CN101471748A - Method for storing and updating LDPC code message - Google Patents
Method for storing and updating LDPC code message Download PDFInfo
- Publication number
- CN101471748A CN101471748A CNA200710307821XA CN200710307821A CN101471748A CN 101471748 A CN101471748 A CN 101471748A CN A200710307821X A CNA200710307821X A CN A200710307821XA CN 200710307821 A CN200710307821 A CN 200710307821A CN 101471748 A CN101471748 A CN 101471748A
- Authority
- CN
- China
- Prior art keywords
- node
- check
- message
- variable
- variable node
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Images
Landscapes
- Error Detection And Correction (AREA)
Abstract
The invention provides a method for storing a variable node message during the decoding of a low density parity check code. The method comprises: utilizing a variable information vector to store a message related to the variable node. The method reduces the consumption of storage units on the premise of preventing the performance loss.
Description
Technical field
The present invention relates to the receiver in the communication system, more specifically, relate to a kind of message stores and update method of LDPC sign indicating number.
Background technology
Chnnel coding researcher is verified: based on belief propagation (belief propagation, brief note is for BP) binary low-density checksum (the low-density parity-check of method, brief note is for LDPC) the iterative decoding performance and the additive white Gaussian noise channel capacity of sign indicating number have only 0.0045dB (referring to S.-Y.Chung, G.D.Forney, Jr., T.J.Richardson, and R.Urbanke.Feb.2001.On the design of low-density parity-check codeswithin 0.0045 dB of the Shannon limit.Vol.5, pp.58-60.IEEE Commun.Lett.) distance.This interative encode method can be described as the message transmission on the LDPC sign indicating number Tanner figure naturally.
Different according to interpretation method and News Category, interative encode method comprises: the exchange of message between calculating, variable node and the check-node of variable node and check-node message on the Tanner figure.Because the sparse characteristic of LDPC sign indicating number Tanner figure, the decoding complexity of its interative computation is lower.(also cry and long-pending interpretation method for the standard BP method, storage side information (comprising variable message and verification message) carries out the iterative decoding computing), its decoding interative computation need be stored all channel informations, variable message (in the BP interpretation method, pass to the message of check-node from variable node) and verification message (the BP interpretation method, passing to the message of variable node from check-node).If the code length of LDPC sign indicating number is represented with N, the average number of degrees d of variable node
vExpression, all message all use the B bit to represent, are (2 * d based on the required memory cell of the ldpc code decoder storing message of standard BP method then
v+ 1) * B * N bit.
Because the LDPC code performance is superior, decoding complexity is low, it becomes forward error correction (forward error correcting, note by abridging the be FEC) sign indicating number of some Modern Communication System first-selections.Second generation digital satellite video broadcast system (Digital video broadcasting satelliteversion 2 for example, brief note is for DVB-S2), digital tv ground broadcasting (the Digitaltelevision terrestrial broadcasting that proposes of Tsing-Hua University, brief note is for China DVB-T), Tai Mei (TiMi) the company mobile multimedia broadcast system (Mobile multimedia broadcasting notes the CMMB for China by abridging) that proposes all used the LDPC sign indicating number as the cascaded code of ISN as its chnnel coding.For the mobile receiver of these broadcast systems, require ldpc code decoder power, space complexity and time complexity low more good more.A kind of method that reduces the ldpc code decoder space complexity is to reduce the total amount of storing message under the prerequisite that does not reduce decoding performance.
Based on minimum and (Min-Sum, brief note is MS) method, list of references 2 (J.Dielissen, A.Hekstra, and V.Berg.Low cost LDPC decoder for DVB-S2.2006.Philips Research, High Tech Campus 5,5656 AE Eindhoven, TheNetherlands.) propose for each one of variable node storage and (all check-nodes that link to each other with this variable node pass to this variable node verification message with), be that each check-node is stored a vector (amplitude that comprises minimum value, the amplitude of sub-minimum, the index of a minimum value and dc symbol).Herein, dc is the number of degrees of check-node, is constant for regular LDPC sign indicating number, and irregular LDPC is a variable.If the code check of LDPC sign indicating number represents that with R adopting the required memory cell of decoder storing message of the method is (B+ (1-R) (2B+log2 (dc)+dc)) * N bit.
(flooding schedule is (referring to F.R.Kschischangand B.J.Frey based on flood information transmission policy, and H.A.Loeliger, Factor graphs and the sum-productalgorithm.2001.IEEE Trans.Inf.Theory, 47 (2), pp.498～519)) BP interpretation method (just above the standard BP method of indication) is well known in the art, so only provide here based on triggering message transmission policy (trigger schedule), it with the check-node BP method at center (check node central, note by abridging be CN_C).Before method is described, provide to give a definition.
(0≤n≤N-1) is defined as the set of the check-node of link variable node n to M (n).(0≤m≤M-1) is defined as the set of the variable node that participates in m verification computing to N (m).M (n) m represent from set M (n), to remove the set of last element behind the element m.N (m) n represent from set N (m), to remove the set of last element behind the element n.In addition, in addition qn → m (0) (qn → m (1)) be based on gather M (n) m, what variable node n passed to check-node m is 0 (1) conditional probability about variable node n; Similar, rm → n (0) (rm → n (1)) be defined as based on set N (m) n, what check-node m passed to variable node n is 0 (1) conditional probability about variable node n.At last, make x=(x0, x1 ..., xN-1) and y=(y0, y1 ..., yN-1) be respectively transmission vector and receive vector.
In the probability territory, the BP interpretation method be input as posterior probability (a posterioriprobabilities, note by abridging be APP) q
N → m(0)=P
X/y(x
n=0|y
n) and q
N → m(1)=P
X/y(x
n=1|y
n), these two values are based on all that the channel statistical property calculation comes out.In addition, definition log-likelihood ratio (Log Likelihood Ratio, note by abridging be LLR) LLRs z
N → m(x
n) ≡ log (q
N → m(0)/q
N → m(1)) (variable message) and L
M → n(x
n) ≡ log (r
M → n(0)/r
M → n(1)) (verification message), CN_C BP method can be described as:
Initialization: for each variable node is given a corresponding channel information, all check informations are initialized as zero.
Information updating:
1. make i=0;
2. calculation check message L
I → n(x
i), n ∈ M (i) wherein;
3. upgrade variable message z
N → j(x
n), n ∈ M (i) wherein, j ∈ N (n) i;
4.i++; If i equals M, finish interative computation one time; If i is not equal to M, turn to step 2.
The decoding judgement:
Utilize Z (x
n) 〉=0 x^
n=0 and Z (x
n)＜0 x^
n=1 obtains sending the estimate vector x^=(x^ of code word
n, x^
1..., x^
N-1).Compute codeword adjoint vector x^H
TIf, x^H
T=0, then stop to decipher computing; Otherwise turn to step 1.If the iterations of decoding surpasses the maximum iteration time it that sets
Max, then declare decoding failure.
Above-mentioned information updating strategy is called triggering message transmission policy.For a kind of interpretation method, can adopt multiple message stores and update strategy (for example method that provides in method of using in the standard BP method and the list of references 2).
For the standard BP interpretation method, need storage all channel information, variable message and verification message, the memory cell that is consumed is (2 * d altogether
v+ 1) * B * N bit.The size of memory is with (2 * d
v+ 1) * and B is that the factor is directly proportional with code length N, when N infinitely increased, memory can't be practical with big.
In order to reduce the quantity of the required memory cell of storing message, list of references 2 has provided a kind of message stores method: for one of each variable node storage and, be vector of each check-node storage.The number of the memory cell that this storage means is required is (B+ (1-R) * (2B+log
2(d
c)+d
c)) * the N bit.For example, be 10 at a code length
4(3,6) regular LDPC sign indicating number, for the standard BP method, need the message memory of 70 * B k bit; For the method in the document 2, need the message of storage 20 * B+45 k bit.Suppose that B equals 6, compare that the method for document 2 can be saved the memory cell of 255k bit with the standard BP method.
Although the message stores method of using in the document 2 has effectively reduced memory space, this method is only applicable to the MS method.As everyone knows, compare with the standard BP method, the MS method has the performance loss of 1dB nearly.In addition, for MS (actual decoder all the is a fixed-point calculation) method of fixed point, use this method to recover further to introduce loss of significance in the verification message in the information updating process, this has more aggravated the deterioration of decoding performance.
From the storage means of two kinds of top message as can be seen: the storage means of standard is used more memory cell, but can obtain good performance; The method that document 2 provides is the conserve memory unit greatly, but has introduced performance loss.
Summary of the invention
According to the present invention, a kind of message stores method is proposed, realized under the prerequisite of not losing performance, using less memory cell.
In order to reduce the use amount of message storage under the condition of not introducing performance loss, the present invention has designed a variable node message vector for variable node, and is corresponding, proposed a kind of replacement criteria of variable node message.Can see that from following argumentation this method is not limited to any interpretation method, and itself can not introduce performance loss.
According to an aspect of the present invention, the storage means of variable node message comprises the related news that use the variable message vector to come the storage of variables node in a kind of low density parity check code decoding.
According to a further aspect in the invention, the update method of variable node message in a kind of low density parity check code decoding, wherein use the variable message vector to come the related news of storage of variables node, described variable message vector comprises channel information and variable node external information, wherein utilize the check-node message upgraded and remove this check-node and the check-node that upgraded all do not upgrade verification message with upgrade described variable node external information.
The method that the present invention proposes is specially adapted to CN_C BP method, for this method, need not to store the iteration update calculation that verification message just can be carried out message.Therefore, for the ldpc code decoder that uses CN_C BP method and message stores provided by the present invention and update method, need ((d
v+ 1) * B+d
vThe memory cell of) * N bit.With above-mentioned is example, and the memory cell of using is the 270k bit, is slightly larger than the required memory cell of list of references 2 methods much smaller than the standard BP method.
Description of drawings
Fig. 1 shows the schematic diagram according to variable node message vector of the present invention
Fig. 2 shows the schematic diagram according to verification message of the present invention and variable external information
Fig. 3 shows the schematic diagram according to variable node message vector v of the present invention
Fig. 4 shows the schematic diagram according to variable node message renewal process of the present invention
Embodiment
Message stores
The present invention has designed an information vector and has been used for the information of storage of variables node.Here, represent the variable node message vector with VN_M, it constitutes as shown in Figure 1.Elements C h_M (B bit) is corresponding to the channel information of this variable node.VM_E is defined as the variable node external information, and VN_F is defined as the variable node mark.VM_E is used for storing and the renewal variable message; VN_F is used to control iterative decoding process.Element VM_E (0) (B bit) is the poor of variable message (this variable node is passed to the variable message of first check-node) and channel massage.The current state of first check-node that element VN_F (0) (1 bit) expression links to each other with this variable node.If first check-node is the current node of handling, VN_F (0)=1 then, otherwise VN_F (0)=0.VM_E (1) ..., VM_E (d
v-1) (B bit) and VN_F (1) ..., VN_F (d
v-1) definition of (1 bit) is similar with VN_F (0) to VM_E (0).
It should be noted that: variable node mark VN_F is the control signal of code check node processing in the CN_C BP method, if use other BP method, this part can be deleted.In addition, Ch_M, VM_E, putting in order of VN_F is arbitrarily.
From the definition of VN_M, can obtain variable message (variable message, note by abridging be V_M).Pass to i (i=0,1 ..., d
v-1) the variable message V_M of individual check-node (i) can calculate with following formula (1):
V_M(i)＝VM_E(i)+Ch_M. (1)
The V_M that use formula (1) obtains can utilize the more new formula of verification message in the standard BP method or its various reduced forms to come verification message is upgraded.In addition, can utilize VN_M to calculate the hard decision information (decision message, note by abridging be D_M) of variable node, as shown in Equation (2).
D_M＝Ch_M+(∑
i∈{0，1，...，dv-1}VM_E(i))/(d
v-1) (2)
Information updating
Below, will provide the update rule of variable node message vector VN_M.Connect d to one
vIndividual check-node c
0, c
1..., c
Dv-1Variable node v, the verification message that defines the it time iteration is respectively C_M (it, 0), C_M (it, 1) ..., C_M (it, d
v-1), the variable external information is VM_E (it, 0), VM_E (it, 1) ..., VM_E (it, d
v-1).Can obtain following formula (3) from the definition of variable external information and check information.
VM_E(it，j)＝∑
k∈N(v)\jC_M(it，k) (3)
According to formula (3), can get verification message and be:
C_M(it，j)＝(∑
k∈N(v)\jVM_E(it，k)-(d
v-2)VM_E(it，j))/(d
v-1)
(4)
Suppose in (it+1) inferior iterative process check-node c
jMessage C_M (it, (it+1, remains unchanged at j) (this moment element VN_F (j)=1 in VN_M) and other verification message is not upgraded j) to be updated to C_M.Utilize to trigger the information updating strategy, message VM_E (it+1, k) k ∈ N (v) j (it+1 should upgrade after j) immediately obtaining C_M.According to formula (3) and (4), the more new formula of message can be write as:
VM_E(it+1，k)＝∑
l∈N(v)\k，jC_M(it，l)+C_M(it+1，j)
＝C_M(it+1，j)+VM_E(it，k)-(∑
l∈N(v)\jVM_E(it，l)-(d
v-2)VM_E(it，j))/(d
v-1)
(5)
Here, j=0,1 ..., d
v-1 and k ∈ N (v) j.
Use triggers the information updating strategy and according to variable message vector update rule of the present invention, the iterative decoding computing for CN_C BP method no longer needs to store verification message.Compare the quantity that has reduced required message storage with the storage means of standard.In addition, message stores proposed by the invention and update method are not subjected to the restriction of verification message update rule, this means that method itself can't introduce performance loss as list of references.
Utilize example to be set forth in the storage and the update method of variable node message under the CN_C BP method situation below.
Connect 3 check-node c for one
0, c
1, c
2Variable node v, in the it time interative computation, the verification message of inflow is respectively C_M (it, 0), C_M (it, 1), C_M (it, 2); The variable external information is respectively VM_E (it, 0), VM_E (it, 1), and VM_E (it, 2), as shown in Figure 2.
Can obtain formula (6)～(8) according to formula (3).
VM_E(it，0)＝C_M(it，1)+C_M(it，2) (6)
VM_E(it，1)＝C_M(it，0)+C_M(it，2) (7)
VM_E(it，2)＝C_M(it，0)+C_M(it，1) (8)
Utilize formula (4) can obtain formula (9)～(11):
C_M(it，0)＝(VM_E(it，1)+VM_E(it，2)-VM_E(it，0))/2 (9)
C_M(it，1)＝(VM_E(it，0)+VM_E(it，2)-VM_E(it，1))/2 (10)
C_M(it，2)＝(VM_E(it，0)+VM_E(it，1)-VM_E(it，2))/2 (11)
Suppose check-node c in (it+1) inferior iterative process
0Message C_M (it, 0) be updated to C_M (it+1,0), and other verification message are not upgraded, and keep C_M (it, 1) and C_M (it, 2) constant (the element VN_F (0)=1 among the VN_M, VN_F (1)=0, VN_F (2)=0) at this moment.Utilize to trigger the information updating strategy, updating message VM_E (it+1,1) and VM_E (it+1,2) at once after obtaining C_M (it+1,0), as shown in Figure 4.According to formula (5), the variable external information is upgraded with following (12) and (13) calculating formula.Utilize formula (2) so only to utilize the variable message vector just to finish the renewal of variable message in the hope of new variable message.
VM_E(it+1，1)＝C_M(it+1，0)+C_M(it，2)
＝C_M(it+1，0)+VM_E(it，1)-(VM_E(it，1)+VM_E(it，2)-VM_E(it，0))/2
＝C_M(it+1，0)+(VM_E(it，0)+VM_E(it，1)-VM_E(it，2))/2
VM_E(it+1，2)＝C_M(it+1，0)+C_M(it，1)
＝C_M(it+1，0)+VM_E(it，2)-(VM_E(it，1)+VM_E(it，2)-VM_E(it，0))/2
＝C_M(it+1，0)+(VM_E(it，0)+VM_E(it，2)-VM_E(it，1))/2
(13)
The present invention proposes a kind of new message stores method.Wherein variable node message is stored as a vectorial VN_M rather than is stored as d
vIndividual variable message.In addition, based on described message stores method a kind of information updating rule has been proposed.
The storage of above-mentioned message and update method are blanket, can be applied to LDPC sign indicating number iterative decoding algorithm arbitrarily.Storage and update method according to message of the present invention are particularly useful for CN_C BP algorithm.For this algorithm, the verification message of using in the iterative decoding process does not need storage.Use technology provided by the invention, ldpc code decoder can be saved a large amount of memory cell, and this is very useful concerning mobile receiver.In addition, message stores provided by the invention and update method can not brought the performance loss of ldpc code decoder.
Claims (10)
1. the storage means of variable node message comprised the related news that use the variable message vector to come the storage of variables node during a low density parity check code was deciphered.
2. method according to claim 1, wherein, described variable message vector comprises channel information and variable node external information.
3. method according to claim 2, wherein, described variable message vector also comprises the variable node mark, is used for controlling the processing of CN_CBP method check-node.
4. method according to claim 2, wherein, the described variable node variable message (V_M (i)) that passes to i check-node equals and the channel information sum of i check-node corresponding variable node external information (VM_E (i)) and described variable node.
5. method according to claim 2, wherein, the decoding decision message (D_M) of described variable node equal the channel information of described variable node and external information that all import this variable node and.
6. the update method of variable node message during a low density parity check code is deciphered, wherein use the variable message vector to come the related news of storage of variables node, described variable message vector comprises channel information and variable node external information, wherein utilize the check-node message upgraded and remove this check-node and the check-node that upgraded all do not upgrade verification message with upgrade described variable node external information.
7. method according to claim 6, wherein, the variable node external information that v variable node passes to k check-node in the it+1 time iterative process equals to add j the check-node information that the it+1 time iteration upgraded except that the check-node information sum k and j the check-node, wherein j is the sequence number of the check-node that has been updated in the it+1 time interative computation, wherein, the verification message of utilizing the corresponding variable node external information of not upgrading not upgraded.
8. method according to claim 7 wherein, is recovered j the check-node information that the it time iteration do not upgraded according to the variable node external information of upgrading and do not upgrade.
9. method according to claim 8 is wherein, that j the check-node information that the it time iteration do not upgraded equals variable node external information all except that j variable node external information and deduct (d
v-2) Bei j variable node external information is divided by (d
v-1) merchant.
10. method according to claim 7, wherein, the it time iteration check-node information sum except that j check-node equals the message that the it time iteration variable node v passes to j check-node.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA200710307821XA CN101471748A (en) | 2007-12-28 | 2007-12-28 | Method for storing and updating LDPC code message |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA200710307821XA CN101471748A (en) | 2007-12-28 | 2007-12-28 | Method for storing and updating LDPC code message |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101471748A true CN101471748A (en) | 2009-07-01 |
Family
ID=40828895
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA200710307821XA Pending CN101471748A (en) | 2007-12-28 | 2007-12-28 | Method for storing and updating LDPC code message |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101471748A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108039891A (en) * | 2017-12-22 | 2018-05-15 | 山东科技大学 | A kind of polarization code BP interpretation methods and device based on multistage more new technological process |
-
2007
- 2007-12-28 CN CNA200710307821XA patent/CN101471748A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108039891A (en) * | 2017-12-22 | 2018-05-15 | 山东科技大学 | A kind of polarization code BP interpretation methods and device based on multistage more new technological process |
| CN108039891B (en) * | 2017-12-22 | 2021-02-12 | 山东科技大学 | Polarization code BP decoding method and device based on multistage updating process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8151171B2 (en) | Operational parameter adaptable LDPC (low density parity check) decoder | |
| US20100138721A1 (en) | Overlapping sub-matrix based LDPC (Low Density Parity Check) decoder | |
| US20080172592A1 (en) | Method and Device For Controlling the Decoding of a Ldpc Encoded Codeword, In Particular For Dvb-S2 Ldpc Encoded Codewords | |
| US9258015B2 (en) | Decoder with selective iteration scheduling | |
| CN104995844A (en) | Bit flipping decoding with reliability inputs for LDPC codes | |
| US10389388B2 (en) | Efficient LDPC decoding with predefined iteration-dependent scheduling scheme | |
| Lewandowsky et al. | Optimum message mapping LDPC decoders derived from the sum-product algorithm | |
| Liu et al. | Improved decoding algorithms of LDPC codes based on reliability metrics of variable nodes | |
| Chen et al. | Erasure-correction-enhanced iterative decoding for LDPC-RS product codes | |
| US9614548B1 (en) | Systems and methods for hybrid message passing and bit flipping decoding of LDPC codes | |
| US20070113163A1 (en) | Belief propagation decoder cancelling the exchange of unreliable messages | |
| Oh et al. | A two-bit weighted bit-flipping decoding algorithm for LDPC codes | |
| CN100539441C (en) | A kind of interpretation method of low density parity check code | |
| CN101471748A (en) | Method for storing and updating LDPC code message | |
| Tsimbalo et al. | CRC error correction for energy-constrained transmission | |
| Shaker | DVB-S2 LDPC finite-precision decoder | |
| Wang et al. | Partial product-LDPC codes without rate loss | |
| WO2020115070A9 (en) | Iterative decoder for decoding a code composed of at least two constraint nodes | |
| CN210075203U (en) | Decoding device of LDPC code | |
| CN110190856B (en) | Decoding method and decoding device of LDPC code | |
| Zolotarev et al. | Usage of divergence within concatenated multithreshold decoding convolutional codes | |
| US9154261B2 (en) | Low density parity check (LDPC) coding in communication systems | |
| Mathews et al. | Decoding LDPC codes with probabilistic local maximum likelihood bit flipping | |
| Park et al. | Rate allocation for component codes of Plotkin-type UEP codes | |
| Kim | Improved trellis-based decoder for non-binary LDPC codes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20090701 |