the preset maximum number of iterations iter max , then the decoding process is terminated and the decoding fails.
  • This embodiment takes an LDPC with a code length of 576 bits and a code rate of 1/2 in the IEEE 802.16e standard as an example, and is described with reference to FIG. 4 to FIG. 7.
  • the channel information obtained by the simulation optimization is quantized by the 4-bit non-uniform hook quantization scheme, and the quantized 4-bit channel information is input into the decoding unit as the prior probability information of the variable node. Since the code length is 576 bits, the decoding unit inputs 576 channel information having a bit width of 4 bits each time, and stores 576 channel information storage units having a bit width of 4 bits, respectively. Simultaneously, The 576 channel information is used to initialize the out-of-variable node information corresponding to the data bit width of 4 bits.
  • the sign bit of the information of the variable node related to the corresponding check node is XORed to obtain the updated symbol of the information outside the check node; the module value update process of verifying the information outside the node
  • the modulus values of the information outside the variable node associated with the check node are grouped into two groups, and the minimum value and the second small value are obtained by hierarchically comparing the modulus values of the information outside the variable node, and then the minimum value is obtained. And the value of the second smallest value and the quantized offset.
  • the quantized offset is zero.
  • the quantized offset is taken as 0; if not, the quantized offset is taken as 1.
  • the magnitude relationship between the difference between the minimum and the minor values and the quantized offset is determined, and the combination of the sign bits is obtained by the check node being passed to the relevant variable node with a data bit width of 4 bits outside the check node. information.
  • the updated check node out information is respectively stored in the corresponding first storage unit having a data bit width of 4 bits.
  • the parity node update unit uses a data bit width of 3 bits.
  • the input and 7-input comparison units implement hierarchical comparison of the modulus values of the information outside the variable node and obtain the minimum and minor values therein.
  • the data bit width in the check node update unit is The 3-bit 6-input comparison unit block diagram is shown in Figure 4, and the 7-input comparison unit block diagram is shown in Figure 5.
  • the comparison unit can obtain the minimum value and the sub-small value of the information outside the variable node and the label of the variable node corresponding to the minimum value of the information outside the variable node.
  • variable node update unit the input 4-bit channel information and the check node out-of-node information According to each of the two values, the hierarchical accumulation process is performed; since the code length of the IEEE 802.16e standard is 576 bits, the LDPC with the code rate of 1/2 includes the connection points of node degrees 2, 3, and 6, Therefore, a hierarchical accumulation unit of 2 inputs, 3 inputs, 4 inputs, 6 inputs, and 7 inputs is required in the variable node update unit in the decoding unit. Among them, the 7-input hierarchical accumulation unit block diagram is shown in Figure 6. The input data of the 7 input accumulation unit is 4-bit channel information and check node out-of-node information.
  • the first six input data are respectively accumulated in three groups, and the accumulated result is 5-bit data. Extend the data bit width of the 7th data by one.
  • the second stage accumulation and the third stage accumulation repeat the process of the first stage accumulation, and the data width of the accumulation sum result is extended to 7 bits.
  • the 7-bit data obtained by the accumulated update is converted into 4-bit data as the updated variable node information and the decoded posterior probability information.
  • a method of converting 7-bit data into 4-bit data using direct truncation clipping is used.
  • variable node updating unit updates each of the decoded posterior probability information or the variable node information, and stores it in the corresponding decoded posterior probability information storage unit with a data bit width of 4 bits. (such as a third storage unit in the decoding unit) or a variable node out-of-memory information storage unit (such as a second storage unit in the decoding unit).
  • the decoding unit extracts the symbol bits of the 576 decoded posterior probability information as the decoded codeword vector obtained by the iterative decoding, and carries it into the check equation for verification. . If the decoded codeword vector satisfies the check equation, the iterative decoding process is ended, and the decoded codeword is output. If the decoded codeword vector satisfies the check equation, if the number of iterations does not exceed the preset maximum iteration number iter max , the iterative decoding process is repeated; if the number of iterations exceeds the preset maximum iteration number iter max , the decoding process is ended. And output the decoded codeword.
  • the bit error rate performance of the LDPC iterative decoding method with low quantization precision designed in the embodiment is as shown in FIG. 7.
  • the dotted line in the figure indicates the bit error rate of the low quantization precision LDPC iterative decoding method proposed in the embodiment of the present invention.
  • the solid line represents the bit error rate of the floating-point BP decoding algorithm, the bit error rate reaches 110-8, the embodiment of the present invention, the bit error iterative LDPC decoding method according to the quantization accuracy is low proposed Compared with the floating-point BP decoding algorithm, the performance difference is about 0.2 dB. This indicates that the low quantization precision LDPC iterative decoding method proposed by the embodiment of the present invention reduces the hardware implementation complexity and storage resource consumption of the decoding unit, and causes a small loss of decoding performance.
  • each module/unit in the foregoing embodiment may be implemented in the form of hardware, or may use software functions.
  • the form of the module is implemented. The invention is not limited to any specific form of combination of hardware and software.
  • the technical solution of the embodiments of the present invention can reduce the storage resource and hardware implementation complexity of the LDPC decoder, and can ensure the operation precision in the decoding process and reduce the decoding performance loss caused by the quantization.

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Probability & Statistics with Applications (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Nonlinear Science (AREA)
  • Error Detection And Correction (AREA)

Abstract

一种迭代译码方法及系统,该方法包括:对接收的信道信息进行非均匀量化处理;对变量节点的先验概率信息进行初始化处理,将量化后的信道信息作为变量节点的先验概率信息;根据变量节点的先验概率信息,并利用自适应的偏移量进行校验节点外信息的更新处理;根据更新后的校验节点外信息,对变量节点外信息进行分级累加的更新处理;判断译码是否结束。该方法能够降低LDPC译码器的存储资源和硬件实现复杂度。

Description

一种迭代译码方法及系统
技术领域
本发明涉及通信领域的数字信号传输与存储系统, 尤其涉及一种迭代译 码方法及系统。
背景技术
在现代数字信号传输与存储系统中 , 由于传输信道噪声或存储媒介的物 理损伤等, 常会造成数字信号的传输或存储的错误, 因此, 为保证数字信号 传输或存储的可靠性, 差错控制编码技术是一项标准技术。
低密度奇偶校验码 ( LDPC, Low Density Parity Check Code )是一种差错 控制编码技术,最早由 Gallager于 1962年提出,但由于受到技术水平的限制, 在很长一段时间里并没有得到重视和推广 ,直到 1996年 D. MacKay和 R. Neal 重新发现 LDPC并证明其具有接近香农极限的良好性能, LDPC才受到越来 越多的关注。
LDPC是一类定义在稀疏矩阵 H上的线性分组码。 H即为 LDPC的校验 矩阵, 对于任何一个合法码字 V, 都满足校验方程 Η·νΤ=0, 其中, LDPC的 校验矩阵 H可用二部图表示。 基于校验矩阵的二部图的软判决译码算法是目 前较常用的 LDPC 的译码方法。 软判决译码算法可以通过基于置信度传播 ( BP, Belief Propagation )译码算法的迭代译码来实现, 并且可以获得良好的 误比特率性能。 虽然 BP译码算法具有较好的译码性能,但算法中涉及到许多 非线性运算, 不适合釆用硬件实现, 因此产生了一些简化的 BP译码算法, 如 最小和( MS, Min-Sum )译码算法,归一化最小和( NMS, Normalized Min-Sum ) 译码算法和偏移最小和(OMS, Offset Min-Sun )译码算法。 最小和译码算法 中主要釆用的运算形式为比较运算和求和运算, 算法复杂度很低, 但其性能 与 BP译码算法相比有较大的损失;归一化最小和译码算法和偏移最小和译码 算法均是在最小和译码算法的基础上提出的改进译码算法。 归一化最小和译 码算法引入了一个小于 1 的校正因子 α, 偏移最小和译码算法引入了一个偏 移量 β。 两种算法在性能上没有明显的差异, 但归一化最小和译码算法在最 小和译码算法的基础上增加了一个乘法运算, 偏移最小和译码算法只增加了 一个比较算法和一个加法运算。 在硬件实现方面, 偏移最小和算法的复杂度 更低。 因此, 为了能够有效降低有限精度 LDPC译码器的硬件逻辑资源, 多 数情况下在译码器中釆用偏移最小和译码算法。
为介绍偏移最小和译码算法, 先给出 LDPC的二部图描述。 分别对应于 校验矩阵 H中的 N列和 M行 LDPC对应的二部图中包含两类节点: N个校 验节点 c和 M个变量节点 V, 二部图中的边对应于校验矩阵 H中的 "1 " , 若校验矩阵 H中的元素 h(m,n)为 1 ,则二部图中的校验节点 cm和变量节点 vn 之间存在一条边。 定义 M(n)为二部图中与变量节点 vn相连接的校验节点的 集合, N(m)为参与校验方程 m的变量节点的集合。 M(n)\m表示集合 M(n)中 除去元素 cm, 同样 N(m)\n表示集合 N(m)中除去元素 vn。 定义 In表示变量节 点 vn的先验概率信息,译码过程中校验节点 cm传递给变量节点 vn的外信息 为 Lmn, 变量节点 vn传递给校验节点 cm的外信息为 znm, 变量节点 vn的后 验概率信息为 LQn。 设从信道接收到的码字向量为 Y=[yl,y2, ... ,yn] , 其中 yi 为接收到的关于第 i个比特的信道信息。 根据以上给出相关概念, 偏移最小 和译码算法的译码过程可分为以下 4个步骤:
第 1步: 初始化, 变量节点的先险^^率信息 ln被初始化为:
/ = V
n ^ n
对于每个满足 {(m,n)|H(m,n)=l }的 (m,n), 变量节点外信息 Znm=ln
第 2步: 校验节点更新, 校验节点外信息 Ι^η更新为:
Figure imgf000004_0001
其中, β为一个较小的正数, 是偏移最小和译码算法的偏移量修正因子, 偏移量修正因子一般根据密度进化方法或仿真方法确定其最优值。
第 3步: 变量节点更新, 变量节点的外信息 Znm和后险^^率信息 LQn更新 为:
Figure imgf000004_0002
LQn = ln + ∑ L
meM(n)
第 4步: 判决, 当 LQn≥0时, „=0, LQn<0,则 „=1 , 这样就生成了译码 码字向量 。 如果校验方程 H = 0 0(12;)成立, 则译码成功并结束。 如果校 验方程不成立, 若迭代次数未超过预设的最大迭代次数, 则重复 2至 4步的 迭代译码过程; 若迭代次数超过预设的最大迭代次数, 则结束译码过程并声 明译码失败。
由 LDPC的迭代译码算法可以看出, LDPC的译码过程可以并行执行, 而现场可编程门阵列 ( FPGA, Field - Programmable Gate Array )和超大规模 集成电路(VLSI, Very Large Scale Integration )具有支持并行操作的优势, 因此, 在 LDPC译码器适合于釆用 FPGA或 VLSI来进行硬件实现。
然而, 釆用硬件实现 LDPC译码器需要考虑硬件资源问题, 特别是片内 存储器(RAM, Random Access Memory )资源和還辑资源。 在存储器资源方 面, 由于 LDPC釆用软信息迭代译码方法, 所以在译码过程中需要使用大量 的存储资源存储外信息。 然而, FPGA 内部的 RAM资源非常有限。 而对于 ASIC来说, 大量的片内 RAM会占用过多的芯片面积, 进而挤占片内逻辑资 源的空间, 限制芯片的逻辑复杂度, 并且会增加芯片的功耗。 在逻辑资源方 面, 为了能够提高译码器的吞吐率, 需要提高 LDPC译码器的译码并行度, 进而增加实现译码器所需要的硬件逻辑资源。 因此, 在译码器实现过程中, 需尽量减少实现译码算法所需的硬件逻辑资源, 以提高译码器的并行度。
为了能够节省存储资源和硬件逻辑资源, 在译码器的硬件实现过程中需 要釆用有限精度的 LDPC译码算法, 即对输入译码器的信道信息需要进行量 化, 在译码过程中釆用限精度的数据运算。
对输入译码器的信道信息进行量化是用预先规定的有限个整数值表示输 入译码器的信道信息值的过程。 预先规定的有限个整数值称为量化值。 量化 值用 q比特有符号的二进制码表示, 二进制码的数据位宽称为量化值的量化 精度。 有符号的二进制码的最高位为符号位, 若量化值为负值, 则最高位取
"1" , 若量化值不为负值, 最高位取 "0" 。 二进制码的其余 q-1 比特表示 量化值的模值。 q 比特有符号的二进制码只能与 M= 2 -l个量化值, 即 [-T,...,-1,0,1,...,T]相对应, 其中 1= 2^ -1。 这就需要把输入译码器的信道信息 值的取值区间划分为与有限个量化值对应的 Μ个量化区间。设 为与量化值 i ( i e [-Τ,...,-1,0,1,...,Τ] )相对应的量化区间的终点值, 称为量阶值。 定义 AV^qrqw为量化间隔。 量化间隔相等的量化方法称为均勾量化, 量化间隔不 相等的量化方法称为非均勾量化。
对信道信息进行量化会造成译码算法的译码性能损失。 量化方案中的量 化精度和量阶值均会影响译码算法的译码性能损失情况。 因此, 对于设计有 限精度 LDPC码译码算法, 信道信息量化方案中量化精度和量阶值的选取是 一项重要工作。
在量化精度的选取方面, 主要工作是平衡量化引起的译码性能损失与译 码器的硬件资源。 若量化方法设计的量化精度较高, 则译码运算过程中的数 据具有较宽的数据位宽, 能够提高译码过程中的数据运算精度, 进而减小量 化造成的译码性能损失。 但较高的量化精度要求译码器的数据总线和外信息 存储单元都具有较宽的数据位宽。 这样译码器的硬件实现需要消耗大量的硬 件逻辑资源和存储资源。 若量化方法设计的量化精度较低, 则能够达到节省 实现译码器的硬件逻辑资源和存储资源的目的, 但译码过程中的数据运算精 度也会随之降低, 从而可能导致译码算法的译码性能损失较大。
在量阶值的设计方面, 主要工作是通过选取合适的量阶值和量化间隔, 尽量减小量化引起的译码性能损失。 当量化精度较高时, 釆用均勾量化方法 即可。 但是, 当量化精度较低时, 釆用均匀量化方法很难同时保证量化数据 的覆盖范围和量化数据的精度, 量化引起的译码性能损失较大, 因此需要釆 用非均匀量化方法。 非均匀量化方法中, 可减小较小的量阶值之间的量化间 隔, 同时扩大较大的量阶值之间的量化间隔。 这样可同时保证量化数据的精 度和覆盖范围。 在非均勾量化方法中, 可通过优化量阶值的设计, 进一步减 小量化引起的译码性能损失。
译码过程中的数据运算为有限精度数据运算, 对有限精度数据进行加法 运算时由于加法进位可能导致数据位宽扩展, 而译码器的外信息存储单元的 数据位宽无法扩展, 因此需要对译码过程中产生数据位宽扩展的数据进行数 据位宽限幅。 对产生数据位宽扩展的数据釆用不同的数据位宽限幅方法, 也 会对译码算法的译码性能产生较大的影响。 针对不同的 LDPC码, 需要通过 仿真优化译码过程中位宽扩展数据的数据位宽限幅方法, 以降低由量化造成 的译码算法的译码性能损失。
发明内容
有鉴于此, 本发明实施例的目的在于提供一种迭代译码方法及系统, 以 降低 LDPC译码器的存储资源和硬件实现复杂度。
为达到上述目的, 本发明实施例的迭代译码系统, 包括:
量化单元, 其设置为对接收的信道信息进行非均勾量化处理; 以及 译码单元, 其设置为对变量节点的先验概率信息进行初始化处理, 将量 化后的信道信息作为变量节点的先险 ¾^率信息; 才艮据变量节点的先险 ¾^率信 息, 并利用自适应的偏移量进行校验节点外信息的更新处理; 才艮据更新后的 校验节点外信息, 对变量节点外信息进行分级累加的更新处理; 以及判断译 码是否结束。
上述系统中, 该系统还可以包括: 信道信息存储单元, 其设置为存储量 化后的信道信息。
上述系统中, 所述译码单元可以包括:
初始化单元, 其设置为对变量节点的先验概率信息进行初始化处理, 将 量化后的信道信息作为变量节点的先险^^率信息;
校验节点更新单元, 其设置为根据变量节点的先验概率信息, 并利用自 适应的偏移量进行校验节点外信息的更新处理;
变量节点更新单元, 其设置为根据更新后的校验节点外信息, 对变量节 点外信息进行分级累加的更新处理; 以及
判断单元, 其设置为判断译码是否结束。
上述系统中, 该系统还可以包括:
第一存储单元 , 其设置为存储更新后的校验节点外信息;
第二存储单元, 其设置为存储更新后的变量节点外信息。
本发明实施例还提供一种迭代译码方法, 包括:
对接收的信道信息进行非均勾量化处理;
对变量节点的先验概率信息进行初始化处理, 将量化后的信道信息作为 变量节点的先险^^率信息;
根据变量节点的先验概率信息, 并利用自适应的偏移量进行校验节点外 信息的更新处理;
根据更新后的校验节点外信息, 对变量节点外信息进行分级累加的更新 处理;
判断译码是否结束。
上述方法中, 所述对输入译码单元的信道信息进行非均勾量化处理的步 骤可包括:
才艮据仿真优化得到的信道信息 N比特非均勾量化方案, 对接收到的用于 输入译码单元的信道信息进行 N比特非均匀量化处理,所述 N为 4或 5或 6; 存储量化后的信道信息。
上述方法中, 所述对接收到的用于输入译码单元的信道信息进行 N比特 非均勾量化处理的步骤可包括: 对接收到的用于输入译码单元的信道信息进行 N比特非均匀量化处理, 将接收到的信道信息量化为:
Figure imgf000009_0001
其中, 为量阶值, i为量化值, ^为接收到的信道信息, 为量化后的 信道信息。
上述方法中, 所述根据变量节点的先验概率信息, 并利用自适应的偏移 量进行校验节点外信息的更新处理的步骤可包括:
偏移量根据译码迭代次数进行自适应调整;
将与校验节点相关联的变量节点外信息按照每两个分为一组, 通过分级 比较变量节点外信息的模值, 得到变量节点外信息的模值中的最小值, 再将 得到的最小值与偏移量的值进行比较, 如果所述最小值大于偏移量的值, 则 更新后的校验节点外信息等于所述最小值与偏移量的值的差值; 如果所述最 小值小于等于偏移量的值, 则更新后的校验节点外信息等于 0;
存储更新后的校验节点外信息。
上述方法中, 所述偏移量根据译码迭代次数进行自适应调整的步骤可包 括:
译码迭代次数小于迭代次数门限值时, 釆用的偏移量为 0; 译码迭代次 数达到迭代次数门限值时, 如果校验节点对应的校验约束关系在前一次迭代 译码过程中得到满足, 则在本次迭代译码过程中, 釆用的偏移量为 0; 如果 校验节点对应的校验约束关系在前一次迭代译码过程中未得到满足, 则在本 次迭代译码过程中, 釆用的偏移量为 1。
上述方法中, 所述根据更新后的校验节点外信息, 对变量节点外信息进 行分级累加的更新处理的步骤可包括:
将校验节点外信息按照每两个值一组进行划分, 将 N比特非均匀量化的 变量节点先验概率信息和每组校验节点外信息进行分级累加处理, 每进行一 级的累加, 就将累加后得到的和值的位宽增加一个比特;
最后,如果变量节点累加更新得到的更新值大于 S^-l , 则更新后的变量 节点外信息为 2^-1 , 如果变量节点累加更新得到的更新值小于 ^-l , 则更 新后的变量节点外信息为 ^-l,如果变量节点累加更新得到的更新值在区间 [ ^-l ^-l]内, 则将变量节点累加更新得到的更新值作为更新后的变量节 点外信息的值; 存储更新后的变量节点外信息。
上述方法中, 所述判断译码是否结束的步骤可包括:
如果校验矩阵 H与译码码字向量 的乘积满足偶校验条件, 则译码成功 并结束译码流程; 如果所述乘积不满足偶校验条件, 则判断迭译码代迭次数 是否超过预设的最大迭代次数 itermax,如果译码迭代次数未超过预设的最大迭 代次数 itermax,则继续进行校验节点外信息、变量节点外信息的更新处理并判 断译码是否结束;如果译码迭代次数超过预设的最大迭代次数 itermax,结束译 码过程并声明译码失败。
本发明实施例提供的迭代译码方法及系统, 对接收的信道信息进行非均 匀量化处理; 对变量节点的先验概率信息进行初始化处理, 将量化后的信道 信息作为变量节点的先验概率信息; 根据变量节点的先验概率信息, 并利用 自适应的偏移量进行校验节点外信息的更新处理; 根据更新后的校验节点外 信息, 对变量节点外信息进行分级累加的更新处理; 判断译码是否结束; 因 此,迭代译码过程中利用均利用 4或 5或 6比特的非均匀量化后的信道信息, 能够节省硬件逻辑资源和存储资源; 此外, 变量节点外信息的更新过程能够 利用分级累加得到高精度的更新的变量节点外信息, 同时釆用自适应的偏移 量进行校验节点外信息的更新, 因此变量节点和校验节点的更新处理方式能 够减小量化引起的译码性能损失; 综上, 本发明的技术方案能够降低 LDPC 译码器的存储资源和硬件实现复杂度,同时能够保证译码过程中的运算精度, 降低由于量化引起的译码性能损失。
附图概述
图 1是本发明实施例实现迭代译码系统的结构示意图;
图 2是本发明实施例实现迭代译码方法的流程示意图;
图 3是本发明实施例中 16输入变量节点外信息的分级累加的更新示意 图;
图 4是本发明实施例中校验节点更新单元中 6输入比较单元的示意图; 图 5是本发明实施例中校验节点更新单元中 7输入比较单元的示意图; 图 6是本发明实施例中变量节点更新单元中 7输入比较单元的示意图; 图 7是本发明实施例实现迭代译码方法的误比特率性能的示意图。
本发明的较佳实施方式
下文中将结合附图对本发明的实施例进行详细说明。 需要说明的是, 在 不冲突的情况下, 本申请中的实施例及实施例中的特征可以相互任意组合。
本发明实施例的方案中, 对接收的信道信息进行非均勾量化处理; 对变 量节点的先验概率信息进行初始化处理, 将量化后的信道信息作为变量节点 的先验概率信息; 根据变量节点的先验概率信息, 并利用自适应的偏移量进 行校验节点外信息的更新处理; 根据更新后的校验节点外信息, 对变量节点 外信息进行分级累加的更新处理; 判断译码是否结束。
图 1是本发明实施例实现迭代译码系统的结构示意图, 如图 1所示, 该 系统包括: 量化单元 10、 译码单元 20; 其中,
量化单元 10, 设置为对接收的信道信息进行非均勾量化处理; 译码单元 20, 设置为对变量节点的先验概率信息进行初始化处理, 将量 化后的信道信息作为变量节点的先险 ¾^率信息; 才艮据变量节点的先险 ¾^率信 息, 并利用自适应的偏移量进行校验节点外信息的更新处理; 才艮据更新后的 校验节点外信息, 对变量节点外信息进行分级累加的更新处理; 以及判断译 码是否结束。
该系统还包括: 信道信息存储单元 30, 设置为存储量化后的信道信息。 所述译码单元 20包括:
初始化单元 21 , 设置为对变量节点的先验概率信息进行初始化处理, 将 量化后的信道信息作为变量节点的先险^^率信息;
校验节点更新单元 22, 设置为根据变量节点的先验概率信息, 并利用自 适应的偏移量进行校验节点外信息的更新处理;
变量节点更新单元 23 , 设置为根据更新后的校验节点外信息, 对变量节 点外信息进行分级累加的更新处理;
判断单元 24, 设置为判断译码是否结束。
译码单元 20还包括:
第一存储单元 25, 设置为存储更新后的校验节点外信息;
第二存储单元 26 , 设置为存储更新后的变量节点外信息。
基于上述系统, 本发明实施例还提供一种迭代译码方法, 图 2是本发明 实施例实现迭代译码方法的流程示意图, 如图 2所示, 该方法包括以下步骤: 步骤 201 , 对接收到的信道信息进行非均勾量化处理, 并存储量化后的 信道信息;
具体的, 量化处理单元根据仿真优化得到的信道信息 N比特非均勾量化 方案, 对自身接收到的用于输入译码单元的信道信息进行 N比特非均匀量化 处理, 其中 N为 4或 5或 6; 将量化后的信道信息存储到信道信息存储单元; 对接收到的用于输入译码单元的信道信息进行 N比特非均勾量化处理, 将接 收到的信道信息量化为:
Figure imgf000012_0001
其中, 为量阶值, i为量化值, ^为接收到的信道信息, 为量化后的 信道信息; 这里, 量化后的信道信息 的符号位与接收的信道信息 yn的符号 位相同。
步骤 202, 对变量节点的先验概率信息进行初始化处理, 将量化后的信道 信息作为变量节点的先险^^率信息;
具体的, 译码单元中的初始化单元对变量节点的先验概率信息 ~ln进行初 始化处理, 将量化后的信道信息作为变量节点的先险 ¾^率信息, 即:
K ~ =yn
对于每个满足 {(m,n)|H(m,n)=l}的 (m,n), 前一次迭代译码中得到的变量节 点外信息 其中, Π1表示校验矩阵 Η的行序号, η表示校验矩阵 Η的 列序号, {(m,n)|H(m,n)=l}表示 Η矩阵中为 1的元素的坐标集合。
步骤 203, 才艮据变量节点的先险 ¾^率信息, 并利用自适应的偏移量进行校 验节点外信息的更新处理;
具体的, 当译码迭代次数小于迭代次数门限值 iterad时, 译码单元中的校 验节点更新单元将校验节点外信息 £皿更新为:
Figure imgf000013_0001
其中, n表示集合 N(m)中除去元素 vn的其他元素, N(m)为 m的变量节点 的集合; min表示取最小值, max表示取最大值; sign为符号函数:
Figure imgf000013_0002
当译码迭代次数达到迭代次数门限值 iterad时, 译码单元中的校验节点更 新单元将校验节点外信息 £皿更新为: sign(z n, -0 ,0 , /zm- = 0(mod2)
.1 ,0 , hm-X = \(mod2)
Figure imgf000013_0003
其中, hm表示校验矩阵 H的第 m行, H矩阵的每一行都是译码码字 需 要满足的一个约束关系;
这里, 偏移量根据译码迭代次数进行自适应调整, 具体为: 当译码迭代 次数小于迭代次数门限值 iterad时, 釆用的偏移量为 0; 当译码迭代次数达到 迭代次数门限值 iterad时, 如果校验节点对应的校验约束关系在前一次迭代译 码过程中得到满足, 则在本次迭代译码过程中, 更新该校验节点外信息时釆 用的偏移量为 0; 如果校验节点对应的校验约束关系在前一次迭代译码过程 中未得到满足, 则在本次迭代译码过程中, 更新该校验节点外信息时釆用的 偏移量为 1 ; 其中, 校验节点外信息 Zm 々符号位可以通过对前一次迭代译码 得到的变量节点外信息^ m ( n e N{m) \ n ) 的符号位进行异或运算得到;
校验节点外信息的更新过程中, 校验节点更新单元中的比较单元首先将 与校验节点相关联的变量节点外信息按照每两个分为一组, 通过分级比较变 量节点外信息的模值 f , , 得到变量节点外信息的模值 f , 中的最小值, 再将 得到的最小值与偏移量的值进行比较, 如果该最小值大于偏移量的值, 则更 新后的校验节点外信息等于该最小值与偏移量的值的差值; 如果该最小值小 于等于偏移量的值, , 更新后的校验节点外信息等于 0;
最后, 译码单元中的第一存储单元存储更新后的校验节点外信息; 实际 应用中, 译码单元可以为译码器, 例如 LDPC译码器。
步骤 204, 根据更新后的校验节点外信息, 对变量节点外信息进行分级累 加的更新处理, 其中数据带宽逐级扩展;
具体的,译码单元中的变量节点更新单元将变量节点外信息 ^皿和后验概 率信息(¾更新为:
znm = ln + ∑ L ,
Qn = + ∑ Lmn 其中, m'表示集合 M(n)中除去元素 cm的其他元素;
在变量节点外信息 5„m和后验概率信息 (¾的更新过程中, 将校验节点外 信息 Z , ( m e M(n) \m )按照每两个值一组进行划分, 将 N比特非均勾量化 的变量节点先验概率信息 /„和每组校验节点外信息进行分级累加处理, 每进 行一级的累加, 就将累加后得到的和值的位宽增加一个比特; 最后, 如果变 量节点累加更新得到的更新值大于 S^-l ,则将更新后的变量节点外信息赋值 为 S^-l , 如果变量节点累加更新得到的更新值小于 ^-l , 则将更新后的变 量节点外信息赋值为 N-1- 1, 如果变量节点累加更新得到的更新值在区间
[ ^-l ^-l]范围内, 则将变量节点累加更新得到的更新值作为更新后的变 量节点外信息的值; 这样, 每次迭代更新后的变量节点外信息仍可用 4比特 的二进制数据表示;
最后, 译码单元中的第二存储单元存储更新后的变量节点外信息。
例如, N为 4时, 将输入的 4比特的信道信息和校验节点外信息每两个 分为一组进行分级累加, 分级累加过程中运算数据的数据位宽随累加级数逐 级扩展, 各级累加结果的数据位宽在前一级数据位宽的基础上扩展一位; 如 果变量节点的节点度数为 dv, 则该变量节点外信息的更新过程中需要累加的 数据包括 1个信道信息和 个校验节点外信息, 共有 +1个累加数据, 求 dv+l个数据的累加和值至少需要进行 log2 « + l)级累加运算, 因此变量节点的 累加更新得到的更新值的数据位宽在输入的数据位宽的基础上扩展 「log2 ( + l 位, 这样, 变量节点外信息的更新过程中的数据计算精度高于 4 比特, 变量节点外信息的更新结束后得到的高量化精度数据的数据位宽为 4+「log2 ( + 1),比特。
对于变量节点外信息的分级累加的更新处理过程,以 16输入的变量节点 为例, 如图 3所示, 输入变量节点更新单元的累加单元的 16个信息数据均为 4比特数据, 将 16个 4比特数据按每两个一组, 分为 8组, 分组进行累加; 第一级累加过程结束后, 8 个累加单元输出的累加结果在输入数据的基础上 增加一位数据位宽, 为 5比特数据; 第二级累加、 第三级累加和第四级累加 的方式与第一级累加相同; 第二级累加后 4个累加单元输出的结果为 6比特 数据, 第三级累加后 2个累加单元输出的结果为 7比特数据, 第四级累加后 累加单元输出结果为 8比特数据, 最后通过查寻优化设计的量化精度转化表 的方法, 将 8比特的累加更新结果转化为 4比特的更新的变量节点外信息。
步骤 205, 判断译码是否结束;
具体的,当后验概率信息 n≥0时,译码码字 „=0, η<0,则译码码字 „=1 , 这样就生成了译码码字向量 ; 如果校验矩阵 H与译码码字向量 的乘积满 足偶校验条件, 即 H = 0 (mod 2;>能够成立, 则译码单元中的判断单元确定译 码成功并结束译码流程; 如果该乘积不满足偶校验条件, 则判断单元判断迭 译码代迭次数是否超过预设的最大迭代次数 itermax,如果译码迭代次数未超过 预设的最大迭代次数 itermax, 则重复步骤 203至步骤 205; 如果译码迭代次数 超过预设的最大迭代次数 itermax, 则结束译码过程并声明译码失败。
应用实例
本实施例以 IEEE 802.16e标准中码长为 576比特, 码率为 1/2的 LDPC 为例, 并结合图 4至图 7进行说明。
硬件实现译码单元之前,通过针对 IEEE 802.16e标准中码长为 576比特, 码率为 1/2的 LDPC仿真, 优化得到信道信息的 4比特非均勾量化方案的 7 个量阶值分别为 =0.0625、 q2=0.1875、 q3=0.375、 q4=0.625、 q5=l、 q6=1.5、 q7=2。
根据仿真优化得到的信道信息 4比特非均勾量化方案对接收的信道信息 进行量化, 将量化后的 4比特的信道信息输入译码单元作为变量节点的的先 验概率信息。 由于码长为 576比特, 所以译码单元每次输入 576个位宽为 4 比特的信道信息, 分别存入 576个位宽为 4比特的信道信息存储单元。 同时, 用 576个信道信息分别初始化与其对应的数据位宽为 4比特的变量节点外信 息。
在校验节点更新单元中, 将与对应的校验节点相关的变量节点外信息的 符号位进行异或运算, 得到更新的校验节点外信息的符号; 校验节点外信息 的模值更新过程中 , 将与校验节点相关联的变量节点外信息的模值按每两个 分为一组,通过分级比较变量节点外信息的模值得到其中的最小值和次小值, 再将最小值和次小值与量化的偏移量的值。 当译码迭代次数未达到迭代次数 门限值 iterad时,量化的偏移量为 0。当译码迭代次数达到迭代次数门限值 iterad 时, 需要判断校验节点对应的校验约束关系在前一次迭代译码过程中是否得 到满足, 若校验约束关系在前一次迭代译码过程中得到满足, 则量化的偏移 量取 0; 如果没有得到满足, 量化的偏移量取 1。
最后判断最小值和次小值与量化的偏移量的差值与 0之间的大小关系, 结合符号位得到校验节点传递给相关的变量节点的数据位宽为 4比特的校验 节点外信息。 将更新的校验节点外信息分别存入对应的数据位宽为 4比特的 第一存储单元。
由于 IEEE 802.16e标准中码长为 576比特, 码率为 1/2的 LDPC码只有 节点度数为 6和 7的校验节点, 因此校验节点更新单元中釆用数据位宽为 3 比特的 6输入和 7输入比较单元实现分级比较变量节点外信息的模值并得到 其中的最小值和次小值。 釆用本专利提出的低量化精度的 LDPC迭代译码方 法的 IEEE 802.16e标准中码长为 576比特, 码率为 1/2的 LDPC译码器中, 校验节点更新单元中数据位宽为 3比特的 6输入比较单元框图如图 4所示, 7 输入比较单元框图如图 5所示。 通过该比较单元可获得输入的变量节点外信 息的最小值和次小值以及变量节点外信息的最小值所对应的变量节点的标 号。
在变量节点更新单元中, 对输入的 4比特的信道信息和校验节点外信息 按每两个值分为一组, 进行分级累加处理; 由于 IEEE 802.16e标准中码长为 576比特, 码率为 1/2的 LDPC中包含节点度数为 2、 3和 6的变连接点, 因 此译码单元中的变量节点更新单元中需要 2输入、 3输入、 4输入、 6输入和 7输入的分级累加单元。 其中, 7输入的分级累加单元框图如图 6所示。 7输 入累加单元的输入数据为 4比特的信道信息和校验节点外信息。 在第一级累 加过程中, 将前 6个输入数据分三组分别累加, 累加结果为 5比特的数据。 将第 7个数据的数据位宽扩展一位。 第二级累加和第三级累加重复第一级累 加的过程, 将累加和结果的数据位宽扩展至 7比特。 最后将累加更新得到的 7 比特数据转化为 4比特数据作为更新的变量节点外信息和译码后验概率信 息。 本实施例中釆用将 7比特数据转化为 4比特数据釆用直接截位限幅的方 法。
变量节点外信息的更新过程中, 变量节点更新单元每更新一个译码后验 概率信息或变量节点外信息, 便将其存入对应的数据位宽为 4比特的译码后 验概率信息存储单元(如译码单元中的第三存储单元)或变量节点外信息存 储单元(如译码单元中的第二存储单元) 。
变量节点外信息的更新过程结束后, 译码单元提取出 576个译码后验概 率信息的符号位作为本次迭代译码得出的译码码字向量, 带入校验方程中进 行校验。 如果译码码字向量满足校验方程, 则结束迭代译码过程, 输出译码 码字。 如果译码码字向量满足校验方程, 若迭代次数未超过预设的最大迭代 次数 itermax, 则重复迭代译码过程; 若迭代次数超过预设的最大迭代次数 itermax, 则结束译码过程并输出译码码字。
实施例中设计的低量化精度的 LDPC迭代译码方法的误比特率性能如图 7所示, 图中虚线表示本发明实施例提出的低量化精度的 LDPC迭代译码方 法的误比特率, 图中实线表示浮点 BP译码算法的误比特率,在误比特率达到 1 10—8时, 本发明实施例提出的低量化精度的 LDPC迭代译码方法的误比特 率性能与浮点 BP译码算法相比较,性能差距约为 0.2 dB。这说明本发明实施 例提出的低量化精度的 LDPC迭代译码方法在降低译码单元的硬件实现复杂 度和存储资源消耗的同时, 引起的译码性能损失较小。
本领域普通技术人员可以理解上述方法中的全部或部分步骤可通过程序 来指令相关硬件完成, 所述程序可以存储于计算机可读存储介质中, 如只读 存储器、 磁盘或光盘等。 可选地, 上述实施例的全部或部分步骤也可以使用 一个或多个集成电路来实现, 相应地, 上述实施例中的各模块 /单元可以釆用 硬件的形式实现, 也可以釆用软件功能模块的形式实现。 本发明不限制于任 何特定形式的硬件和软件的结合。
需要说明的是, 本发明还可有其他多种实施例, 在不背离本发明精神及 的改变和变形, 但这些相应的改变和变形都应属于本发明所附的权利要求的 保护范围。
工业实用性 本发明实施例的技术方案能够降低 LDPC译码器的存储资源和硬件实现 复杂度, 同时能够保证译码过程中的运算精度, 降低由于量化引起的译码性 能损失。

Claims

权 利 要 求 书
1、 一种迭代译码系统, 该系统包括:
量化单元, 其设置为: 对接收的信道信息进行非均勾量化处理; 以及 译码单元, 其设置为: 对变量节点的先验概率信息进行初始化处理, 将 量化后的信道信息作为变量节点的先险 ¾^率信息; 才艮据变量节点的先险 ¾^率 信息, 并利用自适应的偏移量进行校验节点外信息的更新处理; 才艮据更新后 的校验节点外信息, 对变量节点外信息进行分级累加的更新处理; 以及判断 译码是否结束。
2、 根据权利要求 1所述的系统, 该系统还包括: 信道信息存储单元, 其 设置为存储量化后的信道信息。
3、 根据权利要求 1或 2所述的系统, 其中, 所述译码单元包括: 初始化单元, 其设置为对变量节点的先验概率信息进行初始化处理, 将 量化后的信道信息作为变量节点的先险^^率信息;
校验节点更新单元, 其设置为根据变量节点的先验概率信息, 并利用自 适应的偏移量进行校验节点外信息的更新处理;
变量节点更新单元, 其设置为根据更新后的校验节点外信息, 对变量节 点外信息进行分级累加的更新处理; 以及
判断单元, 其设置为判断译码是否结束。
4、 根据权利要求 3所述的系统, 该系统还包括:
第一存储单元, 其设置为存储更新后的校验节点外信息; 以及
第二存储单元, 其设置为存储更新后的变量节点外信息。
5、 一种迭代译码方法, 该方法包括:
对接收的信道信息进行非均勾量化处理;
对变量节点的先验概率信息进行初始化处理, 将量化后的信道信息作为 变量节点的先险^^率信息; 根据变量节点的先验概率信息, 并利用自适应的偏移量进行校验节点外 信息的更新处理;
根据更新后的校验节点外信息, 对变量节点外信息进行分级累加的更新 处理; 以及
判断译码是否结束。
6、 根据权利要求 5所述的方法, 其中, 所述对输入译码单元的信道信息 进行非均勾量化处理的步骤包括:
根据仿真优化得到的信道信息 N比特非均匀量化方案, 对接收到的用于 输入译码单元的信道信息进行 N比特非均匀量化处理,所述 N为 4或 5或 6; 存储量化后的信道信息。
7、 根据权利要求 6所述的方法, 其中, 所述对接收到的用于输入译码单 元的信道信息进行 N比特非均勾量化处理的步骤包括:
对接收到的用于输入译码单元的信道信息进行 N比特非均勾量化处理, 将接收到的信道信息量化为:
-2
Figure imgf000021_0001
其中, 为量阶值, i为量化值, ^为接收到的信道信息, 为量化后的 信道信息。
8、 根据权利要求 5所述的方法, 其中, 所述根据变量节点的先验概率信 息, 并利用自适应的偏移量进行校验节点外信息的更新处理的步骤包括: 偏移量根据译码迭代次数进行自适应调整;
将与校验节点相关联的变量节点外信息按照每两个分为一组, 通过分级 比较变量节点外信息的模值, 得到变量节点外信息的模值中的最小值, 再将 得到的最小值与偏移量的值进行比较, 如果所述最小值大于偏移量的值, 则 更新后的校验节点外信息等于所述最小值与偏移量的值的差值; 如果所述最 小值小于等于偏移量的值, 则更新后的校验节点外信息等于 0; 存储更新后的校验节点外信息。
9、 根据权利要求 8所述的方法, 其中, 所述偏移量根据译码迭代次数进 行自适应调整的步骤包括:
译码迭代次数小于迭代次数门限值时, 釆用的偏移量为 0; 译码迭代次 数达到迭代次数门限值时, 如果校验节点对应的校验约束关系在前一次迭代 译码过程中得到满足, 则在本次迭代译码过程中, 釆用的偏移量为 0; 如果 校验节点对应的校验约束关系在前一次迭代译码过程中未得到满足, 则在本 次迭代译码过程中, 釆用的偏移量为 1。
10、 根据权利要求 5所述的方法, 其中, 所述根据更新后的校验节点外 信息, 对变量节点外信息进行分级累加的更新处理的步骤包括:
将校验节点外信息按照每两个值一组进行划分, 将 N比特非均勾量化的 变量节点先验概率信息和每组校验节点外信息进行分级累加处理, 每进行一 级的累加, 就将累加后得到的和值的位宽增加一个比特;
最后,如果变量节点累加更新得到的更新值大于 S^-l , 则更新后的变量 节点外信息为 2^-1 , 如果变量节点累加更新得到的更新值小于 ^-l , 则更 新后的变量节点外信息为 ^-l,如果变量节点累加更新得到的更新值在区间
[ ^-l ^-l]内, 则将变量节点累加更新得到的更新值作为更新后的变量节 点外信息的值; 存储更新后的变量节点外信息。
11、 根据权利要求 5所述的方法, 其中, 所述判断译码是否结束的步骤 包括:
如果校验矩阵 H与译码码字向量 的乘积满足偶校验条件, 则译码成功 并结束译码流程; 如果所述乘积不满足偶校验条件, 则判断迭译码代迭次数 是否超过预设的最大迭代次数 itermax,如果译码迭代次数未超过预设的最大迭 代次数 itermax,则继续进行校验节点外信息、变量节点外信息的更新处理并判 断译码是否结束;如果译码迭代次数超过预设的最大迭代次数 itermax,结束译 码过程并声明译码失败。
PCT/CN2012/078686 2012-02-07 2012-07-16 一种迭代译码方法及系统 WO2013117076A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201210026479.7A CN102545913B (zh) 2012-02-07 2012-02-07 一种迭代译码方法及系统
CN201210026479.7 2012-02-07

Publications (1)

Publication Number Publication Date
WO2013117076A1 true WO2013117076A1 (zh) 2013-08-15

Family

ID=46351983

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2012/078686 WO2013117076A1 (zh) 2012-02-07 2012-07-16 一种迭代译码方法及系统

Country Status (2)

Country Link
CN (1) CN102545913B (zh)
WO (1) WO2013117076A1 (zh)

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105978578A (zh) * 2016-04-29 2016-09-28 清华大学 低密度奇偶校验码和积译码运算数值的非均匀量化方法
CN109889209A (zh) * 2019-03-08 2019-06-14 上海航天测控通信研究所 一种适用于宇航通信的速率自适应型ldpc译码器
CN109921802A (zh) * 2019-02-26 2019-06-21 北京中科晶上科技股份有限公司 一种qc-ldpc码的译码方法、模块及装置
CN110752852A (zh) * 2019-09-26 2020-02-04 中科睿微(宁波)电子技术有限公司 极化码的bp译码方法、装置、系统、设备及存储介质
CN111475326A (zh) * 2019-01-23 2020-07-31 深圳衡宇芯片科技有限公司 训练人工智能执行低密度奇偶检查码的译码程序的方法
CN111654292A (zh) * 2020-07-20 2020-09-11 中国计量大学 一种基于动态阈值的分裂简化极化码连续消除列表译码器
CN111835364A (zh) * 2020-08-03 2020-10-27 辽宁工程技术大学 一种极化码的低复杂度神经bp译码方法
CN112039534A (zh) * 2020-08-12 2020-12-04 西南交通大学 一种ldpc译码方法、装置、设备及存储介质
CN112350737A (zh) * 2020-11-23 2021-02-09 南京信息工程大学滨江学院 一种基于ldpc码的分组信息更新的传输方法
CN113381769A (zh) * 2021-06-25 2021-09-10 华中科技大学 一种基于fpga的译码器及其设计方法
CN113437979A (zh) * 2021-06-30 2021-09-24 华侨大学 一种基于非均匀信源的原模图ldpc码的结构优化方法及装置
CN113572482A (zh) * 2021-08-03 2021-10-29 南京大学 一种仅基于最小值的ldpc码最小和译码方法
CN113794478A (zh) * 2021-09-06 2021-12-14 深圳市极致汇仪科技有限公司 一种基于噪声功率的ldpc分步译码方法及系统
CN114006621A (zh) * 2021-11-05 2022-02-01 中国传媒大学 一种并行译码方法及系统
CN114584151A (zh) * 2022-02-28 2022-06-03 北京理工大学 基于概率计算的模拟译码电路停止准则的译码方法
CN115632921A (zh) * 2022-10-08 2023-01-20 北京理工大学 基于阈值检测的编码辅助盲帧同步方法及系统
CN113612485B (zh) * 2021-08-03 2024-04-16 深圳宏芯宇电子股份有限公司 一种译码方法、译码装置、设备及存储装置

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102545913B (zh) * 2012-02-07 2015-05-27 中兴通讯股份有限公司 一种迭代译码方法及系统
WO2015168862A1 (zh) * 2014-05-06 2015-11-12 华为技术有限公司 一种数据处理设备和方法
CA2968442A1 (en) * 2014-11-19 2016-05-26 Lantiq Beteiligungs-GmbH & Co.KG Ldpc decoding with finite precision and dynamic adjustment of the number of iterations
BR112017000548B1 (pt) 2015-10-13 2023-04-11 Huawei Technologies Co., Ltd Dispositivo e método de decodificação e sistema de transmissão de sinal
CN105680881A (zh) * 2016-01-08 2016-06-15 广西大学 Ldpc译码方法及译码器
CN106374940A (zh) * 2016-11-14 2017-02-01 中国电子科技集团公司第五十四研究所 一种多进制ldpc译码方法及译码器
CN109831214A (zh) * 2018-12-29 2019-05-31 中国电子科技集团公司第二十研究所 一种全并行ldpc译码器及fpga实现方法
CN111649786B (zh) * 2020-06-03 2022-04-08 东莞深证通信息技术有限公司 数据累计方法、装置、终端设备及存储介质
CN113098531B (zh) * 2021-04-19 2022-04-29 中南林业科技大学 一种基于最小和译码框架的动态偏移补偿方法
CN113285723B (zh) * 2021-04-26 2022-09-30 武汉梦芯科技有限公司 一种ldpc译码过程中校验节点更新方法、系统及存储介质
CN114513211B (zh) * 2022-02-15 2023-06-06 电子科技大学 基于全相关序列的混合概率ldpc译码器
CN114584259B (zh) * 2022-02-18 2024-02-09 阿里巴巴(中国)有限公司 译码方法、装置、设备及存储介质
CN115603761A (zh) * 2022-09-27 2023-01-13 北京邮电大学(Cn) 基于校验置信度的ldpc译码方法和装置

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1773867A (zh) * 2004-11-08 2006-05-17 华为技术有限公司 Turbo码译码方法
CN1852029A (zh) * 2006-05-26 2006-10-25 清华大学 采用可变范围均匀量化的低密度奇偶校验码译码方法
EP1865605A1 (en) * 2006-06-07 2007-12-12 Daewoo Electronics Corporation Method and device for decoding low-density parity check code and optical information reproducing apparatus using the same
CN101534166A (zh) * 2008-03-10 2009-09-16 上海明波通信技术有限公司 准循环低密度奇偶校验码解码器及解码方法
US20100174959A1 (en) * 2009-01-06 2010-07-08 Samsung Electronics Co., Ltd. Decoding method and memory system device using the same
US20110087946A1 (en) * 2009-10-09 2011-04-14 Stmicroelectronics, Sa Low complexity finite precision decoders and apparatus for ldpc codes
US20110145675A1 (en) * 2009-12-15 2011-06-16 International Business Machines Corporation Calculation technique for sum-product decoding method (belief propagation method) based on scaling of input log-likelihood ratio by noise variance
CN102545913A (zh) * 2012-02-07 2012-07-04 中兴通讯股份有限公司 一种迭代译码方法及系统

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101854179B (zh) * 2010-05-26 2012-09-05 厦门大学 一种应用于ldpc译码的5比特量化方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1773867A (zh) * 2004-11-08 2006-05-17 华为技术有限公司 Turbo码译码方法
CN1852029A (zh) * 2006-05-26 2006-10-25 清华大学 采用可变范围均匀量化的低密度奇偶校验码译码方法
EP1865605A1 (en) * 2006-06-07 2007-12-12 Daewoo Electronics Corporation Method and device for decoding low-density parity check code and optical information reproducing apparatus using the same
CN101534166A (zh) * 2008-03-10 2009-09-16 上海明波通信技术有限公司 准循环低密度奇偶校验码解码器及解码方法
US20100174959A1 (en) * 2009-01-06 2010-07-08 Samsung Electronics Co., Ltd. Decoding method and memory system device using the same
US20110087946A1 (en) * 2009-10-09 2011-04-14 Stmicroelectronics, Sa Low complexity finite precision decoders and apparatus for ldpc codes
US20110145675A1 (en) * 2009-12-15 2011-06-16 International Business Machines Corporation Calculation technique for sum-product decoding method (belief propagation method) based on scaling of input log-likelihood ratio by noise variance
CN102545913A (zh) * 2012-02-07 2012-07-04 中兴通讯股份有限公司 一种迭代译码方法及系统

Cited By (26)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105978578B (zh) * 2016-04-29 2019-03-29 清华大学 低密度奇偶校验码和积译码运算数值的非均匀量化方法
CN105978578A (zh) * 2016-04-29 2016-09-28 清华大学 低密度奇偶校验码和积译码运算数值的非均匀量化方法
CN111475326A (zh) * 2019-01-23 2020-07-31 深圳衡宇芯片科技有限公司 训练人工智能执行低密度奇偶检查码的译码程序的方法
CN109921802B (zh) * 2019-02-26 2023-02-07 北京中科晶上科技股份有限公司 一种qc-ldpc码的译码方法、模块及装置
CN109921802A (zh) * 2019-02-26 2019-06-21 北京中科晶上科技股份有限公司 一种qc-ldpc码的译码方法、模块及装置
CN109889209A (zh) * 2019-03-08 2019-06-14 上海航天测控通信研究所 一种适用于宇航通信的速率自适应型ldpc译码器
CN110752852A (zh) * 2019-09-26 2020-02-04 中科睿微(宁波)电子技术有限公司 极化码的bp译码方法、装置、系统、设备及存储介质
CN110752852B (zh) * 2019-09-26 2023-10-03 浙江科睿微电子技术有限公司 极化码的bp译码方法、装置、系统、设备及存储介质
CN111654292A (zh) * 2020-07-20 2020-09-11 中国计量大学 一种基于动态阈值的分裂简化极化码连续消除列表译码器
CN111654292B (zh) * 2020-07-20 2023-06-02 中国计量大学 一种基于动态阈值的分裂简化极化码连续消除列表译码器
CN111835364A (zh) * 2020-08-03 2020-10-27 辽宁工程技术大学 一种极化码的低复杂度神经bp译码方法
CN111835364B (zh) * 2020-08-03 2023-11-14 辽宁工程技术大学 一种极化码的低复杂度神经bp译码方法
CN112039534B (zh) * 2020-08-12 2023-05-02 西南交通大学 一种ldpc译码方法、装置、设备及存储介质
CN112039534A (zh) * 2020-08-12 2020-12-04 西南交通大学 一种ldpc译码方法、装置、设备及存储介质
CN112350737B (zh) * 2020-11-23 2023-12-12 南京信息工程大学滨江学院 一种基于ldpc码的分组信息更新的传输方法
CN112350737A (zh) * 2020-11-23 2021-02-09 南京信息工程大学滨江学院 一种基于ldpc码的分组信息更新的传输方法
CN113381769A (zh) * 2021-06-25 2021-09-10 华中科技大学 一种基于fpga的译码器及其设计方法
CN113437979B (zh) * 2021-06-30 2023-05-16 华侨大学 一种基于非均匀信源的原模图ldpc码的结构优化方法及装置
CN113437979A (zh) * 2021-06-30 2021-09-24 华侨大学 一种基于非均匀信源的原模图ldpc码的结构优化方法及装置
CN113572482A (zh) * 2021-08-03 2021-10-29 南京大学 一种仅基于最小值的ldpc码最小和译码方法
CN113572482B (zh) * 2021-08-03 2024-04-09 南京大学 一种仅基于最小值的ldpc码最小和译码方法
CN113612485B (zh) * 2021-08-03 2024-04-16 深圳宏芯宇电子股份有限公司 一种译码方法、译码装置、设备及存储装置
CN113794478A (zh) * 2021-09-06 2021-12-14 深圳市极致汇仪科技有限公司 一种基于噪声功率的ldpc分步译码方法及系统
CN114006621A (zh) * 2021-11-05 2022-02-01 中国传媒大学 一种并行译码方法及系统
CN114584151A (zh) * 2022-02-28 2022-06-03 北京理工大学 基于概率计算的模拟译码电路停止准则的译码方法
CN115632921A (zh) * 2022-10-08 2023-01-20 北京理工大学 基于阈值检测的编码辅助盲帧同步方法及系统

Also Published As

Publication number Publication date
CN102545913B (zh) 2015-05-27
CN102545913A (zh) 2012-07-04

Similar Documents

Publication Publication Date Title
WO2013117076A1 (zh) 一种迭代译码方法及系统
US8392795B2 (en) Low density parity check codec and method of the same
JP4062435B2 (ja) 誤り訂正符号復号装置
JP4651600B2 (ja) 低密度パリティ検査復号器における検査ノード更新方法
US8359522B2 (en) Low density parity check decoder for regular LDPC codes
JP4627317B2 (ja) 通信装置および復号方法
CN107370490B (zh) 结构化ldpc的编码、译码方法及装置
JP5483875B2 (ja) Ldpc符号のブロックおよびレートに独立な復号の方法および装置
JP4320418B2 (ja) 復号装置および受信装置
JP4519694B2 (ja) Ldpc符号検出装置及びldpc符号検出方法
US9356623B2 (en) LDPC decoder variable node units having fewer adder stages
CN110784232B (zh) 一种空间耦合ldpc码滑窗译码方法
Balatsoukas-Stimming et al. A fully-unrolled LDPC decoder based on quantized message passing
CN109586732B (zh) 中短码ldpc编解码系统和方法
CN110830050B (zh) 一种ldpc译码方法、系统、电子设备及存储介质
Giard et al. Hardware decoders for polar codes: An overview
TW202205815A (zh) 用於自循環置換矩陣之叢集建構之準循環低密度奇偶檢查碼之垂直分層解碼之方法及裝置
KR102092634B1 (ko) Ldpc 부호 복호기 및 복호 방법
JP2008526086A (ja) チャネルコードを用いた復号化装置及び方法
JP5790029B2 (ja) 復号装置、復号方法、およびプログラム
CN107872231B (zh) Ldpc译码方法与装置
JP6395658B2 (ja) 誤り訂正復号装置、受信装置及び誤り訂正復号方法
KR20090012189A (ko) Ldpc 부호의 성능 개선을 위한 스케일링 기반의 개선된min-sum 반복복호알고리즘을 이용한 복호 장치 및그 방법
Ghaffari et al. Probabilistic gradient descent bit-flipping decoders for flash memory channels
Leduc-Primeau et al. High-throughput LDPC decoding using the RHS algorithm

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 12867873

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 12867873

Country of ref document: EP

Kind code of ref document: A1