CN108833052A - Channel-polarization decoding path metric sort method - Google Patents
Channel-polarization decoding path metric sort method Download PDFInfo
- Publication number
- CN108833052A CN108833052A CN201810388527.4A CN201810388527A CN108833052A CN 108833052 A CN108833052 A CN 108833052A CN 201810388527 A CN201810388527 A CN 201810388527A CN 108833052 A CN108833052 A CN 108833052A
- Authority
- CN
- China
- Prior art keywords
- path
- sequence
- channel
- metric value
- odd
- 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.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
- H04L1/0057—Block codes
-
- H—ELECTRICITY
- H03—ELECTRONIC CIRCUITRY
- H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
- H03M13/00—Coding, 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
- H03M13/03—Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
- H03M13/05—Error 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
- H03M13/13—Linear codes
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0045—Arrangements at the receiver end
- H04L1/0052—Realisations of complexity reduction techniques, e.g. pipelining or use of look-up tables
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
- H04L1/0061—Error detection codes
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Theoretical Computer Science (AREA)
- Error Detection And Correction (AREA)
Abstract
A kind of path metric value sort method of channel-polarization decoding disclosed by the invention.Processing and storage equipment can be greatly reduced using the present invention, significantly decreases decoding delay.The technical scheme is that：When polarization code SCL decoding expands to i-th layer, the L path candidate that input is (i-1)-th layer, father path of the corresponding one group of sequence of each path as i-th layer of Path extension, on L father path, each path carries out Path extension by addition bit 0 and 1, obtains 2L strip path candidate；Then decoder is divided into odd sequence and even sequence by serial number, odd-order is classified as ascending sequence, and even-order is classified as unordered sequence to the path metric value of 2L single sub path.Even sequence is sorted from small to large by odd-even sorting network method, output sequence；Again by length be L odd sequence and length be L sequence carry out merger with odd-even merging device, obtain length be 2L ascending sequence.
Description
Technical field
The present invention relates to a kind of (channel-polarization code Polar) polarization code SCL of codec domain in wireless communication system to translate
Code path competing method, and more particularly, to the interpretation method and decoder of Polar code.
Background technique
Since in practical communication, channel has different degrees of noise jamming, therefore received information will appear difference
The mistake of degree, to reduce the reliability of communication system.So the designers of digital communication system, a core of concern
Heart problem is exactly how information to be enabled reliably to reappear in the channel for having noise jamming as far as possible, to reduce our information
Mistake in transmission, while the efficiency of transmission of information system can not be reduced again, it is just to try to accomplish to take into account the effective of communication system
Property and reliability.Channel-polarization is to convert original channel, and channel capacity changes, at the processed of polarization
Journey.Channel-polarization is initially the upper discovery in B-DMC channel, but channel-polarization phenomenon is generally existing, others letter
Channel-polarization can also be carried out under road, on condition that channel must be symmetrical.Road polarization dependent is different in the channel type used
Channel have different channel-polarization methods.BEC channel has simplest polarization method, and the polarization of BSC channel is more complex,
It is difficult to obtain polarization information by general method.The decoding of polarization code and polarization code performance are closely bound up, and the decoding of use is calculated
Method is different, and the performance of polarization code also has very big difference.2008, Arikan was discussed in international information and is put forward for the first time in ISIT meeting
The concept of channel-polarization, and delivered on IEEE Transaction on Information Theory periodical in 2009
A paper in carried out more detailed elaboration, while a kind of coding mode is given based on channel-polarization, is named as Polar
Code.Polar code is channel-polarization (Channel Polarization) code, is compiled because its low decoding complexity has become current channel
One of the research hotspot in code field.Polar code has had been applied in many technologies from the association schemes of other different codes, such as
Deep space communication, magnetic recording channels and optical transmission system etc..The basis of polarization code is channel-polarization, by channel-polarization it
The good channel of selection channel condition is used to transmit information afterwards, so that the performance of polarization code is greatly improved, but polarization code
There is also some inherent shortcomings, its practical application is influenced.And channel-polarization is to rely on particular channel, that is to say, that not
Same channel, has different channel-polarizations, this just needs to be analyzed according to specific channel situation.If the parameter of channel is
Time-varying, polarization code also wants the moment to change, and system can be made to become complicated.In addition under BEC channel, the calculating ratio of channel-polarization
Other than relatively simple, for other kinds of channel, the calculating of channel-polarization is usually all more complicated.To some more complicated letters
Road can't calculate the various parameters of polarisation channel, limit the use of polarization code on these channels.In addition, polarization code one
As used in situation known to channel condition, if channel condition be it is unknown, it is approximate to can be used prediction channel, approximate
Degree it is poor when, polarization code performance can also become very poor.It can only be previous since the SC decoding of polarization code is that sequence carries out
A symbol translates, and can translate the latter symbol.This has been resulted in when code length is longer, and polarization code has biggish time delay,
It is exactly that handling capacity will receive influence.In polarization code, information be sent into channel before, channel will work, generator matrix with
Channel is closely related, and channel can participate in the calculating of generator matrix, and different generator matrixes is had for different channels,
It will form different code words；In decoding, channel also can influence to decode by some parameters.This is that polarization code and general channel are compiled
The difference of code, it is simply that the coding and decoding of polarization code all relies on channel, it is channel dependent.This is because channel
Polarization is the basis of polarization code, and the polarization of channel depends on the parameter of channel type and channel itself, has ultimately caused polarization
The channel dependent characteristic of code.2.1 channel-polarization polarization codes are a kind of construction codes with low complex degree, when being decoded using SC,
The channel capacity of binary system discrete memoryless channel(DMC) (B-DMC) can be reached.Channel-polarization is the basis of this coding method, channel
Polarization be such a phenomenon：For N number of independent B-DMC channel, input bit by a series of linear transformation it
Afterwards, input channel transmits, and in addition to sub-fraction channel, remaining channel can show channel capacity I (W) and be intended to 0 or 1
Phenomenon.Although channel-polarization is proposed based on binary system discrete memoryless channel(DMC), channel-polarization phenomenon is generally existing
, there is also channel-polarization phenomenons for other channels.The deterministic building method of channel-polarization Polar code, and be
One kind, be also known unique one kind " can be reached " channel coding method of channel capacity by Strict Proof.Polar code is 5G
One of the alternative error correcting coding schemes of ultrahigh speed transmission, however Polar code still has one in actual ultrahigh speed transmission application
A little bottlenecks.Two significant challenges that Polar code faces at present are the promotions for decoding handling capacity and middle short code decoding performance.Polar
When proposing using elimination (Successive cancellation, SC) decoding is connected, decoding complexity is low, but belongs to serial
Decoding, delay is big, and belief propagation (Belief Propagation, BP) decoding is a kind of parallel decoding method, can be improved
Polar code decodes handling capacity, but its performance is not obviously improved compared with SC decoding and decodes computation complexity height.The prior art exists
On the basis of original SC decoding, (Successive Cancellation List is decoded by introducing the SC with list
Decoding, SCL) algorithm, the decoding performance of polarization code obtained great promotion.For traditional SC decoder, work as decoding
When information bit, hard decision directly can be done according to corresponding log-likelihood ratio, and continue to decode using the court verdict.
However SCL decoding algorithm can retain L most possible decoding path during decoding.When decoding information bit, paginal translation
Code path carries out downwards branched extensions, is 0 or be 1, thus obtains 2L alternative path, then therefrom select probability is biggish
L paths.When decoding fixed bit, directly judgement is known bit 0.After the decoding judgement for completing N number of bit, from most
Coding sequence corresponding to the maximum path of select probability in the L paths retained eventually obtains the decoding output of SCL decoder.
Since SCL decoder does not do hard decision immediately but a variety of possible values are remained, hard decision can be reduced and bring mistake
Accidentally a possibility that, to promote final decoding performance.The decoding performance of SCL decoder is related to search broadband L, and L is bigger, property
It can be better.The a key algorithm of SCL decoder is path competition, i.e., is ranked up to the metric of 2L path candidate, from
The middle the smallest L paths of selection path metric value.SCL algorithm can also be combined with CRC check, i.e. SCL (the CRC- of CRC auxiliary
Aided SCL, CA-SCL) decoding algorithm, it include CRC check information in source sequence, it is candidate in the L item that SCL decoding terminates
In path sequence, to improve the error correcting capability of decoding algorithm, it can be obtained by the candidate sequence of CRC check by selection
Other more existing coding modes (Turbo, LDPC) quite, even more preferably performance.The delay series of sequence processing will affect
The decoding delay of SCL realizes that the complexity of logic will affect the clock cycle of maximum processing, final to influence gulping down for Polar code
Spit rate.Tradition sequence Processing Algorithm (counting, insertion, selection, exchange etc.) is mostly serial process, and one is counted with n
Serial sort problem, time delay lower bound are D=Ω (nlog2N), that is to say, that its time delay can not be less than the nlog of constant times2n。
When L is larger, the delay series of serial sort is more, limits the handling capacity of poalr code.
Summary of the invention
The purpose of the present invention is in view of the shortcomings of the prior art place, provide it is a kind of can reduce sequence processing prolong
When series, improve Polar code decoding handling capacity, channel-polarization Polar code SCL decode path metric value sort method.
The purpose of the present invention can be achieved through the following technical solutions：A kind of path metric value sequence of channel-polarization decoding
Method, which is characterized in that include the following steps：
When polarization code SCL decoding expands to i-th layer, the L path candidate for (i-1)-th layer is inputted, each path is corresponding
One group of sequenceuk(1≤k≤i-1) be kth layer decoding bit, one as i-th layer of Path extension
Father path, on L father path, each path carries out Path extension by addition bit 0 and 1, obtains 2L strip path candidate；
Then path metric value o=[o of the decoder to 2L single sub path1,o2,...o2L], Ok(1≤k≤2L) is kth single sub path
Path metric value, path metric value sequence sequentially number are divided into odd sequence [o1,o3,...,o2L-1] and even sequence [o2,o4,...,
o2L], by channel-polarization code polar code SC decoding characteristic, odd-order is classified as ascending sequence, and even-order is classified as unordered sequence.By even-order
Arrange [o2,o4,...,o2L] sorted from small to large by odd-even sorting network method, output sequence e=[e1,e2,...,eL], ek
(1≤k≤L) is k-th of element of ascending sequence e；Odd sequence [the o for being again L by length1,o3,...,o2L-1] and length be L
Sequence e=[e1,e2,...,eL] with odd-even merging device merger is carried out, obtain the ascending sequence [p that length is 2L1,p2,...,
p2L], pk(1≤k≤2L) is k-th of element of ascending sequence；The lesser L path candidate of surviving path metric, and meet
p1≤p2≤...≤pLPath candidate metric [p1,p2,...pL], as i-th layer of path candidate and path metric valuepk (i)(1≤k≤2L) is the path metric value of i-th layer of kth path candidate.
The present invention has the advantages that compared with the prior art.
The present invention inputs the L path candidate for (i-1)-th layer when SCL decoding expands to i-th layer, and each path is corresponding
One group of sequenceAs the father path of i-th layer of Path extension, on L father path, each path is by adding
Bit 0 and 1 is added to carry out Path extension, extension obtains 2L strip path candidate；The strategy of this non-serial sequence is by path
Metric is compared parallel, and available L optimal path metrics can reduce the delay series of sequence processing, make letter
Breath can be transmitted reliably in the channel for having noise jamming.Decoding complexity is relatively low, has more excellent error correction
Can, it has the ability to meet requirement of the 5G mobile communication control channel to error-correcting performance.
The present invention is at decoding end to the path metric value o=[o of 2L single sub path1,o2,...o2L], it is divided into odd-order by serial number
Arrange [o1,o3,...,o2L-1] and even sequence [o2,o4,...,o2L], by even sequence [o2,o4,...,o2L] press odd-even sorting network side
Method is sorted from small to large, output sequence e=[e1,e2,...,eL]；Again by odd sequence [o1,o3,...,o2L-1] and sequence e
=[e1,e2,...,eL] with odd-even merging device carry out merger.Pass through letter corresponding to the maximum path of select probability in the paths
Bit sequence is ceased as decoding as a result, if CRC check, the direct maximum road of output probability value is not satisfied in L alternative path
Information bit sequence corresponding to diameter is as decoding result.The strategy of this non-serial sequence is by carrying out simultaneously path metric value
Row compares, and available L optimal path metrics can reduce the delay series of sequence processing, experimental result shows this
Method can provide about 4 times of acceleration when path selection module is in L=16.Low complexity soft exports decoding algorithm, obtains preferable
Processing and storage equipment are greatly reduced while performance, when not only reducing the use of hardware resource, and reducing decoding
Prolong, it has the ability to meet requirement of the 5th third-generation mobile communication control channel to error-correcting performance and throughput.
Detailed description of the invention
Fig. 1 is the basic flow chart of the path metric value sort method of channel-polarization decoding of the present invention.
Fig. 2 is the route searching example schematic diagram of Polar code SCL decoding algorithm.
Fig. 3 is the graphical representation method of two input comparators.
Fig. 4 is one (2,2) odd-even merging device example.
Fig. 5 is one (4,4) odd-even merging device example.
Fig. 6 is the example that the even sequence of a L=4 is ranked up by odd-even sorting network.
Fig. 7 is the example that the even sequence of a L=8 is ranked up by odd-even sorting network.
The odd sequence and e sequence that Fig. 8 is a L=4 are using the progress merger of odd-even merging device and Optimal Example.
Fig. 9 is the path metric value sequence example of a L=4.
Specific embodiment
Refering to fig. 1.According to the present invention, according to the following steps：
Step 201, when calculating i-th layer of SCL decoding, the L path candidate and metric p of (i-1)-th layer of inputi-1, L is
Natural number.
When SCL decoding expands to i-th layer, the L path candidate for (i-1)-th layer, the corresponding one group of sequence of each path are inputted
ColumnAs the father path of i-th layer of Path extension, path metric value isAnd
pi-1For ascending sequence, meet Indicate (i-1)-th layer of the l articles path candidate metric.
Step 202, on L path candidate, each path carries out Path extension by addition bit 0 and 1, and extension obtains 2L
Strip path candidate.
To a father pathBit 0 and 1 is added respectively, is extended to two new path (u1,
u2,...,ui-1, 0) and (u1,u2,...,ui-1, 1), then L father path, is extended to altogether 2L single sub path.2L single sub path pair
Path metric value o=[the o answered1,o2,...o2L], according to the recurrence formula of Polar code path metric value, calculation formula is such as
Under,
In formula, LlFor the likelihood probability on the basis of the l articles father path sequence, calculating i-th layer of bit, 1≤l≤L.
By formula (1) and (2), andThe corresponding path metric value o of 2L single sub path meets
Following two attributes：
o2l-1≤o2(l+1)-1 (3)
o2l-1≤o2l (4)
In formula, 1≤l≤L.
Fig. 2 shows the route searching examples of Polar code SCL decoding algorithm.
The code length of channel-polarization code Polar is N, the full binary tree that all corresponding depth of Polar code is N.Full binary tree
Each layer of side all respectively corresponds an information bit or fixed bit, in addition to leaf node, after each node is with its left and right two
It is respectively labeled as 0 and 1 after the side between node, decoder is equal for the path of N from root node to any leaf node length
A corresponding coding sequence containing fixed bit.Decoder is formed by path from root node to any one node, corresponding
One path metric value.Decoder is extended path from root node, and each layer works as candidate to when next layer of extension
Path is less than or equal to L, all retains path candidate；When path candidate is greater than L, it is ranked up beta pruning, selects to have in current layer
There is the L item compared with path metric value, L is known as search width.After arriving at leaf node layer, by the Sequential output of metric from small to large
Coding sequence corresponding to L paths, as decoding collection of candidate sequences.
In Fig. 2, the code length N=4, u of polar code1For fixed bit, value 0,For information bit,
Value is (0,1,0), channel-polarization code algorithm search width L=2.
101, decoder is using root node as search starting point, the extension on the 1st layer of side of progress, two path candidates 0 and 1, due to
Number of path is equal to 2, retains, and path metric value is 0.13 and+∞.
102, for decoder in the 2nd layer of extension, 2 path candidates are extended to 4 new paths, respectively (0,0),
(0,1), (1,0) and (1,1), path metric value 0.13,0.61 ,+∞ and+∞.Due to path candidate be greater than 2, from this four
Retain two (0,0) and (0,1) having compared with path metric value, path metric value 0.13,0.61 in path candidate.
103, in the 3rd layer of extension, 2 path candidates are extended to 4 new paths, respectively (0,0,0), (0,0,
1), (0,1,0) and (0,1,1), path metric value 0.92,0.13,1.35 and 0.61.Since path candidate is greater than 2, from
Retain two in this four path candidates and have and compared with (0,0,1) of path metric value and (0,1,1), path metric value is
0.13、0.61。
104, in the 4th layer of extension, 2 path candidates are extended to 4 new paths, respectively (0,0,1,0), (0,
0,1,1) (0,1,1,0) and (0,1,1,1), path metric value 0.13,0.86,1.48 and 0.61.Since path candidate is big
In 2, two (0,0,1,0) and (0,1,1,1) having compared with path metric value, path are retained from this four path candidates
Metric is 0.13,0.61.Finally, SCL decoding algorithm output (0,0,1,0) and (0,1,1,1) is as decoding candidate sequence collection
It closes.
During SCL decoding, to the n-th (n=log since root node2L when) layer extends, path candidate number is less than etc.
In search width L, sequence beta pruning is not needed.Start at (n+1)th layer, the path candidate after extension is 2L, needs to be ranked up and cut
Branch retains L path candidate.Therefore sort algorithm of the invention is directed to (n+1)th layer of later calculating of SCL decoding.
Step 203, to the path metric value of sub- path candidate, it is divided into odd, even two sequences by serial number.
To the path metric value o=[o of 2L single sub path1,o2,...o2L], it is divided into odd sequence [o by serial number1,o3,...,
o2L-1] and even sequence [o2,o4,...,o2L].By formula (1), odd sequence [o1,o3,...,o2L-1] beFor ascending sequence, meetSo odd sequence [o1,
o3,...,o2L-1] meet o1≤o3≤...≤o2L-1, as ascending sequence.
Step 204, dual sequence is sorted from small to large by the method for odd-even sorting network.
Dual sequence [o2,o4,...,o2L] sorted from small to large by the method for odd-even sorting network, odd even ordering net
Network is constructed using odd-even merging device.
Odd-even merging device is that Batcher is proposed at the end of the sixties, is defined as follows：
If there are two the ascending sequence [x to merger1,x2,...,xp] and [y1,y2,...,yq], (p, a q) odd even is returned
And device, p >=q can be as follows with recurrence Construction：If pq=1, odd-even merging device is a comparator shown in Fig. 3；If pq >
1, merger two such odd sequenceWithObtain an ordered sequenceWhereinExpression rounds up；Meanwhile merger two such even sequenceWithObtain an ordered sequenceWhereinIt indicates to be rounded downwards.Finally, will
Following sequential element [v after merging1,w1,v2,w2,...,vq,wq,...,vp] input comparator w two-by-two1:v2w2:v3...wq:
vq+1, obtain final ascending sequence [z1,z2,...,zp+q]。
Refering to Fig. 4, in one (2,2) odd-even merging device, two ascending sequence [x to merger1,x2] and [y1,
y2], p=q=2.Because of pq=4>1, two odd sequence [x of merger1] and [y1], due to [x1] and [y1] pq=1 × 1=1, use
Comparator carries out merger, obtains an ordered sequence [v1,v2]；Two even sequence [x of merger2] and [y2], due to [x2] and [y2]
Pq=1 × 1=1, using comparator carry out merger, obtain an ordered sequence [w1,w2]；Finally, to the sequence after merging
[v1,w1,v2,w2] w1:v2Input comparator obtains final ascending sequence [z1,z2,z3,z4]。
Refering to Fig. 5.Scheme (a).In one (4,4) odd-even merging device, two ascending sequence [x to merger1,x2,x3,
x4] and [y1,y2,y3,y4], p=q=4.Because of pq=16>1, two odd sequence [x of merger1,x3] and [y1,y3], using one (2,
2) odd-even merging device carries out merger, obtains an ordered sequence [v1,v2,v3,v4]；Two even sequence [x of merger2,x4] and
[y2,y4], merger is carried out using one (2,2) odd-even merging device, obtains an ordered sequence [w1,w2,w3,w4]；Finally, right
Sequence [v after merging1,w1,v2,w2,v3,w3,v4,w4] w1:v2w2:v3w3:v4Input comparator obtains final ascending order
Arrange [z1,z2,z3,z4,z5,z6,z7,z8].The comparator configuration of Fig. 5 (a) is depicted as Fig. 5 (b), the two is of equal value.The odd even of even sequence
Sorting network is constructed using multiple odd-even merging devices.
The step 204 is even sequence [o2,o4,...,o2L] it is split as L ascending sequence：[o2], [o4]…[o2L], often
The element number of a sequence is 1.Using L/2 (1,1) merger devices to L/2 to ascending sequence [o2] and [o4], [o6] and
[o8] ..., [o2L-2] and [o2L] merger is carried out respectively, obtain the ascending sequence that L/2 length is 2；L/2 length is 2 to pass
Ascending chain combination of two is L/4 to ascending sequence, carries out merger respectively using L/4 (2,2) merger devices, obtains L/4 length
Merger is finally carried out using the ascending sequence that 1 (L/2, L/2) merger device is L/2 to 2 length for 4 ascending sequence ...,
Obtain the ascending sequence e=[e that length is L1,e2,...,eL]。
The delay series D of step 2041：
Refering to Fig. 6, in the even sequence of a L=4 is ranked up by odd-even sorting network, even sequence [o2,o4,...,
o8] it is split as 4 ascending sequences：[o2], [o4], [o6], [o8], the element number of each sequence is 1.Using 2 (1,1) merger
Device is to 2 couples of ascending sequence [o2] and [o4], [o6] and [o8] merger is carried out respectively, the ascending sequence that 2 length are 2 is obtained, then
Merger is carried out using the ascending sequence that 1 (2,2) merger device is 2 to 2 length again, obtains the ascending sequence e=that length is 4
[e1,e2,e3,e4]。
Refering to Fig. 7.It by the even sequence of a L=8, is ranked up by odd-even sorting network, even sequence [o2,o4,...,
o16] it is split as 8 ascending sequences：[o2], [o4]…[o16], the element number of each sequence is 1.
Using 4 (1,1) merger devices to 4 couples of ascending sequence [o2] and [o4], [o6] and [o8], [o10] and [o12], [o14]
[o16] merger is carried out respectively, obtain the ascending sequence that 4 length are 2；The ascending sequence combination of two for being again 2 by 4 length
For 2 pairs of ascending sequences, merger is then carried out respectively using 2 (2,2) merger devices, obtains the ascending sequence that 2 length are 4；Again
Merger is carried out using the ascending sequence that 1 (4,4) merger device is 4 to 2 length, obtains the ascending sequence e=[e that length is 81,
e2,...,e8]。
Step 205, ascending sequence e=[e odd sequence and step 204 exported1,e2,...,eL] use odd-even merging device
Merger is carried out, by ascending sequence [o1,o3,...,o2L-1] and sequence e=[e1,e2,...,eL], it is carried out using (L, L) merger device
Merger obtains the ascending sequence [p that length is 2L1,p2,...,p2L]；By formula (3), in ascending sequence [o1,o3,...,o2L-1]
In first of element o2l-1, then there are L-l+1 element (m containing element in the sequence2l-1) value be greater than or equal to o2l-1, by public affairs
Formula (4), this L-l+1 element can be less than or equal to even sequence [o2,o4,...,o2L] corresponding L-l+1 element, that is,
It says, even sequence [o2,o4,...,o2L] at least L-l+1 element value be greater than or equal to o2l-1.Sequence e=[e1,
e2,...,eL] it is even sequence [o2,o4,...,o2L] ascending sequence after sequence, therefore, sequence e=[e1,e2,...,eL] at least
There is the value of L-l+1 element to be greater than or equal to o2l-1, i.e.,
o2l-1≤el (5)
Optionally, the odd-even merging device in step 205 can be further simplified.
The first order of odd-even merging device in step 205 completes o2l-1And elComparison, l=1,2 ..., L.By formula
(5), this grade of comparator can remove.For 2L path metric value after merger, L before SCL decoder only retains, rear L road
The sequence of diameter metric is unrelated to SCL decoding, therefore can omit the only related comparator with rear L path metric value sequence.
Optionally, the odd-even merging device in step 205 can be further simplified.
The comparators for not influencing preceding L sequence all in (L, L) odd-even merging device are removed.
The delay series D of step 2052：
D2=log2L
Fig. 8 (a) is the odd sequence [o of a L=41,o3,o5,o7] and e=[e1,e2,e3,e4] using odd-even merging device into
The example of row merger, Fig. 8 (b) are its Optimal Examples.
Fig. 8 (a) is by ascending sequence [o1,o3,o5,o7] and sequence e=[e1,e2,e3,e4], using (4,4) odd-even merging device
Merger is carried out, the ascending sequence [p that length is 8 is obtained1,p2,...,p8]；
Fig. 8 (b) removes first order comparator, and does not influence the comparator of preceding 4 sequences in (4,4) odd-even merging device
Remove.
Step 206, retain i-th layer of L path candidate and path metric value.
Retain the lesser L path candidate of path metric value and path candidate metric [p in step 2051,p2,
...pL], meet p1≤p2≤...≤pL, as i-th layer of path candidate and path metric valueIt is full
Foot
The sequence delay of the method for the invention is in step 204 and 205, because of total sequence delay D of the invention the method：
The comparison sheet of serial sort time delay
| Search for broadband L | 2 | 4 | 8 | 16 | 32 | 64 | 128 |
| The present invention | 2 | 8 | 24 | 64 | 160 | 384 | 896 |
| Serial sort | 2 | 5 | 9 | 14 | 20 | 27 | 35 |
Total sequence delay and the lower bound (minimum value) of serial algorithm time delay are carried out according to the comparison sheet of serial sort time delay
Compare, with the increase of search broadband L, serial algorithm time delay is quicklyd increase, and the time delay advantage of the method for the invention is more invented
It is aobvious.
Below with reference to specific example, the sequencer procedure of the more detailed description embodiment of the present invention.
Refering to Fig. 9.In path metric value sequence in the case where L=4,
The first step, using (i-1)-th layer of 4 path candidates as father path, path metric value
Every father path is carried out subpath extension by addition bit 0 and 1, obtains 8 strip path candidates by second step.Road
Diameter metric is calculated by formula (1) and (2), obtains 8 strip path candidate metric o=[o1,o2,o3,o4,o5,o6,o7,
o8]。
Third step, path metric value o=[o1,o2,o3,o4,o5,o6,o7,o8], it is divided into odd sequence [o by serial number1,o3,o5,
o7], even sequence [o2,o4,o6,o8].Wherein odd sequence [o1,o3,o5,o7] beMeet o1≤o3≤
o5≤o7。
4th step, even sequence [o2,o4,o6,o8] be ranked up by odd-even sorting network, export ascending sequence e=[e1,e2,
e3,e4]。
5th step, odd sequence [o1,o3,o5,o7] and ascending sequence e=[e1,e2,e3,e4] carried out using (4,4) merger device
Merger, wherein the first order comparator of odd-even merging device removes and all comparators for not influencing preceding 4 sequences remove.
6th step, lesser 4 path candidates of surviving path metric and path candidate metric [p1,p2,p3,p4], it is full
Sufficient p1≤p2≤p3≤p4, as i-th layer of path candidate and path metric valueMeetAlways sequence delay D=5 in the case where L=4.
The embodiment of the present invention has been described in detail above, and specific embodiment used herein carries out the present invention
It illustrates, method of the invention that the above embodiments are only used to help understand；Meanwhile for the general technology of this field
Personnel, according to the thought of the present invention, there will be changes in the specific implementation manner and application range, in conclusion this theory
Bright book content should not be construed as limiting the invention.
Claims (10)
1. a kind of path metric value sort method of channel-polarization decoding, which is characterized in that include the following steps：
When polarization code SCL decoding expands to i-th layer, the L path candidate for (i-1)-th layer is inputted, each path is one group corresponding
Sequenceuk(1≤k≤i-1) is the decoding bit of kth layer, a father as i-th layer of Path extension
Path, on L father path, each path carries out Path extension by addition bit 0 and 1, obtains 2L strip path candidate；Then
Path metric value o=[o of the decoder to 2L single sub path1,o2,...o2L], Ok(1≤k≤2L) is the path of kth single sub path
Metric, path metric value sequence sequentially number are divided into odd sequence [o1,o3,...,o2L-1] and even sequence [o2,o4,...,o2L], by
Channel-polarization code polar code SC decoding characteristic, odd-order are classified as ascending sequence, and even-order is classified as unordered sequence.By even sequence [o2,
o4,...,o2L] sorted from small to large by odd-even sorting network method, output sequence e=[e1,e2,...,eL], ek(1≤k
≤ L) be ascending sequence e k-th of element；Odd sequence [the o for being again L by length1,o3,...,o2L-1] and length be L sequence e
=[e1,e2,...,eL] with odd-even merging device merger is carried out, obtain the ascending sequence [p that length is 2L1,p2,...,p2L], pk(1
≤ k≤2L) be ascending sequence k-th of element；The lesser L path candidate of surviving path metric, and meet p1≤p2
≤...≤pLPath candidate metric [p1,p2,...pL], as i-th layer of path candidate and path metric valuepk (i)(1≤k≤2L) is the path metric value of i-th layer of kth path candidate.
2. the path metric value sort method of channel-polarization decoding as described in claim 1, which is characterized in that i-th layer of path
The father path metric of extension isAnd pi-1For ascending sequence, meet Indicate (i-1)-th layer of the l articles path candidate metric.
3. the path metric value sort method of channel-polarization decoding as described in claim 1, which is characterized in that a road Tiao Fu
DiameterBit 0 and 1 is added respectively, is extended to two new path (u1,u2,...,ui-1, 0) and (u1,
u2,...,ui-1, 1), then L father path, is extended to altogether 2L single sub path, and L is natural number.
4. the path metric value sort method of channel-polarization decoding as claimed in claim 3, which is characterized in that 2L single sub path
Corresponding path metric value o=[o1,o2,...o2L], the recurrence formula according to channel-polarization code Polar code path metric value
In formula, LlFor the likelihood probability on the basis of the l articles father path sequence, calculating i-th layer of bit, and 1≤l≤L.
5. the path metric value sort method of channel-polarization decoding as described in claim 1, which is characterized in that channel-polarization code
The code length of Polar is N, the full binary tree that all corresponding depth of Polar code is N.Each layer of side of full binary tree all respectively corresponds
One information bit or fixed bit, the side difference in addition to leaf node, between each node and its left and right two descendant node
It is marked as 0 and 1, decoder corresponds to one from root node to any leaf node length for the path of N and contains fixed bit
Coding sequence.
6. the path metric value sort method of channel-polarization as described in claim 1 decoding, which is characterized in that decoder is from root
Node is formed by path to any one node, corresponds to a path metric value.
7. the path metric value sort method of channel-polarization as described in claim 1 decoding, which is characterized in that decoder is from root
Node sets out, and is extended to path, and each layer is to when next layer of extension, and when path candidate is less than or equal to L, all reservation is waited
Routing diameter；When path candidate is greater than L, it is ranked up beta pruning, selects the L item for having compared with path metric value in current layer, L is known as
Search width.After arriving at leaf node layer, by coding sequence corresponding to the Sequential output L paths of metric from small to large, make
To decode collection of candidate sequences.
8. the path metric value sort method of channel-polarization decoding as described in claim 1, which is characterized in that decoder is with root
Node is search starting point, carries out the extension on the 1st layer of side, two path candidates 0 and 1；Decoder is in the 2nd layer of extension, 2 candidates
Path is extended to 4 new paths, respectively (0,0), (0,1), (1,0) and (1,1), path metric value 0.13,
0.61 ,+∞ and+∞；In the 3rd layer of extension, 2 path candidates are extended to 4 new paths, respectively (0,0,0), (0,
0,1), (0,1,0) and (0,1,1), path metric value 0.92,0.13,1.35 and 0.61；In the 4th layer of extension, 2 times
Routing diameter is extended to 4 new paths, respectively (0,0,1,0), (0,0,1,1) (0,1,1,0) and (0,1,1,1), road
Diameter metric is 0.13,0.86,1.48 and 0.61.
9. the path metric value sort method of channel-polarization decoding as described in claim 1, which is characterized in that antithetical phrase candidate road
The path metric value of diameter is divided into odd, even two sequences by serial number；To the path metric value o=[o of 2L single sub path1,o2,
...o2L], it is divided into odd sequence [o by serial number1,o3,...,o2L-1] and even sequence [o2,o4,...,o2L]；Odd sequence [o1,o3,...,
o2L-1] beFor ascending sequence, meet
10. the path metric value sort method of channel-polarization decoding as described in claim 1, which is characterized in that, dual sequence
It is sorted from small to large by the method for odd-even sorting network；Dual sequence [o2,o4,...,o2L] by the side of odd-even sorting network
Method is sorted from small to large, and odd-even sorting network is constructed using odd-even merging device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810388527.4A CN108833052B (en) | 2018-04-26 | 2018-04-26 | Channel polarization decoding path metric value sorting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810388527.4A CN108833052B (en) | 2018-04-26 | 2018-04-26 | Channel polarization decoding path metric value sorting method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108833052A true CN108833052A (en) | 2018-11-16 |
| CN108833052B CN108833052B (en) | 2021-02-26 |
Family
ID=64155706
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810388527.4A Active CN108833052B (en) | 2018-04-26 | 2018-04-26 | Channel polarization decoding path metric value sorting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108833052B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109818627A (en) * | 2019-01-21 | 2019-05-28 | 中国地质大学（武汉） | A kind of the simple gate limit pruning method and system of Polar code SCL decoding algorithm |
| CN110808740A (en) * | 2019-11-01 | 2020-02-18 | 北京航空航天大学 | Low-complexity decoding method based on polarization code under abridged channel |
| CN111130566A (en) * | 2019-12-18 | 2020-05-08 | 清华大学 | Circuit implementation method for searching L maximum path metric values in Polar code decoder |
| CN112861145A (en) * | 2021-01-06 | 2021-05-28 | 华控清交信息科技(北京)有限公司 | Data processing method and device and data processing device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016168962A1 (en) * | 2015-04-20 | 2016-10-27 | 华为技术有限公司 | Decoding method and decoding apparatus for polar code |
| CN106209113A (en) * | 2016-07-29 | 2016-12-07 | 中国石油大学(华东) | A kind of decoding method of polarization code |
| CN107636973A (en) * | 2015-05-31 | 2018-01-26 | 华为技术有限公司 | Path merging method, device and the code translator of polarization code |
-
2018
- 2018-04-26 CN CN201810388527.4A patent/CN108833052B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016168962A1 (en) * | 2015-04-20 | 2016-10-27 | 华为技术有限公司 | Decoding method and decoding apparatus for polar code |
| CN107636973A (en) * | 2015-05-31 | 2018-01-26 | 华为技术有限公司 | Path merging method, device and the code translator of polarization code |
| CN106209113A (en) * | 2016-07-29 | 2016-12-07 | 中国石油大学(华东) | A kind of decoding method of polarization code |
Non-Patent Citations (1)
| Title |
|---|
| 魏一鸣: "极化码性能研究及其SCL译码算法的FPGA实现", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109818627A (en) * | 2019-01-21 | 2019-05-28 | 中国地质大学（武汉） | A kind of the simple gate limit pruning method and system of Polar code SCL decoding algorithm |
| CN109818627B (en) * | 2019-01-21 | 2020-08-07 | 中国地质大学（武汉） | Single-threshold pruning method and system for Polar code SC L decoding algorithm |
| CN110808740A (en) * | 2019-11-01 | 2020-02-18 | 北京航空航天大学 | Low-complexity decoding method based on polarization code under abridged channel |
| CN110808740B (en) * | 2019-11-01 | 2021-08-10 | 北京航空航天大学 | Low-complexity decoding method based on polarization code under abridged channel |
| CN111130566A (en) * | 2019-12-18 | 2020-05-08 | 清华大学 | Circuit implementation method for searching L maximum path metric values in Polar code decoder |
| CN111130566B (en) * | 2019-12-18 | 2021-05-11 | 清华大学 | Circuit implementation method for searching L maximum path metric values in Polar code decoder |
| CN112861145A (en) * | 2021-01-06 | 2021-05-28 | 华控清交信息科技(北京)有限公司 | Data processing method and device and data processing device |
| CN112861145B (en) * | 2021-01-06 | 2023-12-12 | 华控清交信息科技(北京)有限公司 | Data processing method and device for data processing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108833052B (en) | 2021-02-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108833052A (en) | Channel-polarization decoding path metric sort method | |
| Niu et al. | Stack decoding of polar codes | |
| CN107846261A (en) | Decoded and the continuous apparatus and method for cancelling list-decoding for parallel continuous cancel | |
| CN109842418B (en) | Polarization code belief propagation decoding method based on bit flipping | |
| US20060090121A1 (en) | Trellis-based receiver | |
| CN107017892B (en) | A kind of verification cascade polarization code encoding method and system | |
| CN108462558A (en) | A kind of polarization code SCL interpretation methods, device and electronic equipment | |
| CN105978577A (en) | Serial list decoding algorithm based on bit flipping | |
| CN109347487A (en) | Freeze the polarization code SCL interpretation method of auxiliary based on bit | |
| US20070266303A1 (en) | Viterbi decoding apparatus and techniques | |
| CN107896137A (en) | A kind of sort method suitable for the division of polarization code decoding path | |
| CN104025459A (en) | Decoding processing method and decoder | |
| CN109936377A (en) | A kind of the polarization code coding and interpretation method of segmentation CRC auxiliary | |
| CN109547034B (en) | Decoding method and device, decoder | |
| CN110138390A (en) | A kind of polarization code SSCL algorithm decoder based on deep learning | |
| JP2001237716A (en) | Method and device for estimating sequence | |
| CN110492974A (en) | A kind of parallel polarization code coding method and device | |
| Song et al. | Efficient successive cancellation stack decoder for polar codes | |
| CN110661533A (en) | Method for optimizing decoding performance of decoder for storing polarization code | |
| Song et al. | Low-complexity segmented CRC-aided SC stack decoder for polar codes | |
| Rezaei et al. | Implementation of ultra-fast polar decoders | |
| CN104821863A (en) | Forward stack decoding method for rateless Spinal code | |
| Goyal et al. | Sequence reconstruction problem for deletion channels: A complete asymptotic solution | |
| CN110765730A (en) | Method for simplifying combined logic circuit | |
| CN110324111A (en) | A kind of interpretation method and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |