CN109217984B

CN109217984B - Efficient blind detection decoding method and decoder for polarization code

Info

Publication number: CN109217984B
Application number: CN201811144879.1A
Authority: CN
Inventors: 张川; 任雨青; 束峰; 尤肖虎
Original assignee: Southeast University
Current assignee: Southeast University
Priority date: 2018-09-29
Filing date: 2018-09-29
Publication date: 2020-11-20
Anticipated expiration: 2038-09-29
Also published as: CN109217984A

Abstract

The invention discloses an efficient blind detection decoding method and a decoder for a polarization code. The invention realizes the simplified blind detection of the polarization code, can avoid the receiver from carrying out a complex decoding algorithm on all polarization code system candidates, and reduces the power consumption, the complexity and the delay as much as possible. In the simulation of simulating a real transmission environment, the error rate of a decoded frame can be in E on the basis of the missing code length N by simplifying a blind detection algorithm_b/N₀4dB, and reaches 0.021.

Description

Efficient blind detection decoding method and decoder for polarization code

Technical Field

The invention relates to a polarization code blind detection decoding method and a decoder, belonging to the technical field of channel coding.

Background

With the rapid development of modern communication technology, polarization codes have been regarded as a great breakthrough in coding theory since the beginning of the invention. As a known error correction code that has proven to fully reach the capacity of binary discrete memoryless channels, polar codes are chosen by 3GPP as a control channel code to enhance the mobile broadband application scenario. In previous mobile communications, blind detection has been a solution for certain circumstances. The schemes used in LTE rely on concatenation of CRC and convolutional codes, as well as blind detection algorithms for other codes, such as BCH codes and LDPC codes. However, since the control information specified in 5G is encoded by the polarization code, the problem of blind detection in the context of internet of things in the research of the polarization code becomes particularly prominent.

In the literature (Giard P, Ballsoukas-Stimming A, Burg A. Blank detection of polar codes [ C ]// IEEE International work on Signal Processing systems. IEEE,2017:1-6), a means for blind detection of a polar code is proposed, which distinguishes a polar code from a non-polar code encoded Signal by SSC decoding. By blind detection it is possible to avoid complex decoders running modern error correction codes for all candidates, i.e. preferably eliminating most candidates earlier, to minimize complexity, delay power consumption. However, at present, there is no better solution for blind detection and decoding of polarization codes of different systems, the problem of blind detection and decoding of frame information of multiple polarization code systems in a channel is solved, and great significance is provided for improving decoding efficiency and reducing decoding power consumption.

Disclosure of Invention

The purpose of the invention is as follows: the invention aims to provide an efficient blind detection decoding method and a decoder for a polar code, which mainly solve the problem of blind detection decoding of the polar code in a scene of enhancing mobile bandwidth.

The technical scheme is as follows: in order to achieve the purpose, the invention adopts the following technical scheme:

an efficient blind detection decoding method for a polarization code comprises the following steps:

(1) sending the received polarization code to the front end of a blind detection decoder for blind detection decoding, wherein the blind detection decoding adopts an SSC decoder structure to realize the decoding of a plurality of candidate codes with different code lengths and the same code rate in a polarization code candidate code set;

(2) to frozen code node N in the blind detection decoding process⁰The bit number is accumulated to obtain a measurement D, the likelihood ratio is mapped before accumulation, and the difference of the change of the receiving sequence is enlarged to eliminate the influence of the decoder;

(3) and detecting a threshold value of the measurement D, identifying the polarization code signal according to a set threshold value interval, and sending the polarization code signal to a correct decoder for decoding.

In a specific scheme, the polar code received in step (1) is a polar code with a known code rate and an unknown code length.

In the specific scheme, different codes in a polarization code candidate code set are in a multiple relation, the front end of a blind detection decoder is designed according to the maximum polarization code length in the polarization code candidate code set, the folding characteristic of a polarization code decoder is utilized, and a decoder with a larger code length is formed by combining decoders with a smaller code length; the front end of the blind detection decoder inputs a frame sequence with the maximum code length in a group of candidate code sets or polarization code frame sequences with other code lengths in a plurality of groups of candidate code sets in parallel.

In a specific scheme, in the step (2), the likelihood ratio in the frozen bit is mapped to be 1 when the likelihood ratio is greater than 0, and mapped to be-1 when the likelihood ratio is less than 0.

In a specific scheme, the threshold detection formula in the step (3) is as follows:

wherein d is₁...d_nFor a threshold interval determined according to simulation statistics, H₁...H_nFor each hypothesis of the candidate for a polar code, H is the polar code determined from the metric D threshold.

The invention provides a high-efficiency blind detection decoder of a polarization code, which comprises a blind detection decoder front end, a mapping unit, a judging unit and at least two decoders for respectively decoding the polarization codes with different coding modes;

the front end of the blind detection decoder adopts an SSC decoder structure to decode a plurality of candidate codes with different code lengths and the same code rate in a polarization code candidate code set, inputs the received polarization code and outputs a frozen code node N⁰A likelihood ratio of (d);

the mapping unit is used for mapping the likelihood ratio and expanding the difference of the change of the receiving sequence so as to eliminate the influence of the decoder;

the decision unit is used for determining the node N of the mapped frozen code⁰And accumulating the likelihood ratios to obtain a metric D, carrying out threshold detection on the metric D, identifying the polarization code signal according to a set threshold interval, and sending the polarization code signal into a correct decoder for decoding.

Has the advantages that: compared with the prior art, the invention can complete the blind detection and decoding of the polarization code of the information transmitted in the channel under the condition of missing part of prior information, can avoid the receiver from carrying out complex decoding operation on all polarization code candidates, and reduces the power consumption, complexity and delay as much as possible. The method can well distinguish the polarization code signal from a non-polarization code signal (such as an all-zero signal or a random signal) and the polarization code signals with different coding modes by simulating a real channel transmission environment according to a simulation result. When the conventional channel transmission is simulated, under the condition that the code length of the polarization code transmitted in the channel is unknown by a receiver, the simplified blind detection method designed by the invention can reach the frame error rate of 0.021 under the condition that the signal-to-noise ratio is equal to 4 dB.

Drawings

Fig. 1 is a block diagram of a decoding method for blind detection of polarization codes.

Fig. 2 is a block diagram of a polar code blind detection decoder.

Fig. 3 is a diagram of the channel capacity after channel polarization with N-1024.

Fig. 4 is a graph of the metric D probability distribution function using a blind detection decoding algorithm for (128,16), (256,32), (512,64) polar codes and all-zero and random signals.

Fig. 5 is a graph of the metric D probability density function using a blind detection decoding algorithm for (128,16), (256,32), (512,64) polar codes and all-zero and random signals.

Fig. 6 is an FER performance diagram of the blind detection decoding method for polar codes according to the present invention.

Detailed Description

The invention is further described with reference to the following figures and specific embodiments.

The invention is based on the algorithm of the continuous serial elimination decoding (SC decoding) and the simplified continuous serial elimination decoding (SSC decoding) of the polarization code, and reasonably processes and accumulates the likelihood ratio of the special position generated in the process.

As described in the literature (Arikan E.channel Polarization: A Method for structuring Capacity-Achieving Codes for symmetry Binary-Input memory Channels [ J ]. IEEE Transactions on Information Theory,2009,55(7):3051-3073), the processing formula for the parity-split sub-Channels in the SC coding algorithm is as follows:

wherein the L function is a log-likelihood ratio decision formula derived on the basis of a channel transition probability formula, y is a received sequence,

to estimate the bits, N denotes the code length, i denotes the sequence number of the sequence, and o and e denote odd and even numbers, respectively.

SSC(Alamdar-Yazdi A,Kschischang F R.A Simplified Successive-Cancellation Decoder for Polar Codes[J]IEEE Communications Letters,2011,15(12):1378-⁰And information code node N¹According to N⁰And N¹The regular pruning of the code tree can reduce the operation times and the waiting time, and the pruning of the huge full binary code tree has great significance on the blind detection and decoding of the polarization code.

Blind detection of the polarization code is actually the setting by the receiver of a metric D by some correlation in the decoding of the received signal in the absence of all or part of the a priori information encoded by the received signal. This metric is then threshold detected, thereby enabling blind detection of the polarization code. The efficient blind detection decoding framework of the polar code provided by the embodiment of the invention is shown in fig. 1, polar code signals of a plurality of different coding modes exist in a channel, and under the condition of losing part of prior information of the polar code signals, the polar code signals can be sent to a corresponding decoder for decoding after the measurement D judgment of the polar code blind detection decoder designed by the invention.

Verification is performed with reference to a parity check matrix which is actually used for blind detection of LDPC. But it is impractical to recover a large parity check matrix without any a priori information, i.e. totally blind, so for a transceiver that requires a pre-given set of modulation and coding, the candidate set is known to both the receiver and the transmitter. The blind detection method of the polarization code designed by the invention also utilizes the thought, and the receiver stores the prior information of the candidate code set of the polarization code in advance, such as the code length, the code rate and the like of each candidate. In the decoding process, similar to the Path Metric value processing of SCL, the Metric D is accumulated, and the correct encoding format is found from the candidate set by the value of D, so that the correct decoding is performed.

Because if the received information is to be blindly detected, it is necessary to adopt the method for receiving the information

The metric D is sufficiently sensitive to change. Since as the code length N increases, the decoder risks underflow because the transition probability is smaller and smaller. Therefore, before accumulating the previous metric D, a one-step mapping relationship is added:

wherein w₀And w₁The probability formula is taking values of 0,1 for channel transition.

Because the absolute value of LLR is greatly influenced by the structure of the decoder, the influence of the structure of the decoder is eliminated through the mapping relation, and the gravity center of blind detection is attributed to the received information code element.

Furthermore, the LLRs for the information bits are discarded when calculating the accumulation, since the information bits do not have information redundancy, and only the LLRs for the frozen bits are retained for the accumulated distribution of metric D in the blind detection, i.e., N in SSC decoding⁰Likelihood ratio information of the node is retained, N¹The node likelihood ratio information of (1) will be discarded.

1) Code rate-0 code: rate-0 codes that consist entirely of frozen bits are not true codes, i.e., they are previously known to be all-zero vectors. In noiseless transmission, the LLR for a rate 0 node should consist of all positive LLRs. Therefore, we introduce D the following rule for updating:

wherein A is^cTo freeze a set of bit positions, N_stopIs a manually controlled cutoff bit.

2) Code rate-1 code: by definition, rate-1 codes do not contain any frozen bits, i.e. no redundancy is added to the information. This makes rate-1 codes useless for detection purposes, so they are ignored in the calculation of the detection metric.

Based on the above analysis, the efficient blind detection decoding method for the polarization code disclosed by the embodiment of the invention mainly comprises the following implementation steps:

step (1): the candidate information of the polarization code is uniformly sent to blind detection decoding, and for the blind detection decoder designed by the embodiment of the invention, the code length information N of the candidate signal of the polarization code is unknown.

Step (2): the likelihood ratio accumulation is carried out on a specific frozen digit in the decoding process, and the frozen code node N of SSC decoding is mainly utilized⁰And information code node N¹There is a property of information redundancy, and the distribution of frozen bits is different for polar code candidate signals of different coding formats, so the accumulated values are also different.

And (3): reasonably mapping the likelihood ratio to L_symMapping the likelihood ratio L larger than 0 into 1 and mapping the likelihood ratio L smaller than 0 into-1, expanding the difference of the received sequence change, thereby eliminating the influence of the decoder itself and leading the blind detection to be returned to the information sequence

Itself.

And (4): the set metric D is accumulated and thus threshold detection is performed. For the polar code signals falling into the corresponding threshold interval, blind detection decoding of the polar code frame information in the channel can be realized, and the polar code frame information is sent to a correct decoder for decoding.

Fig. 3 is the channel capacity distribution after N1024 channel polarization. From the above, it is known that for N1024, the blind detection metric D should continuously increase in the forward direction during the calculation of channels 0-300, so that D exceeds a certain value D_AWe can then estimate the current code length to be N-1024. If the likelihood ratio LLRs continue to be accumulated, the amplification of the metric D is slowed down or even stopped due to the start of the intervention of a large number of information bits. After a large number of simulations, for the polar codes with different code lengths, D of the polar codes will fall into different intervals, and threshold detection is performed, and the formula is as follows:

wherein d is₁...d_nIs a threshold interval, the value of which can be determined from simulation statistics, H₁...H_nAre hypotheses for respective polarization code candidates.

In combination with the above idea, we will propose a decoder structure for a blind detection algorithm for polar codes. Because the algorithm mainly focuses on the node likelihood ratio study on the code rate-0 and involves the mapping relation L_symAnd the accumulation of the metric D, the decoding structure is improved on the basis of the original SSC decoding, and the mapping and decision gating processes are added. Specifically, the efficient blind detection decoder for the polarization codes disclosed by the embodiment of the invention mainly comprises a blind detection decoder front end, a mapping unit, a judging unit and at least two decoders for respectively decoding the polarization codes with different coding modes. Wherein, the decoder front end is used for carrying out blind detection decoding on the received candidate information of the polarization code and outputting a frozen code node N⁰The front end of the decoder adopts an SSC decoder structure to realize the decoding of a plurality of candidate codes with different code lengths and the same code rate in a polarization code candidate code set; a mapping unit for mapping the likelihood ratio and expanding the difference of the received sequence variation to eliminate the influence of the decoder; a decision unit for mapping the node N of the frozen code⁰And accumulating the likelihood ratios to obtain a metric D, carrying out threshold detection on the metric D, identifying the polarization code signal according to a set threshold interval, and sending the polarization code signal into a correct decoder for decoding.

Different codes in the polarization code candidate code set are in a multiple relation, the front end of the blind detection decoder is designed according to the maximum polarization code length in the polarization code candidate code set, the folding characteristic of the polarization code decoder is utilized, and the decoder with the larger code length is formed by combining the decoders with the smaller code length; the front end of the blind detection decoder inputs a frame sequence with the maximum code length in a group of candidate code sets or polarization code frame sequences with other code lengths in a plurality of groups of candidate code sets in parallel.

As shown in fig. 2, blind detection is performed for N-128, N-256, and N-512 polarization codes. When the polar code format information of the received signal is unknown, the decoder firstly assumes that the received information is a polar code with the length of N being 512 codes, and then the polar code is sent to the front end of the decoder to be decoded. In this way, for a polar code with a code length N256 and N128, the input of the decoder is used

Respectively composed of four groups of frame sequences

With two sets of frame sequences

And (4) forming. And then, carrying out measurement D accumulation according to the code rate and the digit distribution of the N-512 polarization code stored in the receiver, wherein the final decision device plays a gating role, and gates a polarization code decoder with a corresponding correct code length at the rear end according to an output result. The folding characteristic of the polar code decoder is utilized, and the structure of the blind detection decoder is greatly simplified.

In the 3GPP RAN1 conference, as shown in the standard, a maximum code length N of 512 and a low-rate (generally R of 1/8) polar code are often discussed and used, so the polar code rate participating in the test is set to 1/8. The simulation scene is that three polarization codes with different code lengths are simultaneously decoded through blind detectionThe coders, receivers implement blind detection of the code length N they do not know, knowing their code rate R. Let the polarization code of the three formats be N-512, N-256, N-128, and R-1/8. The structure of the decoder is substantially as shown in figure 2. Transmission scenario is E_b/N₀2dB AWGN channel and BPSK modulation.

For simulation results, the present invention considers the following transmission scheme:

1) no transmission (NoTx): this is to simulate the situation where there is no data transmission on the channel, i.e. there is only white gaussian noise in the channel.

2) Random transmission (RndTx): this is a scenario where random data is transmitted over a channel. It simulates the situation where the channel is in use, but contains data that does not have the structure inherent to the polar code to be detected, i.e. the transmission data is not encoded in the polar code.

3) Regular transmission (RegTx): the scenario simulates the transmission of frames encoded using polar codes over a channel. This indicates that the channel contains a polar code block, and the polar code frame should be detected for decoding by a corresponding decoder.

As shown in fig. 4, there are five signals in the figure, and the signals are respectively, from left to right in the display order: random signal, polar code (512,64), polar code (256,32), polar code (128,16) and all-zero signal, and fig. 5 is a metric D probability density function. For frozen node N, as shown in

equations

1 and 2⁰When decoding, will be paired

The mapping is done to accumulate the metric D.

The polarization code images of (512,64), (256,32), (128,16) in fig. 4 were analyzed. The accumulation of the measurement D from small to large is carried out according to the descending order of the code length N, and the three can be well separated in the dimension of the measurement D. According to the original idea, if the polar codes with three code lengths pass through the (512,64) decoder, when the first 384 bits of the (512,64) decoder are accumulated, the polar codes (512,64) should theoretically take more L with frozen bit bits_symThereby accumulatingThe accumulation speed should be much faster than the code length N256 and N128 of the polar code. When the code length N of the polar code is 256 or 128, a large number of information bits may be flooded in the corresponding number of bits when decoding is performed by the (512,64) decoder. Because the information is not controllable for the information bits, the D value is made to be slow or even reversed, but the simulation result is just contrary to theory. This phenomenon occurs because the channel capacity for transmitting the frozen bits is close to 0 after channel polarization, and the channel quality is very poor. This allows us to transmit 0 bits in the frozen bit, if not according to A^cThe prior information is directly judged, most of the prior information is decoded into 1 bit at a decoding end, so that the phenomenon that the measurement D is increased along with the decreasing of the code length N just occurs.

The random signal and the all-zero signal in fig. 4 were analyzed. The random signal refers to a random signal transmitted at a transmitting end in a code format without any polarization code. An all-zero signal refers to sending an all-0 signal at the transmitting end. It can be seen that the metric D of the random signal is substantially distributed around 0, which corresponds to the characteristics of the random signal and is well distinguished from the three polar code signals with the coding format. The all-zero signals are distributed on the rightmost side of the image, and the accumulation speed of the metric D is the fastest, and the metric D also accords with the performance of the metric D.

With signal-to-noise ratio E_b/N₀The metric D discrimination between different signals shows an improvement to different extents. It also shows an increasing tendency to the right, which is more pronounced for all-zero signals. Therefore, on the basis of a large number of simulations, in combination with equation 3, three signal metric D threshold regions are specified as (60,190), (191,264), (265,330), and a decision is made on the actual blind detection D. And on this basis, a simulation of the frame error rate FER was thus performed, as shown in fig. 6. It can be seen from the FER result of fig. 6 that the blind detection decoding method for the polar code proposed by the present invention can be implemented in E_b/N₀With 4dB, FER performance of 0.021 is achieved.

Claims

1. An efficient blind detection decoding method for a polarization code is characterized in that: the method comprises the following steps:

(2) to frozen code node N in the blind detection decoding process⁰Accumulating the likelihood ratio to obtain a metric D, mapping the likelihood ratio before accumulation, and expanding the difference of the change of the receiving sequence to eliminate the influence of the decoder;

2. The efficient blind detection decoding method for polar codes according to claim 1, wherein: the polarization code received in the step (1) is a polarization code with known code rate and unknown code length.

3. The efficient blind detection decoding method for polar codes according to claim 1, wherein: different codes in the candidate code set of the polarization codes have a multiple relation, the front end of the blind detection decoder is designed according to the maximum polarization code length in the candidate code set of the polarization codes, the folding characteristic of the polarization code decoder is utilized, and the decoder with the larger code length is formed by combining the decoders with the smaller code length; the front end of the blind detection decoder inputs a frame sequence with the maximum code length in a group of candidate code sets or polarization code frame sequences with other code lengths in a plurality of groups of candidate code sets in parallel.

4. The efficient blind detection decoding method for polar codes according to claim 1, wherein: and (2) mapping the likelihood ratio of the frozen bit to be greater than 0 into 1 and mapping the likelihood ratio of the frozen bit to be less than 0 into-1.

5. The efficient blind detection decoding method for polar codes according to claim 1, wherein: the threshold detection formula in the step (3) is as follows:

wherein d is₁...d_nFor a threshold interval determined according to simulation statistics, H₁...H_nFor each hypothesis of the candidate for the polarization code, H is the polarization code signal judged according to the metric D threshold, and N is the code length.

6. An efficient blind detection decoder for a polar code, comprising: the device comprises a blind detection decoder front end, a mapping unit, a judgment unit and at least two decoders for decoding polarization codes with different coding modes respectively;