CN106603084B - It is a kind of for punching the preprocess method of LDPC Hard decision decoding - Google Patents

Abstract

The present invention provides a kind of for punching the preprocess method of LDPC Hard decision decoding, firstly, being determined according to coding punching algorithm to the punching check information position after coding；Then, the position generator matrix G correlation unit is found according to punch position；The information received is recycled, does precoding processing, the check information that completion punching is lost according to LDPC encryption algorithm；Finally, the reception information of completion is done LDPC work decoding by hard decision algorithm, reach code correction performance.The present invention is few to the consumption of hardware resource, and is easy to Project Realization；The case where for punching, can provide the decoding gain of about 0.5dB, and especially in the situation that punch position is more, channel condition is poor, error-correcting performance is more significant.

Description

It is a kind of for punching the preprocess method of LDPC Hard decision decoding

Technical field

The invention belongs to field of channel coding, are related to a kind of for punching the preprocess method of LDPC Hard decision decoding.

Background technique

Low density parity check code (LDPC) shows the error correction so far closest to shannon limit in a communications system Performance, have been widely used for optic communication, satellite communication, deep space communication, forth generation mobile communication, at a high speed with very high speediness rate number In the systems such as user's line and magnetic recording.But compared with other error correcting codes, the encoding and decoding of LDPC code usually require a large amount of hardware Resource consumption and processing time, this just greatly limits its practical application.In recent years, the striving direction of research is concentrated mainly on drop Low encoding and decoding complexity using Embedded micro-system and improves speed of service etc. of electronic system, however level of hardware Limitation is so that numerous studies personnel are more concerned about improvement or substitution to existing algorithm.Hardware is reduced by improvement or alternate algorithm The complexity of realization is wise and reasonable selection, and improved algorithm can reduce the requirement to hardware to a certain extent, It reduces cost to reach, be easy to practical purpose.

In LDPC decoding, it is broadly divided into Hard decision decoding and Soft decision decoding two major classes, Hard decision decoding does not need to appoint Decoding can be realized in what Soft Inform ation, and structure is simple, and operand is smaller, is easy to Project Realization, but compared with Soft decision decoding error-correcting performance About low 2~3dB.Soft decision decoding has good performance, but resources requirement is big, and implementation complexity is higher.

Carrying out punching processing to LDPC code is a kind of method for realizing variable bit rate LDPC, and thought is verified by deleting Some information bits in information, to achieve the purpose that realize adjustment code rate and code length.In practical applications, it is frequently encountered tune It the case where rate to match communication of whole code length or code rate, is mainly decoded at present using the method for Soft decision decoding.So And under conditions of code length is longer, hardware resource is nervous, the realization difficulty of Soft decision decoding it is abnormal greatly even not possible with, if It using hard decision LDPC interpretation method, decodes error correction gain and just has biggish loss, especially punching information is more, channel Hard decision LDPC decoding can even lose decoding error correcting capability in the case that condition is poor.

Summary of the invention

For overcome the deficiencies in the prior art, the present invention provides a kind of for punching the pretreatment side of LDPC Hard decision decoding Method, in conjunction with generator matrix, carries out precoding processing to the information received, then decoded, finally according to punching information position Obtain decoding result.

The technical solution adopted by the present invention to solve the technical problems the following steps are included:

(1) according to punching algorithm is encoded, position K1, K2 ... the Kn of check information are punched after confirmation coding, from reception information In find out encoded information C, i.e. information bit；

(2) it is converted to obtain generator matrix G according to check matrix H, be found out in generator matrix G corresponding with punching information Ki Ki column Gi, wherein corresponding relationship refers to that the Ki of the Ki information bit and generator matrix G arrange corresponding relationship each other；

(3) encoded information unit C, mould 2 in the corresponding reception code word of element non-zero in all generator matrix Ki column are asked With check information Pki after obtaining precoding；It is done using encoded information C and recompiles processing partially to restore perforation Check information；

(4) step (2), (3) are repeated, each value of i is traversed, finds out all punch position check information recovery values respectively Pki；

(5) the Ki punching check information in information is received come completion with the Pki found out respectively, recovers complete code word S'；

(6) by S ' as new Receiving coded information, decoding information is solved with LDPC Hard decision decoding algorithm.

The beneficial effects of the present invention are:

(1) preprocess method proposed by the present invention, docking, which collects mail to cease to have done, before decoding partially recompiles processing, extensive Override interrupt partial information, provides more reliable informations for decoding processing, directly carries out doing for Hard decision decoding compared to previous Method can have the decoding gain of about 0.5dB, and especially when punching, channel condition are more harsh, effect is more obvious；

(2) preprocess method proposed by the present invention only recompiles processing to punching information relevant portion, due to The H-matrix and generator matrix G of LDPC code are all sparse matrixes, i.e., most of unit is 0 in matrix, and a small amount of unit is 1, according to square Battle array multiplication rule, need to only take the information of non-zero unit in G correlation unit, simplify operation, only need to increase a small amount of hardware resource, It is few to the consumption of hardware resource many compared to the way of application Soft decision decoding algorithm, more meet practical application and Cheng Benfang The considerations of face.

Detailed description of the invention

Fig. 1 is the process flow diagram that the present invention carries out preprocess method；

Fig. 2 is preprocess method application process figure of the present invention；

Fig. 3 is the carry out pretreatment energy simulation comparison figure in specific embodiment.

Specific embodiment

Present invention will be further explained below with reference to the attached drawings and examples, and the present invention includes but are not limited to following implementations Example.

The technical solution adopted by the present invention to solve the technical problems be according to punching information position, in conjunction with generator matrix, Precoding processing is carried out to the information that receives, then is decoded, finally obtain decoding as a result, specifically includes the following steps:

(1) it initializes

Algorithm is punched according to coding, position K1, K2 ... the Kn of check information are punched after confirmation coding, is looked for from receiving in information Encoded information C out, i.e. information bit.

(2) the generator matrix G unit answered is determined

It is converted to obtain generator matrix G according to check matrix H, finds out in generator matrix G corresponding with punching information Ki Ki column Gi, wherein corresponding relationship refers to that the Ki of the Ki information bit and generator matrix G arrange corresponding relationship each other.

(3) precoding processing

Encoded information unit C, mould 2 in the corresponding reception code word of " 0 " element non-in all generator matrix Ki column are summed, Check information Pki after obtaining precoding.According to coding principle S=C × G, wherein S is code word after coding, and C is to receive in information Encoded information, G is generator matrix.It due to receiving code word S is spliced by encoded information C and check information J, using connecing Encoded information C in code word is received, part is done and recompiles processing to restore the check information of perforation, i.e. precoding processing.

(4) (2) are repeated, (3) step obtains all recovery value Pki

I is accumulated to n from 1, (2), (3) step is repeated, finds out all punch position check information recovery values respectively Pki。

(5) completion punch position check information restores code word S '

It is recovered complete code word S ' come the Ki punching check information in completion code word with the Pki found out respectively and connect Collection of letters breath+punching information recovery value.

(6) LDPC Hard decision decoding

By S ' as new received encoded information, then LDPC Hard decision decoding algorithm is used, according to Hard decision decoding Step solves decoding information.

For (1250, the 2500) quasi-cyclic LDPC code for being below 1/2 with code rate, to the tool of preprocess method of the invention Body embodiment is illustrated.Since (1250,2500) LDPC code is not the more regular code of a code length, in real system In, traffic rate 62.5kbps, in order to match traffic rate, has done at punching LDPC along with the CRC check after coding Reason, carrying out decoding processing using pretreatment of the invention in this case, specific step is as follows:

The originally determined punch position of step 1 and encoded information

The last 150bits of check bit has been done at punching in order to meet the requirement of rate-matched after system completes coding Reason is that a unit carries out with 50bits since (1250,2500) LDPC code is formed with the cyclic code of 50bits*50bits Processing, we select the 23rd, 24,25 perforation unit, and the data grouping to receiving, and every 50bit is one group, and first 25 groups are For receive encoded information C1, C2 ... C25.

Step 2 determines the corresponding generator matrix G unit group of 3 groups of punchings

The corresponding G matrix of the H-matrix is calculated by Matlab software, finds the 23rd corresponding G matrix of perforation unit Unit G23.Under normal circumstances, needing to find out generator matrix G just can be carried out subsequent step, but corresponding some special LDPC codes For, G matrix is readily available, such as quasi-cyclic LDPC (QC-LDPC) code, lower triangular structure LDPC code, double diagonal arrangements LDPC code, especially some LDPC codes etc. for directly constructing G matrix.It, can be with reference encoder part in the code that G matrix is difficult to obtain The G matrix that coding generates is borrowed, many time and efforts can be saved.

Step 3 does precoding processing to the 23rd punching relevant information, the value that is restored P23.

According to coding principle S=C × G, wherein S is code word after coding, and C is the encoded information received in information, and G is to generate Matrix.It due to receiving code word s is spliced by encoded information C and check information J, utilizes encoded information in reception code word s C does part and recompiles processing to restore the check information of perforation, i.e. precoding processing.

In G matrix, in the 23rd column matrix unit, the only the 5th, 11,17,25 matrix unit is non-zero unit, is followed Ring shift count is respectively: 17,18,5,34.Encoded information unit C5, C11, C17, C25 for receiving in code word are made respectively 17,18,5,34 circulative shift operations to the right do 2 He of mould to encoded information C ' 5, C ' 11, C ' 17, the C ' 25 after displacement by column Operation has obtained the recovery value P23 of the 23rd perforation unit.

Step 4 is restored value P24, P25 according to (2), (3) step respectively

The 24th, 25 column G matrix unit G24, G25 are taken respectively according to step (2), wherein in G24 the 6th, 11,20,21 to be non- Zero location, cyclic shift number is respectively: 25,13,1,9；Wherein in G25 the 7th, 12,18,25 be non-zero unit, circulation Shift count is respectively: 28,29,43,14.According to step (3), mould 2 and place are done after shifting respectively for corresponding encoded information Reason, obtains recovery value P24, P25 of the 24th, 25 single hole unit.

Step 5 restores 2500bits code word S '

The code word for receiving dissection process shares 2350bits, and P23, P24, P25 that step (4) is obtained are mended to reception code word Below, i.e., the position of the the 23rd, 24,25 perforation unit, recovers the code word S ' of 2500bits.

Step 6 weighted bit overturns the decoding of hard decision algorithm

Application weighting bit reversal Hard decision decoding algorithm, maximum number of iterations are selected as 20 times, are done to the code word S ' of recovery Decoding processing obtains decoding result C '.

For the quasi- circulation in (1250,2500) for being 1/2 with code rate punches LDPC code, using the pretreatment side of invention Method, the decoding gain performance that software emulation obtains are analyzed compared with Hard decision decoding such as Fig. 3, it can be seen that the preprocess method Compared to Hard decision decoding method, lower its error-correcting performance of signal-to-noise ratio (Eb/N0) improve it is more, with the raising of signal-to-noise ratio, The raising of decoding gain is becoming smaller, when signal-to-noise ratio (Eb/N0) is lower than 5.5dB, which can improve about 0.5~ The decoding gain of 1.0dB.

It is directed to this code simultaneously, this preprocess method is applied in concrete practice, in the Cyclone of altera corp It is realized on III EP3C120F84I7FPGA, Soft decision decoding, the resource occupation pair for pre-processing Hard decision decoding, Hard decision decoding Table 1 for example than situation, wherein Soft decision decoding information bit uses 6 quantification treatments, and Hard decision decoding is hard using weighted bit overturning Adjudicate decoding algorithm.It can be seen that being about the 1/ of soft-decision decoding method demand using hardware resource needed for the preprocess method 3, compared with untreated Hard decision decoding, it need to only increase the hardware resource less than 1%.

1 three kinds of processing method resource occupation contrast tables of table

Processing method	Logic elements	Registers	Memory bits
				Pre-process hard decision	7078 (6%)	3846	14721 (< 1%)
Soft decision decoding	23596 (20%)	16091	142771 (4%)
				Hard decision decoding	5803 (5%)	3611	10625 (< 1%)

Claims (1)

1. a kind of for punching the preprocess method of LDPC Hard decision decoding, it is characterised in that include the following steps:

(1) algorithm is punched according to coding, position K1, K2 ... Ki ... Kn, 1≤i≤n of check information is punched after confirmation coding, Encoded information C, i.e. information bit are found out from receiving in information；

(2) it is converted to obtain generator matrix G according to check matrix H, finds out generation square corresponding with the punching position Ki of check information Ki column Gi in battle array G, wherein corresponding relationship refers to that the Ki information bit arranges corresponding each other close with the Ki of generator matrix G System；

(3) encoded information unit C, mould 2 in the corresponding reception code word of element non-zero in all generator matrix Ki column are summed, is obtained Check information Pki after to precoding；It is done using encoded information C and recompiles processing partially to restore the verification of perforation and believe Breath；

(4) step (2), (3) are repeated, each value of i is traversed, finds out all punch position check information recovery value Pki respectively；

(5) the Ki punching check information in information is received come completion with the Pki found out respectively, recovers complete code word S '；

(6) by S ' as new Receiving coded information, decoding information is solved with LDPC Hard decision decoding algorithm.