CN103124181A - Turbo code decoding iteration cease method based on cosine similarity - Google Patents
Turbo code decoding iteration cease method based on cosine similarity Download PDFInfo
- Publication number
- CN103124181A CN103124181A CN2013100227699A CN201310022769A CN103124181A CN 103124181 A CN103124181 A CN 103124181A CN 2013100227699 A CN2013100227699 A CN 2013100227699A CN 201310022769 A CN201310022769 A CN 201310022769A CN 103124181 A CN103124181 A CN 103124181A
- Authority
- CN
- China
- Prior art keywords
- iteration
- cosine similarity
- external information
- component decoder
- decoder
- 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
Abstract
The invention discloses a turbo code decoding iteration cease method based on cosine similarity. The method includes: (1) performing once iteration decoding between two component decoders, (2) storing external information sequence for decoding block, which is output by a component decoder II and defined as external information vector, (3) calculating cosine similarity between external information vectors of current iteration and last iteration, (4) comparing the cosine similarity calculated through step (3) with a preset threshold, entering step (5) if the cosine similarity is more than the threshold or the current number of iterations reaches a preset maximum and repeating step (1)-(4) if not and (5) performing de-interleaving and hard decision for log likelihood ratio which is generated by the iteration component decoder II the last time and outputting the log likelihood ratio as decoding result of the decoder. According to the method, iteration decoding speed is increased significantly during Turbo code pattern decoding of LTE (Long Term Evolution) / LET-A (Long Term Evolution-Advanced) system. A certain value is for designing high speed Turbo decoder to satisfy requirements of the LTE / LET-A system.
Description
Technical field
The present invention relates to the LTE/LTE-A technology, relate to Turbo code high-speed coding technology, be specifically related to a kind of Turbo code decoding iteration stopping method based on the cosine similarity.
Background technology
Turbo code is a kind of channel coding technology of superior performance.Turbo code exchanges the raising that external information obtains performance between two component decoders by mutual iterative process, iterations is more, and decoding performance is better, but complexity is higher simultaneously, delays time larger.Modern Communication System is more and more higher to transmission rate request, how effectively to reduce the decoding delay of Turbo code and how balance decoding performance and time-delay are all very important research topics.
Chnnel coding in 3GPP Long Term Evolution LTE system and LTE-A system has all adopted Turbo code.Require descending peak rate 1Gbit/s in LTE-A system standard, up peak rate 500Mbit/s.Requirement for the high transfer rate that meets the LTE-A 1Gbit/s of system need to improve traditional turbo coder, and design and checking meet the novel decoding algorithm framework of high transfer rate requirement in future broadband wireless communication systems.The iteration stopping strategy is one of key technology that improves decoding speed, can in the situation that the bit error rate performance loss is less significantly reduce the Turbo decode procedure in mean iterative number of time, improve decoding speed.
The error bit ability of Turbo code is along with the carrying out of iteration constantly reduces, but after certain iterations, the performance of turbo code no longer improves along with the carrying out of iteration, and continue iteration this moment only can increase Time Delay of Systems in vain.And for some data sequences, just can realize zero defect decoding through iterations seldom; Also have some data sequences, wrong too many due to what exist, carry out all error correction fully of how many times iteration.Therefore, using conventional methods is all sequences to be decoded all to be arranged identical fixed number of times there is no need, and can cause system resource and waste of time; And dynamically determine can effectively reduce the decoding iterations by the decoding iterations of each data sequence under the little condition of Turbo code performance impact according to certain iteration stopping strategy, improve the average decoding speed of Turbo decoder.
Mainly contain the complexity of decoding speed (mean iterative number of time of the every frame of decoding), error performance (error rate and frame error ratio), stopping criterion etc. for the standard of weighing different iteration stopping methods.The iteration stopping method requires in the situation that the error performance loss is less significantly reduce the Turbo decode procedure in mean iterative number of time, improve decoding speed, simultaneously the complexity of algorithm taken in.
Auxiliary (HDA) criterion of the iteration stopping method that some are commonly used such as hard decision, auxiliary criterion (IHDA) criterion of improved hard decision, different ratio (SDR) criterion of symbol complexity are lower.The HDA criterion stops iteration when the hard decision symbol of the soft information of double iteration output of component decoder II no longer changes; The IHDA criterion is improved the HDA criterion, and purpose is the storage that reduces last iterative information, thereby reduces memory requirement; The SDR criterion determines whether to continue iteration by prior information bit number and the thresholding different with the external information symbol of a relatively more same iteration component decoder I or component decoder II.As tolerance, mean iterative number of time and desirable criterion gap are larger based on the symbol of soft information in iterative process for these several criterions.The Cyclic Redundancy Check criterion is carried out CRC check to the hard decision result, and result is 0 and stops iteration; CRC criterion speed is higher, but complexity is higher.
Summary of the invention
The purpose of this invention is to provide a kind of Turbo code decoding iteration stopping method based on the cosine similarity, in the situation that the bit error rate performance loss is less, significantly reduce mean iterative number of time in the Turbo decode procedure, further improve decoding speed.
To achieve these goals, technical scheme of the present invention is:
A kind of Turbo code decoding iteration stopping method based on the cosine similarity comprises the steps: to carry out iterative decoding one time between the component decoder I of (1) Turbo decoder and component decoder II; (2) external information of each bit of the piecemeal to be decoded of component decoder II output in storing step (1), first bit to the external information sequence of last bit in the piecemeal to be decoded of component decoder II output is regarded as a vector, be called the external information vector, the mould that calculates and store this external information vector is long; The cosine similarity of the external information vector that (3) obtains since the 2nd iteration calculation procedure (2) and the external information vector of last iteration; (4) step (3) is calculated the cosine similarity of gained and the default corresponding thresholding of current iteration number of times and compared, if this cosine similarity reaches default maximum iteration time greater than thresholding or current iteration number of times, enter step (5); Otherwise, repeating step (1), (2), (3), (4); (5) log-likelihood ratio that component decoder II in last iteration is generated carries out deinterleaving and hard decision, and this hard decision is exported as the final decode results of Turbo code decoder.
Step of the present invention (1) concrete grammar is: (11) before iteration, the prior information of component decoder I is initialized as 0 for the first time; (12) check bit of information bits, component decoder I and prior information are input to the component decoder I and carry out MAP decoding, generate external information and the log-likelihood ratio of component decoder I; The external information of the component decoder I after (13) process QPP interweaves is as the prior information of component decoder II, be input to the component decoder II with check bit through the information bits that interweaves and component decoder II and carry out MAP decoding, generate external information and the log-likelihood ratio of component decoder II; (14) through the external information of the component decoder II after the QPP deinterleaving, will use as the new prior information of component decoder I.
External information vector in step of the present invention (2) during the n time iteration is the K n dimensional vector n that in this piecemeal of component decoder II output, first bit to the external information sequence of last bit forms, and K is the block sort size.
The arithmetic square root of the quadratic sum of the external information of all bits in long this piecemeal exported for the component decoder II of the mould of the external information vector in step of the present invention (2) during the n time iteration;
In step of the present invention (3), the cosine similarity of the external information vector of the n time iteration of employing and the n-1 time iteration is as tolerance, and wherein the cosine similarity of two vectors is defined as:
In formula,
With
Be two vectors that dimension is identical,
With
Be respectively vector
With
Mould long, A
iAnd B
iBe respectively
With
I component.
When calculating the cosine similarity in step of the present invention (3), the order of magnitude of external information absolute value is 10-100, for excessive generation of the median of avoiding computational process overflowed, can be with all external informations divided by 100 rear participation computings, the cosine similarity size that finally obtains is constant;
One group of threshold value corresponding with the current iteration frequency n in step of the present invention (4), namely the thresholding during the n time iteration is θ
n, n=2,3 ..., N
max-1; N wherein
maxBe default maximum iteration time.θ
nScope be θ
n∈ [0.85,1], and θ
n-1≤ θ
nθ
nConcrete size need to be determined according to emulation; θ corresponding to different iterationses
nIdentical value can be got, also two or more sets different values can be made as by the height of iterations.
In another embodiment of the present invention, the cosine similarity of gained and corresponding thresholding θ will be calculated in step (3)
nCompare, if stop iterative process greater than thresholding, namely the iteration stopping condition is:
cosα＞θ
n
Wherein, α is the angle of two external information vectors, and iterations n satisfies 2≤n≤N
max-1.
In another embodiment of the present invention, default maximum iteration time N
max, when iterations reaches N
maxWhether time satisfies the iteration stopping condition and all stops iterative process.
In another embodiment of the present invention, the information sequence in the decode procedures such as external information, prior information or log-likelihood ratio that two dope vectors in the cosine similarity are component decoder or it is made information sequence after conversion;
Compared with prior art, the present invention as tolerance, has further improved decoding speed with the cosine similarity between the external information vector in iterative decoding process.The present invention in the situation that bit error rate performance loss is less further reduce the Turbo decode procedure in mean iterative number of time, improve decoding speed.
Description of drawings
Fig. 1 is the Turbo code decoding iteration stopping method flow diagram that the present invention is based on the cosine similarity.
Fig. 2 is the decoding architecture figure of the Turbo code decoding iteration stopping method based on the cosine similarity shown in Figure 1.
Embodiment
By following description also by reference to the accompanying drawings, it is more clear that the present invention will become, and these accompanying drawings are used for explaining embodiments of the invention.
With reference now to accompanying drawing, describe embodiments of the invention, in accompanying drawing, similar element numbers represents similar element.
The present invention is based on the Turbo code decoding iteration stopping method of cosine similarity, the Turbo code type in LTE/LTE-A of being based on realizes; Channel type is additive white Gaussian noise channel (AWGN); Modulation system adopts binary phase shift keying (BPSK); Decoding algorithm adopts MAP decoding; The Turbo code generator polynomial is (13,15); Code check 1/3.
The below illustrates the present embodiment based on the flow process of the Turbo code decoding iteration stopping method of cosine similarity.In conjunction with Fig. 1, Fig. 2, said method comprising the steps of:
Step S1, the prior information of initialization component decoder I is 0, the cosine similarity is 0;
Step S2, relatively and judge the n time and the n-1 time iteration external information vector between the cosine similarity whether greater than the default thresholding θ of correspondence
n, or whether iterations n is greater than maximum iteration time N
max2≤n≤N
max-1, if having at least a condition to set up, go to step S8; Otherwise, enter next step;
Step S3, iterations n increases by 1;
Step S4, after the external information process QPP deinterleaving of component decoder II last iteration, prior information (prior information of iteration real component decoder I is initialization value 0 first) as the component decoder I, be input to the component decoder I and carry out MAP decoding together with the check bit of information bits, component decoder I, generate external information and the log-likelihood ratio of component decoder I;
Step S5, the external information of the component decoder I after interweaving through QPP is as the prior information of component decoder II, be input to the component decoder II with check bit through the information bits that interweaves and component decoder II and carry out MAP decoding, generate external information and the log-likelihood ratio of component decoder II;
Step S6, the external information vector of storage component decoder II output, the mould that calculates and store the external information vector is long, the external information vector is the K n dimensional vector n that in this piecemeal of component decoder II output, first bit to the external information sequence of last bit forms, K is the block sort size, the long arithmetic square root for the quadratic sum of the external information of all bits in this piecemeal of component decoder II output of the mould of external information vector;
Step S7 calculates the cosine similarity of the external information vector of the n time iteration and the n-1 time iteration; The cosine similarity of two vectors is defined as:
In formula,
With
Be two vectors that dimension is identical,
With
Be respectively vector
With
Mould long, A
iAnd B
iBe respectively
With
I component.Be back to step S2;
Step S8 stops iterative process, and the last log-likelihood ratio that generates of component decoder II is carried out deinterleaving and hard decision, obtains final decode results; Finish.
As seen from the above, the present embodiment measures with the cosine similarity of the external information of two component decoders in iterative decoding process the ability that two decoders continue to improve by the exchange of external information decoding performance based on the Turbo code decoding iteration stopping method of cosine similarity, thereby judgement is enough to stop iterative process, to improve decoding speed.The present invention in the situation that bit error rate performance loss is very little further reduce the Turbo decode procedure in mean iterative number of time, improve decoding speed.
Above invention has been described in conjunction with most preferred embodiment, but the present invention is not limited to the embodiment of above announcement, and should contain various modification, equivalent combinations of carrying out according to essence of the present invention.
Claims (8)
1. Turbo code decoding iteration stopping method based on the cosine similarity is characterized in that comprising the steps: carrying out iterative decoding one time between the component decoder I of (1) Turbo decoder and component decoder II; (2) external information of each bit of the piecemeal to be decoded of component decoder II output in storing step (1), first bit to the external information sequence of last bit in the piecemeal to be decoded of component decoder II output is regarded as a vector, be called the external information vector, the mould that calculates and store this external information vector is long; The cosine similarity of the external information vector that (3) obtains since the 2nd iteration calculation procedure (2) and the external information vector of last iteration; (4) step (3) is calculated the cosine similarity of gained and the default corresponding thresholding of current iteration number of times and compared, if this cosine similarity reaches default maximum iteration time greater than thresholding or current iteration number of times, enter step (5); Otherwise, repeating step (1), (2), (3), (4); (5) log-likelihood ratio that component decoder II in last iteration is generated carries out deinterleaving and hard decision, and this hard decision is exported as the final decode results of Turbo code decoder.
2. the Turbo code decoding iteration stopping method based on the cosine similarity according to claim 1, it is characterized in that: described step (1) concrete grammar is: (11) before iteration, the prior information of component decoder I is initialized as 0 for the first time; (12) check bit of information bits, component decoder I and prior information are input to the component decoder I and carry out MAP decoding, generate external information and the log-likelihood ratio of component decoder I; The external information of the component decoder I after (13) process QPP interweaves is as the prior information of component decoder II, be input to the component decoder II with check bit through the information bits that interweaves and component decoder II and carry out MAP decoding, generate external information and the log-likelihood ratio of component decoder II; (14) through the external information of the component decoder II after the QPP deinterleaving, will use as the new prior information of component decoder I.
3. the Turbo code decoding iteration stopping method based on the cosine similarity according to claim 1, it is characterized in that: the external information vector in described step (2) during the n time iteration is the K n dimensional vector n that in this piecemeal of component decoder II output, first bit to the external information sequence of last bit forms, and K is the block sort size.
4. the Turbo code decoding iteration stopping method based on the cosine similarity according to claim 3 is characterized in that: the arithmetic square root of the quadratic sum of the external information of all bits in long this piecemeal exported for the component decoder II of the mould of the external information vector in described step (2) during the n time iteration.
5. the Turbo code decoding iteration stopping method based on the cosine similarity according to claim 1, it is characterized in that: in described step (3), the cosine similarity of the external information vector of the n time iteration of employing and the n-1 time iteration is as tolerance, and wherein the cosine similarity of two vectors is defined as:
In formula,
With
Be two vectors that dimension is identical,
With
Be respectively vector
With
Mould long, A
iAnd B
iBe respectively
With
I component.
6. the Turbo code decoding iteration stopping method based on the cosine similarity according to claim 5, it is characterized in that: when calculating the cosine similarity in described step (3), the order of magnitude of external information absolute value is 10-100, for excessive generation of the median of avoiding computational process overflowed, divided by 100 rear participation computings, the cosine similarity size that finally obtains is constant with all external informations.
7. the Turbo code decoding iteration stopping method based on the cosine similarity according to claim 1 is characterized in that: the information sequence in the decode procedures such as external information, prior information or log-likelihood ratio that two dope vectors in described cosine similarity are component decoder or it is made information sequence after conversion.
8. the Turbo code decoding iteration stopping method based on the cosine similarity according to claim 1 is characterized in that: one group of threshold value corresponding with the current iteration frequency n in described step (4), namely the thresholding during the n time iteration is θ
n, n=2,3 ..., N
max-1; N wherein
maxBe default maximum iteration time; θ
nScope be θ
n∈ [0.85,1], and θ
n-1≤ θ
n
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013100227699A CN103124181A (en) | 2013-01-22 | 2013-01-22 | Turbo code decoding iteration cease method based on cosine similarity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013100227699A CN103124181A (en) | 2013-01-22 | 2013-01-22 | Turbo code decoding iteration cease method based on cosine similarity |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103124181A true CN103124181A (en) | 2013-05-29 |
Family
ID=48455059
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2013100227699A Pending CN103124181A (en) | 2013-01-22 | 2013-01-22 | Turbo code decoding iteration cease method based on cosine similarity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103124181A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104579369A (en) * | 2014-12-18 | 2015-04-29 | 中国科学院自动化研究所 | Turbo iterative decoding method and device |
| CN106533454A (en) * | 2015-09-14 | 2017-03-22 | 展讯通信（上海）有限公司 | Decoding and iteration control method and device for Turbo codes |
| CN108650643A (en) * | 2018-05-10 | 2018-10-12 | 福建科立讯通信有限公司 | A kind of effective method for reducing the signaling response time in DMR/MPT terminal devices |
| CN112217525A (en) * | 2020-10-15 | 2021-01-12 | 天津津航计算技术研究所 | Automatic updating method for iterative times of Turbo decoding |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020184596A1 (en) * | 2001-04-09 | 2002-12-05 | Motorola, Inc. | Iteration terminating using quality index criteria of turbo codes |
| US20070300139A1 (en) * | 2006-06-09 | 2007-12-27 | Samsung Electronics Co., Ltd. | Unified stopping criteria for binary and duobinary turbo decoding in a software-defined radio system |
| EP2501048A1 (en) * | 2011-03-14 | 2012-09-19 | Commissariat à l'Énergie Atomique et aux Énergies Alternatives | State metrics based stopping criterion for turbodecoding |
| US20120272125A1 (en) * | 2011-04-19 | 2012-10-25 | Lin Shou-Sheu | Stopping Methods for Iterative Signal Processing |
| CN102832954A (en) * | 2012-09-17 | 2012-12-19 | 华中科技大学 | Turbo code iterative decoding stopping method based on soft information average minimum value |
-
2013
- 2013-01-22 CN CN2013100227699A patent/CN103124181A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020184596A1 (en) * | 2001-04-09 | 2002-12-05 | Motorola, Inc. | Iteration terminating using quality index criteria of turbo codes |
| US20070300139A1 (en) * | 2006-06-09 | 2007-12-27 | Samsung Electronics Co., Ltd. | Unified stopping criteria for binary and duobinary turbo decoding in a software-defined radio system |
| EP2501048A1 (en) * | 2011-03-14 | 2012-09-19 | Commissariat à l'Énergie Atomique et aux Énergies Alternatives | State metrics based stopping criterion for turbodecoding |
| US20120272125A1 (en) * | 2011-04-19 | 2012-10-25 | Lin Shou-Sheu | Stopping Methods for Iterative Signal Processing |
| CN102832954A (en) * | 2012-09-17 | 2012-12-19 | 华中科技大学 | Turbo code iterative decoding stopping method based on soft information average minimum value |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104579369A (en) * | 2014-12-18 | 2015-04-29 | 中国科学院自动化研究所 | Turbo iterative decoding method and device |
| CN104579369B (en) * | 2014-12-18 | 2018-06-15 | 北京思朗科技有限责任公司 | A kind of Turbo iterative decodings method and code translator |
| CN106533454A (en) * | 2015-09-14 | 2017-03-22 | 展讯通信（上海）有限公司 | Decoding and iteration control method and device for Turbo codes |
| CN106533454B (en) * | 2015-09-14 | 2019-11-05 | 展讯通信（上海）有限公司 | Turbo code decoding iteration control method and device |
| CN108650643A (en) * | 2018-05-10 | 2018-10-12 | 福建科立讯通信有限公司 | A kind of effective method for reducing the signaling response time in DMR/MPT terminal devices |
| CN108650643B (en) * | 2018-05-10 | 2021-06-15 | 福建科立讯通信有限公司 | Method for effectively reducing signaling response time in DMR/MPT terminal equipment |
| CN112217525A (en) * | 2020-10-15 | 2021-01-12 | 天津津航计算技术研究所 | Automatic updating method for iterative times of Turbo decoding |
| CN112217525B (en) * | 2020-10-15 | 2023-03-03 | 天津津航计算技术研究所 | Automatic updating method for iterative times of Turbo decoding |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106849960B (en) | Segmentation CRC check storehouse interpretation method and framework based on polarization code | |
| CN108847850A (en) | A kind of segmentation polarization code coding/decoding method based on CRC-SSCL | |
| CN107689801B (en) | The early method of shutting down of LDPC code ADMM iterative decoding | |
| CN108429599B (en) | Method and apparatus for data processing in a communication system | |
| CN106330207A (en) | Joint detection and decoding algorithm based on Turbo-SCMA system | |
| CN107612560A (en) | Polarization code earlier iterations method of shutting down based on partial information bit log likelihood ratio | |
| Zhang et al. | Efficient adaptive list successive cancellation decoder for polar codes | |
| CN103124181A (en) | Turbo code decoding iteration cease method based on cosine similarity | |
| CN102832954B (en) | Turbo code iterative decoding stopping method based on soft information average minimum value | |
| EP2717477B1 (en) | Channel decoding method and decoder for tail-biting codes | |
| Miao et al. | A low complexity multiuser detection scheme with dynamic factor graph for uplink SCMA systems | |
| CN111726202B (en) | Early termination iteration method for polarization code belief propagation decoding | |
| Song et al. | Low-complexity segmented CRC-aided SC stack decoder for polar codes | |
| CN103812510A (en) | Decoding method and device | |
| WO2023116504A1 (en) | Double-factor correction turbo decoding method based on simulated annealing algorithm | |
| CN101969308A (en) | Method and device for decoding tail-biting convolutional codes | |
| CN103701475B (en) | Decoding method for Turbo codes with word length of eight bits in mobile communication system | |
| Yuan et al. | Double-Parity-Check CA-SCL encoding and decoding for polar codes | |
| Tajima et al. | Iterative decoding based on concatenated belief propagation for CRC-aided polar codes | |
| CN114421976A (en) | TPC iterative decoding method and decoder based on probability calculation | |
| Reddy et al. | A low complexity stopping criterion for reducing power consumption in turbo decoders | |
| CN109474283B (en) | Polar code decoding method with early termination and confidence enhancement and architecture thereof | |
| CN111835363A (en) | LDPC code decoding method based on alternative direction multiplier method | |
| Xia et al. | A two-staged adaptive successive cancellation list decoding for polar codes | |
| Han et al. | Simplified multi-bit SC list decoding for polar 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 |
Application publication date: 20130529 |