WO2021132983A1 - 통신 시스템에서 극부호의 복호화 장치 및 방법 - Google Patents

통신 시스템에서 극부호의 복호화 장치 및 방법 Download PDF

Info

Publication number
WO2021132983A1
WO2021132983A1 PCT/KR2020/018557 KR2020018557W WO2021132983A1 WO 2021132983 A1 WO2021132983 A1 WO 2021132983A1 KR 2020018557 W KR2020018557 W KR 2020018557W WO 2021132983 A1 WO2021132983 A1 WO 2021132983A1
Authority
WO
WIPO (PCT)
Prior art keywords
bit
decoding
bits
code
polar code
Prior art date
Application number
PCT/KR2020/018557
Other languages
English (en)
French (fr)
Inventor
장민
김종환
Original Assignee
삼성전자 주식회사
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 삼성전자 주식회사 filed Critical 삼성전자 주식회사
Priority to US17/788,526 priority Critical patent/US20230034299A1/en
Priority to EP20904850.3A priority patent/EP4075672A4/en
Publication of WO2021132983A1 publication Critical patent/WO2021132983A1/ko

Links

Images

Classifications

    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, 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
    • H03M13/03Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
    • H03M13/05Error 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
    • H03M13/13Linear codes
    • H03M13/15Cyclic codes, i.e. cyclic shifts of codewords produce other codewords, e.g. codes defined by a generator polynomial, Bose-Chaudhuri-Hocquenghem [BCH] codes
    • H03M13/151Cyclic codes, i.e. cyclic shifts of codewords produce other codewords, e.g. codes defined by a generator polynomial, Bose-Chaudhuri-Hocquenghem [BCH] codes using error location or error correction polynomials
    • H03M13/155Shortening or extension of codes
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, 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
    • H03M13/03Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
    • H03M13/05Error 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
    • H03M13/13Linear codes
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, 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
    • H03M13/37Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
    • H03M13/3707Adaptive decoding and hybrid decoding, e.g. decoding methods or techniques providing more than one decoding algorithm for one code
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, 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
    • H03M13/37Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
    • H03M13/39Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes
    • H03M13/3905Maximum a posteriori probability [MAP] decoding or approximations thereof based on trellis or lattice decoding, e.g. forward-backward algorithm, log-MAP decoding, max-log-MAP decoding
    • H03M13/3927Log-Likelihood Ratio [LLR] computation by combination of forward and backward metrics into LLRs
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, 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
    • H03M13/63Joint error correction and other techniques
    • H03M13/635Error control coding in combination with rate matching
    • H03M13/6362Error control coding in combination with rate matching by puncturing
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, 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
    • H03M13/65Purpose and implementation aspects
    • H03M13/6502Reduction of hardware complexity or efficient processing
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, 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
    • H03M13/03Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
    • H03M13/05Error 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
    • H03M13/11Error 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 using multiple parity bits
    • H03M13/1102Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
    • H03M13/1105Decoding
    • H03M13/1108Hard decision decoding, e.g. bit flipping, modified or weighted bit flipping
    • HELECTRICITY
    • H03ELECTRONIC CIRCUITRY
    • H03MCODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00Coding, 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
    • H03M13/03Error detection or forward error correction by redundancy in data representation, i.e. code words containing more digits than the source words
    • H03M13/05Error 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
    • H03M13/11Error 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 using multiple parity bits
    • H03M13/1102Codes on graphs and decoding on graphs, e.g. low-density parity check [LDPC] codes
    • H03M13/1148Structural properties of the code parity-check or generator matrix
    • H03M13/1174Parity-check or generator matrices built from sub-matrices representing known block codes such as, e.g. Hamming codes, e.g. generalized LDPC codes
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00Arrangements for detecting or preventing errors in the information received
    • H04L1/004Arrangements for detecting or preventing errors in the information received by using forward error control
    • H04L1/0041Arrangements at the transmitter end
    • H04L1/0042Encoding specially adapted to other signal generation operation, e.g. in order to reduce transmit distortions, jitter, or to improve signal shape

Definitions

  • the present disclosure relates generally to resource allocation in a communication system, and more particularly to an apparatus and method for decoding a polar code in a communication system.
  • the 5G communication system or the pre-5G communication system is called a 4G network after (Beyond 4G Network) communication system or an LTE (Long Term Evolution) system after (Post LTE) system.
  • 5G communication systems are being considered for implementation in very high frequency (mmWave) bands (eg, 60 gigabytes (60 GHz) bands).
  • mmWave very high frequency
  • FD-MIMO Full Dimensional MIMO
  • array antenna, analog beam-forming, and large scale antenna technologies are being discussed.
  • an evolved small cell in the 5G communication system, an evolved small cell, an advanced small cell, a cloud radio access network (cloud radio access network, cloud RAN), an ultra-dense network (ultra-dense network) ), Device to Device communication (D2D), wireless backhaul, moving network, cooperative communication, Coordinated Multi-Points (CoMP), and reception interference cancellation ) and other technologies are being developed.
  • cloud radio access network cloud radio access network, cloud RAN
  • ultra-dense network ultra-dense network
  • D2D Device to Device communication
  • wireless backhaul moving network
  • cooperative communication Coordinated Multi-Points (CoMP), and reception interference cancellation
  • FQAM Hybrid Frequency Shift Keying and Quadrature Amplitude Modulation
  • SWSC Sliding Window Superposition Coding
  • ACM Advanced Coding Modulation
  • FBMC Filter Bank Multi Carrier
  • NOMA Non Orthogonal Multiple Access
  • SCMA Sparse Code Multiple Access
  • error detection codes and error correcting codes (ECC) schemes as coding schemes designed to handle errors generated by a communication channel in a receiver.
  • ECC error correcting codes
  • the error detection code is a method that allows the receiver to check whether or not an error is included in the received data
  • the error correction code is a technique that allows the receiver to correct the error included in the received data by itself.
  • the error correction code is often referred to as channel coding.
  • Various methods exist in the error correction coding technique For example, convolutional coding, turbo coding, low-density parity-check coding (LDPC coding), and polar coding schemes exist.
  • the turbo code, the low-density parity check code, and the polar code are excellent codes having performance close to the theoretical channel capacity and are being used in various communication systems today.
  • polar code achieves point-to-point channel capacity with low decoding complexity based on channel polarization that occurs during sequential cancellation (SC) decoding. This is the first theoretically proven code.
  • SC sequential cancellation
  • the present disclosure provides an apparatus and method for reducing the complexity and delay of decoding using a polar code in a communication system.
  • the present disclosure provides an apparatus and method for reducing the number of operations of a decoding operation using a polar code in a communication system.
  • the present disclosure provides an apparatus and method for determining a bit unnecessary for operation in a decoding operation using a polar code in a communication system.
  • a method of operating a receiving apparatus in a communication system includes a process of receiving data including a plurality of bits encoded using a polar code, and decoding of the plurality of bits. It includes the steps of checking at least one bit that is unnecessary for operation, and performing decoding on at least some of the remaining bits except for the at least one bit.
  • a receiving apparatus in a communication system includes a transceiver and at least one processor connected to the transceiver.
  • the at least one processor receives data including a plurality of bits encoded using a polar code, identifies at least one bit that does not require a decoding operation among the plurality of bits, and the at least one Decoding is performed on at least some of the remaining bits except for bits of .
  • the apparatus and method according to various embodiments of the present disclosure may reduce decoding complexity and delay by omitting a portion of a decoding operation.
  • FIG. 1 illustrates a wireless communication system according to various embodiments of the present disclosure.
  • FIG. 2 illustrates an example of a configuration of a receiving end device in a wireless communication system according to various embodiments of the present disclosure.
  • FIG. 3 illustrates an example of an encoding process of a polar code in a communication system according to various embodiments of the present disclosure.
  • FIG. 4 illustrates an example of a polar code decoding process in a communication system according to various embodiments of the present disclosure.
  • FIG. 5 illustrates an example of a graph corresponding to a pole code in a communication system according to various embodiments of the present disclosure.
  • FIG. 6 illustrates an LR calculation process in a communication system according to various embodiments of the present disclosure.
  • FIG. 7 illustrates an example of a decoding binary tree in a communication system according to various embodiments of the present disclosure.
  • FIG. 8 illustrates an example of a binary decision tree in a communication system according to various embodiments of the present disclosure.
  • FIG. 9 is a flowchart illustrating a decoding at a receiving end in a communication system according to various embodiments of the present disclosure.
  • FIG. 10 is a diagram illustrating another flowchart for decoding by a receiving end in a communication system according to various embodiments of the present disclosure.
  • FIG. 11 illustrates an example of a distribution of a start bit index in a communication system according to various embodiments of the present disclosure.
  • FIG. 12 illustrates another example of a distribution of a start bit index in a communication system according to various embodiments of the present disclosure.
  • FIG. 13 illustrates another example of a distribution of a start bit index in a communication system according to various embodiments of the present disclosure.
  • FIG. 14 illustrates an example of an end bit index distribution when a shortened bit exists in a communication system according to various embodiments of the present disclosure.
  • FIG. 15 illustrates a configuration of a decoder in a communication system according to various embodiments of the present disclosure.
  • 16 and 17 show performance of a decoding technique according to various embodiments of the present disclosure.
  • the present disclosure relates to an apparatus and method for decoding using a polar code in a communication system.
  • the present disclosure describes a technique for reducing the complexity of a polar code-based decoding operation in a communication system.
  • the present disclosure provides an error-correcting code that corrects and restores when errors and loss occur or may occur due to various causes such as noise and interference in the process of transmitting or storing data. codes).
  • the present disclosure relates to decoding performed by a receiver in a communication system using a polar code, and relates to a technique for decoding a codeword encoded with a polar code with low complexity and delay.
  • a term that refers to parameters used in the following description a term that refers to a redanduncy bit (eg, parity check bit), a term that refers to an information bit, a term that refers to a channel, a term that refers to control information, and a network object
  • a term that refers to a redanduncy bit eg, parity check bit
  • a term that refers to an information bit e.g. parity check bit
  • a term that refers to an information bit a term that refers to a channel
  • a term that refers to control information e.g., control information
  • a network object referring to (network entities), terms referring to components of an apparatus, and the like are exemplified for convenience of description. Accordingly, the present disclosure is not limited to the terms described below, and other terms having equivalent technical meanings may be used.
  • the polar code is an error correction code proposed by E. Arikan in 2008 and proves that it achieves the channel capacity, the data transmission limit, in all B-DMCs (binary discrete memoryless channels) while having low coding/complexity performance. It is the first error-correcting code that has been
  • the polar code has advantages in error-correction performance and decoding complexity when transmitting a shorter code compared to other channel capacity-approaching codes, such as turbo code and low-density parity-check (LDPC) codes. have. Due to these advantages, it was decided to use the polar code for the purpose of transmitting control information of a short length in the 3GPP NR standardization currently in progress for 5G mobile communication in 2017.
  • a basic polar code decoding method is sequential removal (SC) decoding.
  • the SC decoder sequentially decodes the encoded input bit sequence bit by bit, and uses the bit result decoded up to that point in decoding a certain bit.
  • SCL (SC-list) decoding, SCS (SC-stack) decoding, and SCF (SC-flip) decoding used to improve performance also basically use a sequential decoding method. Due to the characteristics of these sequential decoding operations, SC-based decoders are not suitable to be implemented with a parallelized architecture or a pipelined architecture, and the Turbo code and There is a disadvantage that a clock cycle required for decoding the LDPC code is larger. That is, the SC-based decoder of the polar code generally has a low decoding throughput.
  • the SCL decoder maintains as many candidate groups as the number of preset lists and performs independent decoding on each candidate group. In the decoding process of every bit, each list is decoded independently, and a list with high probability is maintained based on the decoding result and a list with low probability is excluded. Therefore, the number and time of operations performed by the SCL decoder increases in proportion to the number of lists.
  • An object of the present invention is to reduce decoding complexity and delay of SC and SCL decoders of polar code systems that support various coding parameters (code length, code rate, etc.) and settings.
  • the 3GPP NR polar code system encodes bits from length 12 to 1706 and supports various coding rates.
  • An object of the present invention is to effectively reduce the complexity and delay of SC and SCL decoders used in polar code systems supporting various code lengths and code rates.
  • Another object of the present invention is to minimize costs such as additional hardware, computational amount, and computational time involved in achieving the above object.
  • the sign analysis process and operation added to construct the simplified SCL decoder based on the given sign parameter can increase the overall amount of computation and the computation time, contrary to the original purpose of simplifying the decoder.
  • the present invention solves the problem that the prior art as described above is not suitable or efficient for a communication system supporting various parameters. That is, the main purpose of the present invention is to minimize the delay caused by the operation other than the original SCL decoding operation that is additionally performed in supporting various code parameters and code structures.
  • 1 illustrates a wireless communication system according to various embodiments of the present disclosure.
  • 1 illustrates a transmitting device 110 and a receiving device 120 as some of the nodes using a wireless channel in a wireless communication system.
  • 1 shows one transmitter 110 and one receiver 120, it may include a plurality of transmitters or a plurality of receivers.
  • the transmitter 110 and the receiver 120 are described as separate entities, but the functions of the transmitter 110 and the receiver 120 may be interchanged.
  • the transmitting end 110 may be a terminal, and the receiving end 120 may be a base station.
  • the transmitting end 110 may be a base station, and the receiving end 120 may be a terminal.
  • the transmitter 110 may perform encoding using the polar code of the information bits and then transmit.
  • the receiving end 120 may restore the information bits by decoding the coded information transmitted from the transmitting end 110 using a polar code.
  • FIG. 2 illustrates an example of a configuration of a receiving end device in a wireless communication system according to various embodiments of the present disclosure. That is, the configuration illustrated in Fig. 2 may be understood as the configuration of the receiving end 120.
  • Terms such as '... unit' and '... group' used below process at least one function or operation. means a unit, which may be implemented by hardware or software, or a combination of hardware and software.
  • the receiving terminal 120 may include a communication unit 210 , a storage unit 220 , and a control unit 230 .
  • the communication unit 210 may perform functions for transmitting and receiving signals through a wireless channel.
  • the communication unit 210 may perform a function of converting a baseband signal and a bit stream according to a physical layer standard of the system.
  • the communication unit 210 may generate complex symbols by encoding and modulating the transmitted bit stream.
  • the communication unit 210 may restore the received bit stream by demodulating and decoding the baseband signal.
  • the communication unit 210 may up-convert the baseband signal into a radio frequency (RF) band signal, transmit it through the antenna, and down-convert the RF band signal received through the antenna into a baseband signal.
  • RF radio frequency
  • the communication unit 210 may include a transmit filter, a receive filter, an amplifier, a mixer, an oscillator, a digital to analog converter (DAC), an analog to digital converter (ADC), and the like. Also, the communication unit 210 may include a plurality of transmission/reception paths. Furthermore, the communication unit 210 may include at least one antenna array including a plurality of antenna elements. In terms of hardware, the communication unit 310 may be composed of a digital unit and an analog unit, and the analog unit is divided into a plurality of sub-units according to operating power, operating frequency, etc. can be configured.
  • the communication unit 210 transmits and receives signals as described above. Accordingly, the communication unit 210 may be referred to as a 'transmitter', a 'receiver' or a 'transceiver'. In addition, in the following description, transmission and reception performed through a wireless channel are used in the meaning of including processing as described above by the communication unit 210 . Also, the communication unit 210 may include a backhaul communication unit for communication with another network entity connected through a backhaul network.
  • the communication unit 210 may include a decoder (not shown) to perform decoding according to various embodiments of the present disclosure.
  • the decoder may perform a decoding operation based on the polar code.
  • the storage unit 220 may store data such as a basic program, an application program, and setting information for the operation of the receiving terminal 120 .
  • the storage unit 220 may be configured as a volatile memory, a nonvolatile memory, or a combination of volatile memory and nonvolatile memory.
  • the storage 220 may provide the stored data according to the request of the controller 230 .
  • the controller 230 may control overall operations of the receiver 120 .
  • the control unit 230 may transmit and receive signals through the communication unit 220 .
  • the controller 230 may write or read data to or from the storage 230 .
  • the controller 230 may include at least one processor or microprocessor, or may be a part of the processor.
  • the control unit 230 may control the operation of each component included in the communication unit 220 .
  • the controller 230 may control the receiver 120 to perform operations according to various embodiments to be described later.
  • FIG. 3 illustrates an example of an encoding process of a polar code in a communication system according to various embodiments of the present disclosure. Among the operations illustrated in FIG. 3 , a part of the process may be omitted or a separate operation may be added according to system requirements and characteristics.
  • the number of information bits to be transmitted by the transmitter is expressed as A
  • the number of codeword bits transmitted through the channel after encoding is expressed as E.
  • Each step of encoding performed at the transmitting end is as follows.
  • An information bit sequence b ⁇ b 0 ,b 1 ,...,b A-1 ⁇ of length A to be transmitted is given.
  • the information bit sequence may be a part of the entire information to be transmitted, that is, a segment.
  • the information bit sequence b is first coded into a concatenated outer code to improve performance.
  • a concatenated outer code is usually used to improve the performance of a decoder that performs decoding in consideration of a plurality of codeword candidates, such as SCL (SC-list) decoding of polar codes.
  • SCL decoding will be described in detail below in the operation of the receiving end.
  • the external code used in conjunction with the polar code includes an error detection code such as a cyclic redundancy check (CRC) code, an error correction code such as a BCH code, a parity check (PC) code, and a convolutional code. Only one outer code may be used, or two or more outer codes may be used in combination.
  • the polar code used for uplink control information is used together with the 11-bit CRC code if the information bit length is 20 bits or more according to the length of the information bit, and the information bit If the length is 19 bits or less, it is used together with a 3-bit PC code and a 6-bit CRC code.
  • N is the size of the mother polar code, which is a power of 2, and is determined by a predetermined criterion among values larger than K.
  • the bit sequence u is an encoder input bit sequence of a polar code encoder, and bits of b' are mapped to u according to a predetermined method and criteria ( 304 ).
  • Each bit of the encoded input bit sequence u can then be interpreted as if passing through a sub-channel, which is a virtual channel of different quality, by channel polarization by the operation of the transmitting end and the receiving end. .
  • each subchannel is also referred to as a synthetic channel.
  • the channel capacity of each subchannel, the Bhatacharayya parameter, the result of density evolution, etc. are used.
  • a rate-matching operation performed later is considered. Because of this feature, the process of mapping b' to u is also referred to as a subchannel assignment process.
  • a bit of u corresponding to a subchannel to which b' is mapped is referred to as a non-frozen bit, and a bit of u corresponding to the remaining subchannel is referred to as a frozen bit.
  • the freeze bit has a fixed value, which is usually zero.
  • the encoded input bit sequence u of length N is multiplied by a generator matrix G of the polar code, thereby generating the encoded output bit sequence x of the same length N.
  • the generative matrix G was defined as [Equation 1].
  • the operation means Kronecker power n times.
  • B N is a size N ⁇ N bit-reversal permutation matrix.
  • a vector of length 8 ⁇ a 0 ,a 1 ,a 2 ,a 3 ,a 4 ,a 5 ,a 6 ,a 7 ⁇ and an 8 ⁇ 8 bit-reversal permutation matrix B 8 is multiplied to obtain ⁇ a 0 ,a 4 ,a 2 ,a 6 ,a 1 ,a 5 ,a 3 ,a 7 ⁇ in which the index is bit-reversal permutated.
  • a simple form of a generation matrix such as [Equation 2] except for B N may be used.
  • the code rate adjusting unit 308 an operation of generating a bit sequence of length E to be transmitted from the generated encoded output bit sequence x is performed.
  • a transmission bit sequence obtained by adjusting the code rate in this way is represented by c.
  • the encoding output bit sequence x may be readjusted to improve the performance of the polar code.
  • the coded output bit sequence x is interleaved in units of 32 sub-blocks, stored in a circular buffer, and sequentially extracted so that the code of length E is A sequence is created. If the length E of the codeword is smaller than the size E of the polar code parent code, one of puncturing and shortening may be performed.
  • FIG. 4 illustrates an example of a polar code decoding process in a communication system according to various embodiments of the present disclosure. Among the operations illustrated in FIG. 4 , other operations may be added or included operations may be omitted according to system requirements.
  • Probability information corresponding to the transmitted bit c is generated by demodulating the received signal. Probability information is given as values such as a probability vector, a likelihood ratio (LR), and a log-likelihood ratio (LLR). For brevity, the LLR is considered below unless otherwise noted. However, the present invention is not limited to the LLR-based decoding operation.
  • the LLR sequence l of length E is input to the polar code decoder 404 of length N, and undergoes an inverse process of code rate-adjustment at the transmitting end. If puncturing occurs at the code rate adjusting end of the transmitting end, the LLR value for the corresponding bit is determined to be 0. If shortening occurs, the LLR value for the corresponding bit is the maximum LLR value corresponding to the bit value 0. determined by value. If repetition occurs for a specific bit, the LLR value for the corresponding bit is determined by combining all the corresponding LLR values.
  • SC-based decoding includes general SC decoding, SCL (SC-list) decoding, and SCS (SC-stack) decoding. SC-based decoding sequentially decodes each bit of the encoded input sequence one bit according to the index order. These decoders perform decoding on each bit in the order of the index values of the encoded input bit sequence, that is, in the order of u 0 , u 1 , ..., u N-1. Specifically, the decoding of the i-th bit, u i is performed by the following procedure.
  • each bit is performed based on the bit value that has been previously decoded and estimated. For example, when bit u i is decoded, it is an estimated value of the previously decoded bit (u 0 ,...,u i-1 ). and probability information about it or the cumulative value of a value corresponding to the probability information, etc. are used.
  • each partial bit sequence is referred to as a list or path, and the accumulated probability information calculated while decoding for each path or a value corresponding to the probability information is referred to as a path-metric (PM).
  • the SCL decoder has a list of L number of list sizes set when bit u i is decoded. It is a method of decoding while maintaining .
  • the SCL decoder maintains a list Calculate probability information for u i values 0 and 1 based on . And, each list considered when calculating the probability information of each u i Updates the probability information or values corresponding to the probability information for 0 and 1 to the u i bit values in PM for , and a total of 2L lists Calculate the PM for . If u i is a frozen bit, the estimated value is the determined bit value promised by the transmitter/receiver regardless of the calculated PM this is decided If u i corresponds to the information bit among the unfrozen bits, the L lists determined to have high probability based on the PM value among the total 2L lists.
  • the SCL decoder maintains a total of L lists for each bit decoding and performs decoding. And, when decoding of all bits is completed, a codeword with the highest probability among the finally obtained L lists, that is, candidate bit sequences, is selected based on the PM. If L is set to 1, the SCL decoder operates the same as the basic SC decoder. In the middle or after completion of SCL decoding, error correction performance can be improved by using concatenated external encoding. For example, when CRC codes are concatenated and used, a codeword with the highest probability while satisfying the CRC code constraint condition among the L lists obtained after decoding is estimated as the final decoding result.
  • the estimated encoded input bit sequence A message bit at a predetermined position is obtained from Finally, the estimated message bit sequence is is marked with
  • the generator matrix multiplication process of the encoding process described with reference to FIG. 3 and the SC-based decoding process of the decoding process described with reference to FIG. 4 may be described using a graph corresponding to the polar code.
  • 5 illustrates an example of a graph corresponding to a pole code in a communication system according to various embodiments of the present disclosure. 5 illustrates a bipartite graph for a polar code of size 8.
  • the binary graph includes variable nodes 511 represented by circles, check nodes 512 represented by squares, and edges connecting these nodes. (513).
  • a variable node represents an individual bit, and in the decoding process, each variable node represents probability information such as LLR for the corresponding bit.
  • a check node represents a linear constraint of connected neighboring variable nodes, and a modulo-2 sum of all bits corresponding to a neighboring variable node of one check node is 0.
  • a polar sign graph of length N may consist of log 2 N+1 stages.
  • the symbol t indicates the index of a step in the graph
  • One of the characteristics of the polar graph is that the whole graph is composed of connections of 2 ⁇ 2 elementary elements 531 having a Z shape.
  • the connection between basic elements in the entire polar code graph is determined from the generative matrix of [Equation 2].
  • a generator matrix multiplication process among the encoding process described with reference to FIG. 3 may be understood from the graph of FIG. 5 .
  • the leftmost N variable nodes 521 correspond to the encoded input bit sequence u of length N of the encoding process of FIG. 3
  • the rightmost N variable nodes 522 are encoded output of length N Corresponds to the bit sequence x.
  • the encoding process represents a process of calculating the rightmost encoded output bit sequence x from the leftmost encoded input bit sequence u to satisfy the relational expression of the check node. That is, the encoding process may be understood as an operation of updating the bit value in stages according to the relationship between the variable nodes and the check nodes connected from left to right of the graph illustrated in FIG. 5 .
  • the SC-based decoding process may also be understood as belief-propagation on the graph of FIG. 5 .
  • soft-value probability information such as code rate-inverse-adjusted N intrinsic LLRs are input to the rightmost N variable nodes 521 of the graph.
  • soft-value probability information is an LLR for concise description, but this assumption does not limit the scope of application of the present invention.
  • SC-based decoding is performed according to a predetermined procedure.
  • Sequential decoding of each bit of the encoded input bit sequence u is a process of calculating an a-posteriori LLR (AP-LLR) for the corresponding bit and estimating a bit value based on the calculated LLR.
  • AP-LLR a-posteriori LLR
  • the AP-LLR of each bit is calculated from the intrinsic LLR value given on the rightmost side of the graph. This process sequentially calculates the LLR values for the variable nodes for each stage from the right to the left of the graph.
  • the function sgn(x) is a function that outputs the sign of x, and outputs 1 if the value of x is greater than 0, and -1 if it is less than 0.
  • a function for calculating the LLR of a 1 as in [Equation 4] or [Equation 5] is referred to as an f function. Estimated value of bit a 1 as decoding progresses When this is obtained, as shown in the right 620 of FIG. 6, through the following [Equation 6] to calculate
  • a function for calculating the LLR of a 2 as shown in Equation 6 may be referred to as a g function.
  • bit is a non-frozen bit
  • a hard-decision value of the updated LLR value is substituted.
  • the estimated bits are transferred from the left to the right of the graph to the variable nodes in the possible range in the current decoding step, and are used in the LLR calculation of other variable nodes as in the process of [Equation 6].
  • 7 illustrates an example of a decoding binary tree in a communication system according to various embodiments of the present disclosure.
  • 7 exemplifies a decoded binary tree widely used to efficiently explain the types and procedures of operations performed by the aforementioned SC-based decoder, and is an example of decoding a polar code of size 8.
  • This binary tree is useful for identifying the types of operations performed by the SC decoder and understanding the order in which each operation is performed.
  • the depth corresponds to the stage of the polar sign graph of FIG. 5 .
  • the SC-based decoding procedure can be understood as visiting or searching each leaf node from the root node in a binary tree in a depth-first search (DFS) manner.
  • DFS depth-first search
  • the LLR values of the corresponding variable nodes are calculated by the f function of [Equation 4] or [Equation 5], and the right node of the tree
  • the LLR values of the corresponding variable nodes are calculated by the g function of [Equation 6].
  • the LLR values of half of the variable nodes indicated by the child node are placed at the node location using the LLRs of the variable nodes indicated by the parent node.
  • the function f or g function If the LLR value of the i-th leaf node is calculated by performing the above process in the DFS method, this value becomes the AP-LLR for the corresponding i-th coded input bit u i .
  • each Z-shaped base It enables g function operation such as [Equation 6] on the element.
  • the type and number of operations for each step on the polar code graph performed for decoding the i-th encoded input bit u i can be confirmed. For example, when decoding u 0 in a polar code of length 8, f, f, and f operations are performed 4 times, 2 times, and 1 time in steps 2, 1, and 0 of the polar code graph, respectively. As another example, when decoding u 6 , g, g, f operations are performed 4 times, 2 times, and once in steps 2, 1, and 0 of the polar code graph. If this is generalized, the type of operation for each step performed for decoding u i can be identified from the binary representation of the bit index i.
  • step t the number of variable nodes for which the LLR value is calculated in step t, that is, the total number of operations, is to be.
  • the total number of f function and g function operations required for SC decoding of the polar code is N(N-1).
  • N-1 operations For example, considering decoding u 4 and u 5 from the polar code of length 8, g,f,f function operations occur in steps 2, 1, and 0 respectively in steps 2, 1, and 0 when bit u 4 is decoded, and u 5 During decoding, g, f, g function operations are performed in steps 2, 1, and 0, respectively.
  • the leaf node corresponding to 5 is found.
  • SCL decoding is performed as follows.
  • SCL decoding is a method of maintaining L candidate groups determined in advance in the step of calculating and estimating the LLR value for each encoding input bit.
  • the SCL decoder decodes bit u i , the L bits obtained in the decoding process of the preceding bits u 0 ,u 1 ,...,u i-1
  • a path-metric PM is probability information for each list, and in the case of an LLR-based SCL decoder, the lower the PM of the list, the higher the probability of the list.
  • the SCL decoder initializes the PM of each list to a predetermined value before starting decoding, and the initial value is usually 0.
  • the SCL decoder uses each list 2L lists by calculating probability information such as LLR of u i for u i , and LLR etc. for a case where u i is 0 and 1, respectively , Update the PM for Updated 2L list Among them, L lists are selected based on PMs and the rest are pruned. These processes are sequentially performed according to the bit index.
  • the number of lists considered in the process of decoding the first u 0 is one, and in the case of the beginning of decoding where the index of the bit to be decoded is small, the number of lists considered is less than L.
  • 8 illustrates an example of a binary decision tree in a communication system according to various embodiments of the present disclosure.
  • 8 is an embodiment of a binary decision tree depicting a process in which an SCL decoder estimates sequential bits for a polar code of length 8.
  • the binary decision tree of FIG. 8 is different from the polar code graph of FIG. 5 and the decoded binary tree of FIG. 6 .
  • SCL decoding decodes encoded input bits in the order of bit indexes. This operation of SCL decoding can be understood as an operation of sequentially descending by one depth while maintaining a maximum of L among possible paths starting from the root node in the binary decision tree of FIG. 8 .
  • FIG. 8 The example of FIG.
  • bits u 0 , u 1 , u 2 are freeze bits, the bit value is determined to be 0 regardless of the calculated value of the LLR. Assuming that bits u 0 , u 1 , and u 2 are all 0, an LLR value for the non-frozen bit u 3 is calculated. shown in Fig. is the encoded input bit u i is the value It is called a branch metric (BM) as a probability value for the case of having , and is obtained from the LLR value. PM is obtained by considering the BM for each bit together.
  • BM branch metric
  • bit u 4 is a frozen bit
  • u 4 is a freeze bit and the value is 0, the values of (u 0 ,u 1 ,u 2 ,u 3 ) assumed in each list are different, so different LLRs and BMs can be calculated. .
  • decoding of the non-frozen bit u 5 is performed. In the example of FIG.
  • the decoder selects and outputs one list determined to have the highest probability based on PM. If external encoding such as a CRC code is concatenated together, a list determined to have the highest probability based on PM while satisfying the CRC code condition from among the list is selected and output.
  • the decoding technique may allow the SCL decoder to have low complexity and delay without loss of performance. To this end, an operation performed by the SCL decoder is analyzed in more detail to identify areas that can be improved in terms of complexity/delay. A decoding process according to an embodiment will be described with reference to the binary decision tree of FIG. 8 .
  • the BM calculated according to the estimated value is considered in the process of selecting the maximum L paths or lists in the binary decision tree.
  • the BM calculated for the frozen bits before the first non-freezing bit is included as the same value in all PMs thereafter, which acts as a kind of constant included between different PMs. do. Accordingly, the BM calculated for the frozen bits before the first non-frozen bit in the SCL decoder has no effect in selecting the L lists while sequential decoding is performed thereafter.
  • the encoded input bit are frozen bits, and there is no need to update the PM for the corresponding bits based on the above description, and furthermore, there is no need to perform decoding on the corresponding bits.
  • a shortening operation may be performed to adjust the code rate.
  • the shortening of the error correction code is an operation of reducing the length (eg, number of output bits) and dimension (eg, number of input bits) of the code by the same amount.
  • the value of the shortened bit is fixed the same as the freeze bit, and is generally 0.
  • the shortened bits and the frozen bits are not distinguished, but the shortened bits and the non-freeze bits are used for other purposes.
  • bits to be shortened can be determined by a binary domination rule. Considering shortening J bits in the polar code of length N, the index set of shortened bits is when said, According to the nature of binary domination, ⁇ 0,... ,N-1 ⁇ is composed of indices having a large value. When bits having a large index are shortened, it means that later bits in a sequential decoding order are shortened.
  • One motivation of the decoding technique is that there is no need to update the PM for shortened bits located after the last unfrozen bit in the shortened polar code. Since the value of the shortened bit is usually fixed to 0, it does not contain information. Accordingly, the decoder does not need to perform decoding on the corresponding bit. In addition, due to the causality of bit generation of the polar code, the shortened bit at the back in the decoding order does not affect the generation and decoding of other bits located at the front. If the largest value among the non-frozen bit indexes in the shortened polar code is q max , the encoding input bit are shortened bits, and there is no need to update the PM for the shortened bits as described above. That is, decoding of the shortened bits may not be performed.
  • the decoding operation performed based on the above-described frozen bits and shortened bits, that is, the operation of the receiving end, is as shown in FIG. 9 or FIG. 10 below.
  • 9 is a flowchart illustrating a decoding at a receiving end in a communication system according to various embodiments of the present disclosure. 9 illustrates an operation method of the receiving end 120 .
  • the receiving end receives data.
  • Data includes a plurality of bits encoded using a polar code.
  • the receiving end may receive downlink control information (DCI) prior to data reception.
  • DCI downlink control information
  • UCI uplink control information
  • the receiving end checks at least one bit of the previous stage that is not required to be decoded.
  • the receiving end may check at least one bit based on information included in DCI or UCI.
  • the receiving end decodes at least some of the remaining bits. For example, when at least one bit unnecessary for decoding also exists at the rear end of the plurality of bits, the receiving end may end the decoding operation before the last bit of the plurality of bits.
  • 10 is a diagram illustrating another flowchart for decoding by a receiving end in a communication system according to various embodiments of the present disclosure. 10 illustrates an operation method of the receiving end 120 .
  • step 1001 the receiving end checks the number of input bits K and the number of output bits E of the polar sign. In other words, the receiving end identifies a code dimension corresponding to the number of input bits of a code, which is a given code parameter, and a code length corresponding to the number of output bits.
  • the receiving end configures the code of the polar code to be used. For example, the receiving end determines a mother code of the polar code to be used, a rate matching method, and the like based on the code parameter.
  • the receiving end determines a start bit index s and an end bit index e of encoding.
  • the receiving end may determine the bit index s to start the SCL decoder and the bit index e to end the SCL decoder based on the determined sign parameter, sign setting, and the like.
  • various methods may be utilized. According to an embodiment, s and e may be determined by an equation related to a sign parameter and a setting. According to another embodiment, s and e may be determined by a reference table related to a sign parameter and a setting. Also, according to another embodiment, s and e may be determined by an appropriate function related to a sign parameter and setting. According to another embodiment, s and e may be determined based on at least one of machine learning, deep learning, and reinforcement learning. Based on the determined start index and end bit index, SCL decoding is performed.
  • step 1007 the receiving end starts decoding from bit u s.
  • SCL decoding starts by calculating an LLR for u s . That is, prior to decoding u s , no operation may be performed. Therefore, since the calculated result to the dictionary that can be used for LLR calculations of u s not, and all the operations necessary for decoding the u s can be performed. That is, based on the intrinsic LLR of step log 2 N given through the channel when calculating the LLR of u s, at each step t The LLR values for the variable nodes are sequentially calculated. In this case, the total number of operations becomes N-1 times as in [Equation 7].
  • step 1009 the receiving end sequentially decodes up to bit u e. That is, the receiving end performs decoding on bits u s+1 ,...,u e. In this case, when decoding each bit, a result of an operation performed on bits decoded before the corresponding bit may be utilized.
  • the receiving end selects an estimated encoded input bit and outputs it. That is, when decoding of u e is completed, the receiving end ends SCL decoding and outputs one of the lists.
  • the output list may be selected based on at least one of a decoding result for PM and external encoding.
  • the SCL decoder of the receiving end does not perform decoding on at least one bit before u s and at least one bit after u e , and may omit it, thereby reducing the amount of computation and delay required for decoding. is reduced
  • the start/end bit indexes s and e may be determined as follows.
  • N-1 the rate-matching method of the polar code
  • the decoder identifies the smallest unfrozen bit index q min and the largest unfrozen bit index q max of the polar code configured according to the given sign parameter and setting, and, accordingly, 0 ⁇ s ⁇ q min, respectively , s and e may be determined within a range satisfying q max ⁇ e ⁇ N-1 .
  • 11 illustrates an example of a distribution of a start bit index in a communication system according to various embodiments of the present disclosure.
  • 11 illustrates the first non-frozen bit index of the polar code, ie, q min , constructed according to the number of encoded input bits A and the number of output bits E after code rate-adjustment in a polar code system for 3GPP NR downlink.
  • the value of q min is indicated by light and dark, the brighter the value is, the larger the value is, and the darker the value is, the smaller the value is.
  • the q min values are the same in a range of (A, E) combinations.
  • the q min values for similar parameter (A,E) combinations using the same parent code size N are also similar.
  • high index bits are shortened, and q min is decreased because bits of a relatively low index become non-frozen bits. Utilizing these features, it is possible to efficiently determine the start bit index s based on the code parameters (A, E), the parent code size N, the used code rate-adjustment method, and the like.
  • the complexity and processing time of polar code decoding are proportional to the size N of the parent code.
  • the start bit index s may be adaptively determined according to the determined N. The following shows an example of determining the start bit index s in the polar code system for downlink as shown in FIG. 12 . It should be noted that the following embodiment is for the simplest system for convenience, and an actual system may be configured by more various branches.
  • the following shows another example of determining the start bit index s in the polar code system for downlink as shown in FIG. 13 .
  • 14 shows the last non-frozen bit index of the polar code, ie, q max , constructed according to the number of encoded input bits A and the number of output bits E after code rate-adjustment in a polar code system for 3GPP NR downlink. 14 shows q max only for the case where the configuration of the pole code uses a short axis. In this way, based on a pre-determined based on the information of the q max, it is determined from e q max ⁇ e ⁇ N-1 range.
  • Representative methods are a simplified SC (simplified SC, SSC) decoding technique and a simplified SSCL (simplified SCL, SSCL) decoding technique. These methods modify the decoded binary tree shown in FIG. 5 according to the sign parameter, combine some nodes to make a special node, and efficiently process the special node as a separate method, thereby reducing the overall operation.
  • Representative special nodes include a rate-0 node, a rate-1 node, a repetition node, and a single parity-check (SPC) node.
  • the SSC decoder or SSCL decoder In order to implement the SSC decoder or SSCL decoder, it is required to precede the operation of scheduling the operation of the decoder, such as analysis and modification of the decoded binary tree according to the sign parameter. If the number of combinations of code parameters to be supported by the system is small, by storing previously scheduled information in a memory or the like, efficient implementation is possible. However, when considering countless combinations of code parameters as in the 3GPP NR polar code system, it is difficult to store previously scheduled information in the memory, and the decoder must analyze the decoded binary tree in advance for each code parameter and perform scheduling. do. In addition, a separate function, process, and memory are required to process special nodes (eg, rate-0, rate-1, repetition, SPC node, etc.).
  • special nodes eg, rate-0, rate-1, repetition, SPC node, etc.
  • the decoder may be implemented more simply than the aforementioned SSC decoder or SSCL decoder.
  • 15 illustrates a configuration of a decoder in a communication system according to various embodiments of the present disclosure.
  • the decoder includes a decoding controller 1520 and an SCL decoder 1570 .
  • the decoding controller 1520 determines a start bit index or an end bit index according to one of the above-described embodiments.
  • the SCL decoder 1570 performs a decoding operation based on the start bit index or the end bit index determined by the decoding controller 1520 .
  • Scheduling of decoding required for a decoder according to various embodiments is simpler than SSC and SSCL decoding.
  • the decoding operation according to various embodiments of the present disclosure does not require analysis of the decoded binary tree, but only requires to determine a start bit index and an end bit index.
  • the decoding method according to various embodiments of the present disclosure does not consider a special node, it is not required to configure a separate function, process, memory, etc. according to the type and size of the special node.
  • PM and the like are calculated for each special node, but the decoder according to various embodiments of the present disclosure does not perform the operation itself on some bits.
  • 16 and 17 show performance of a decoding technique according to various embodiments of the present disclosure.
  • 16 and 17 consider a 3GPP NR uplink polar code system.
  • 16 and 17 show a latency reduction ratio and complexity of a decoding technique according to various embodiments of the present disclosure compared to general SCL decoding according to the number of polar code input bits A and the number of code rate-adjusted output bits E reduction ratio) respectively.
  • the degradation rate in each parameter is obtained by calculating the actual number of operations and steps performed by the general SCL decoder and the SCL decoder of the present invention.
  • a degradation rate of 0 means that there is no degradation in complexity and delay
  • a degradation rate of 1 means that all calculations are omitted.
  • the decoding complexity and delay reduction according to the present invention are increased up to a level of 40%.
  • uplink control information (UCI) transmitted through the 3GPP NR uplink is generally 100 bits or less, a significant reduction in decoding complexity and delay can be expected in most UCI transmission situations.
  • the decoding technique in decoding the polar code coded codeword with an SCL decoder or the like, unnecessary operations are omitted, thereby reducing the amount of required calculations and the calculation time.
  • unnecessary operations are omitted, thereby reducing the amount of required calculations and the calculation time.
  • the decoding technique according to various embodiments of the present disclosure can fully and comprehensively support decoding of codewords encoded with a wide range of code dimensions, code lengths, and code rates. have.
  • when a low code rate is used, about half or more of operations are omitted, and thus decoding complexity and delay can be reduced by about 40%.
  • a computer-readable storage medium storing one or more programs (software modules) may be provided.
  • One or more programs stored in the computer-readable storage medium are configured to be executable by one or more processors in an electronic device (device).
  • One or more programs include instructions for causing an electronic device to execute methods according to embodiments described in a claim or specification of the present disclosure.
  • Such programs include random access memory, non-volatile memory including flash memory, read only memory (ROM), electrically erasable programmable ROM (electrically erasable programmable read only memory, EEPROM), magnetic disc storage device, compact disc-ROM (CD-ROM), digital versatile discs (DVDs), or other It may be stored in an optical storage device or a magnetic cassette. Alternatively, it may be stored in a memory composed of a combination of some or all thereof. In addition, each configuration memory may be included in plurality.
  • non-volatile memory including flash memory, read only memory (ROM), electrically erasable programmable ROM (electrically erasable programmable read only memory, EEPROM), magnetic disc storage device, compact disc-ROM (CD-ROM), digital versatile discs (DVDs), or other It may be stored in an optical storage device or a magnetic cassette. Alternatively, it may be stored in a memory composed of a combination of some or all thereof. In addition, each configuration memory may be included in plurality.
  • the program is transmitted through a communication network consisting of a communication network such as the Internet, an intranet, a local area network (LAN), a wide area network (WAN), or a storage area network (SAN), or a combination thereof. It may be stored on an attachable storage device that can be accessed. Such a storage device may be connected to a device implementing an embodiment of the present disclosure through an external port. In addition, a separate storage device on the communication network may be connected to the device implementing the embodiment of the present disclosure.
  • a communication network such as the Internet, an intranet, a local area network (LAN), a wide area network (WAN), or a storage area network (SAN), or a combination thereof. It may be stored on an attachable storage device that can be accessed.
  • Such a storage device may be connected to a device implementing an embodiment of the present disclosure through an external port.
  • a separate storage device on the communication network may be connected to the device implementing the embodiment of the present disclosure.
  • the present disclosure relates generally to resource allocation in a communication system, and more particularly to an apparatus and method for decoding a polar code in a communication system.

Landscapes

  • Physics & Mathematics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Algebra (AREA)
  • General Physics & Mathematics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Error Detection And Correction (AREA)

Abstract

본 개시는 LTE(Long Term Evolution)와 같은 4G(4th generation) 통신 시스템 이후 보다 높은 데이터 전송률을 지원하기 위한 5G(5th generation) 또는 pre-5G 통신 시스템에 관련된 것이다. 본 개시는 통신 시스템에서 극부호(polar code)의 복호화를 위한 것으로, 수신 장치의 동작 방법은, 극부호(polar)를 이용하여 부호화된, 복수의 비트들을 포함하는 데이터를 수신하는 과정과, 상기 복수의 비트들 중 복호 연산이 불필요한 적어도 하나의 비트를 확인하는 과정과, 상기 적어도 하나의 비트를 제외한 나머지 비트들 중 적어도 일부에 대한 복호를 수행하는 과정을 포함한다.

Description

통신 시스템에서 극부호의 복호화 장치 및 방법
본 개시(disclosure)는 일반적으로 통신 시스템에서 자원 할당에 관한 것으로, 보다 구체적으로 통신 시스템에서 극부호(polar code)의 복호화를 위한 장치 및 방법에 관한 것이다.
4G(4th generation) 통신 시스템 상용화 이후 증가 추세에 있는 무선 데이터 트래픽 수요를 충족시키기 위해, 개선된 5G(5th generation) 통신 시스템 또는 pre-5G 통신 시스템을 개발하기 위한 노력이 이루어지고 있다. 이러한 이유로, 5G 통신 시스템 또는 pre-5G 통신 시스템은 4G 네트워크 이후(Beyond 4G Network) 통신 시스템 또는 LTE(Long Term Evolution) 시스템 이후(Post LTE) 시스템이라 불리어지고 있다.
높은 데이터 전송률을 달성하기 위해, 5G 통신 시스템은 초고주파(mmWave) 대역(예를 들어, 60기가(60GHz) 대역과 같은)에서의 구현이 고려되고 있다. 초고주파 대역에서의 전파의 경로손실 완화 및 전파의 전달 거리를 증가시키기 위해, 5G 통신 시스템에서는 빔포밍(beamforming), 거대 배열 다중 입출력(massive MIMO), 전차원 다중입출력(Full Dimensional MIMO, FD-MIMO), 어레이 안테나(array antenna), 아날로그 빔형성(analog beam-forming), 및 대규모 안테나(large scale antenna) 기술들이 논의되고 있다.
또한, 시스템의 네트워크 개선을 위해, 5G 통신 시스템에서는 진화된 소형 셀, 개선된 소형 셀(advanced small cell), 클라우드 무선 액세스 네트워크(cloud radio access network, cloud RAN), 초고밀도 네트워크(ultra-dense network), 기기 간 통신(Device to Device communication, D2D), 무선 백홀(wireless backhaul), 이동 네트워크(moving network), 협력 통신(cooperative communication), CoMP(Coordinated Multi-Points), 및 수신 간섭제거(interference cancellation) 등의 기술 개발이 이루어지고 있다.
이 밖에도, 5G 시스템에서는 진보된 코딩 변조(Advanced Coding Modulation, ACM) 방식인 FQAM(Hybrid Frequency Shift Keying and Quadrature Amplitude Modulation) 및 SWSC(Sliding Window Superposition Coding)과, 진보된 접속 기술인 FBMC(Filter Bank Multi Carrier), NOMA(Non Orthogonal Multiple Access), 및 SCMA(Sparse Code Multiple Access) 등이 개발되고 있다.
일반적으로, 통신 시스템 또는 방송 시스템에서 송신기와 수신기 사이에 데이터를 송신 및 수신하는 경우, 통신 채널에 존재하는 잡음으로 인해 데이터 오류가 발생할 수 있다. 이처럼 통신 채널에 의해 발생된 오류를 수신기에서 처리할 수 있도록 설계된 부호화 방식으로 오류 검출 부호(error detection codes) 및 오류 정정 부호(error correcting codes, ECC) 방식이 존재한다. 오류 검출 부호는 수신기가 수신한 데이터에 오류가 포함되어 있는지 여부를 확인할 수 있도록 하는 방법이며, 오류 정정 부호는 수신기가 수신한 데이터에 포함된 오류를 스스로 정정할 수 있도록 하는 기법이다. 여기서 오류 정정 부호는 흔히 채널 부호화(channel coding) 라고도 불린다.
오류 정정 부호 기법에는 다양한 방식들이 존재한다. 예컨대, 컨볼루션 부호(convolutional coding), 터보 부호(turbo coding), 저밀도 패리티 검사 부호(low-density parity-check coding, LDPC coding), 및 극부호(polar coding) 방식 등이 존재한다. 특히 터보 부호, 저밀도 패리티 검사 부호, 극부호는 이론적인 채널 용량(channel capacity) 에 거의 근접하는 성능을 갖는 우수한 부호로 오늘날 다양한 통신 시스템에서 활용되고 있다. 이러한 오류 정정 부호 기법들 중 polar code 은 순차적 제거(successive cancellation, SC) 복호 시 발생하는 채널 양극화(channel polarization) 현상을 바탕으로 낮은 복호 복잡도로 점대점(point-to-point) 채널 용량을 달성함이 이론적으로 증명된 최초의 부호이다. 또한, 짧은 길이의 극부호는 SCl(SC-list) 복호 등을 사용하였을 때 다른 채널 부호 대비 성능이 우수함이 확인되었다. 이러한 이유로 5G 통신 표준인 3GPP 릴리즈(release)-15 NR에서 제어 정보를 전송할 때 극부호를 사용하고 있다.
상술한 바와 같은 논의를 바탕으로, 본 개시(disclosure)는, 통신 시스템에서 극부호(polar code)를 이용한 복호화의 복잡도 및 지연을 낮추기 위한 장치 및 방법을 제공한다.
또한, 본 개시는, 통신 시스템에서 극부호를 이용한 복호화 동작의 연산 횟수를 줄이기 위한 장치 및 방법을 제공한다.
또한, 본 개시는, 통신 시스템에서 극부호를 이용한 복호화 동작에서 연산이 불필요한 비트를 결정하기 위한 장치 및 방법을 제공한다.
본 개시의 다양한 실시 예들에 따르면, 통신 시스템에서 수신 장치의 동작 방법은, 극부호(polar)를 이용하여 부호화된, 복수의 비트들을 포함하는 데이터를 수신하는 과정과, 상기 복수의 비트들 중 복호 연산이 불필요한 적어도 하나의 비트를 확인하는 과정과, 상기 적어도 하나의 비트를 제외한 나머지 비트들 중 적어도 일부에 대한 복호를 수행하는 과정을 포함한다.
본 개시의 다양한 실시 예들에 따르면, 통신 시스템에서 수신 장치는, 송수신기와, 상기 송수신기와 연결된 적어도 하나의 프로세서를 포함한다. 상기 적어도 하나의 프로세서는, 극부호(polar)를 이용하여 부호화된, 복수의 비트들을 포함하는 데이터를 수신하고, 상기 복수의 비트들 중 복호 연산이 불필요한 적어도 하나의 비트를 확인하고, 상기 적어도 하나의 비트를 제외한 나머지 비트들 중 적어도 일부에 대한 복호를 수행한다.
본 개시의 다양한 실시 예들에 따른 장치 및 방법은, 복호 동작의 일부를 생략함으로써 복호 복잡도와 지연을 낮출 수 있다.
본 개시에서 얻을 수 있는 효과는 이상에서 언급한 효과들로 제한되지 않으며, 언급하지 않은 또 다른 효과들은 아래의 기재로부터 본 개시가 속하는 기술 분야에서 통상의 지식을 가진 자에게 명확하게 이해될 수 있을 것이다.
도 1은 본 개시의 다양한 실시 예들에 따른 무선 통신 시스템을 도시한다.
도 2는 본 개시의 다양한 실시 예들에 따른 무선 통신 시스템에서 수신단 장치의 구성의 예를 도시한다.
도 3은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 극부호의 부호화 과정의 일 예를 나타낸다.
도 4는 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 극부호의 복호화 과정의 일 예를 나타낸다.
도 5는 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 극부호에 대응하는 그래프의 일 예를 도시한다.
도 6는 본 개시의 다양한 실시 예에 따른 통신 시스템에서 LR 계산 과정을 도시한다.
도 7은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 복호 이진 트리(decoding binary tree)의 일 예를 도시한다.
도 8은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 이진 결정 트리(binary decision tree)의 예를 도시한다.
도 9는 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 수신단의 복호화를 위한 흐름도를 도시한다.
도 10은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 수신단의 복호화를 위한 다른 흐름도를 도시한다.
도 11은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 시작 비트 인덱스의 분포에 대한 일 예를 도시한다.
도 12은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 시작 비트 인덱스의 분포에 대한 다른 예를 도시한다.
도 13은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 시작 비트 인덱스의 분포에 대한 또 다른 예를 도시한다.
도 14는 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 단축된 비트가 존재하는 경우 종료 비트 인덱스 분포에 대한 예를 도시한다.
도 15는 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 복호기의 구성을 도시한다.
도 16 및 17은 본 개시의 다양한 실시 예들에 따른 복호 기법의 성능을 도시한다.
본 개시에서 사용되는 용어들은 단지 특정한 실시 예를 설명하기 위해 사용된 것으로, 다른 실시 예의 범위를 한정하려는 의도가 아닐 수 있다. 단수의 표현은 문맥상 명백하게 다르게 뜻하지 않는 한, 복수의 표현을 포함할 수 있다. 기술적이거나 과학적인 용어를 포함해서 여기서 사용되는 용어들은 본 개시에 기재된 기술 분야에서 통상의 지식을 가진 자에 의해 일반적으로 이해되는 것과 동일한 의미를 가질 수 있다. 본 개시에 사용된 용어들 중 일반적인 사전에 정의된 용어들은, 관련 기술의 문맥상 가지는 의미와 동일 또는 유사한 의미로 해석될 수 있으며, 본 개시에서 명백하게 정의되지 않는 한, 이상적이거나 과도하게 형식적인 의미로 해석되지 않는다. 경우에 따라서, 본 개시에서 정의된 용어일지라도 본 개시의 실시 예들을 배제하도록 해석될 수 없다.
이하에서 설명되는 본 개시의 다양한 실시 예들에서는 하드웨어적인 접근 방법을 예시로서 설명한다. 하지만, 본 개시의 다양한 실시 예들에서는 하드웨어와 소프트웨어를 모두 사용하는 기술을 포함하고 있으므로, 본 개시의 다양한 실시 예들이 소프트웨어 기반의 접근 방법을 제외하는 것은 아니다.
이하 본 개시는 통신 시스템에서 극부호(polar code)를 이용한 복호화를 위한 장치 및 방법에 관한 것이다. 구체적으로, 본 개시는 통신 시스템에서 극부호에 기반한 복호 동작의 복잡도를 낮추기 위한 기술을 설명한다. 보다 상세하게, 본 개시는 데이터를 전송하거나, 저장하는 과정에서 잡음, 간섭 등의 다양한 원인으로 인해 오류 및 소실이 발생하거나 발생할 여지가 있을 때, 이를 정정, 복원하는 오류-정정 부호(error-correcting codes)에 관한 것이다. 다시 말해, 본 개시는 극부호를 사용하는 통신 시스템에서 수신기가 수행하는 복호화와 관련된 것으로서, 극부호로 부호화된 부호어를 낮은 복잡도와 지연으로 복호화하는 기술에 대한 것이다.
이하 설명에서 사용되는 파라미터들을 지칭하는 용어, 리던던시(redanduncy) 비트를 지칭하는 용어(예: 패리티 체크 비트), 정보 비트를 지칭하는 용어, 채널을 지칭하는 용어, 제어 정보를 지칭하는 용어, 네트워크 객체(network entity)들을 지칭하는 용어, 장치의 구성 요소를 지칭하는 용어 등은 설명의 편의를 위해 예시된 것이다. 따라서, 본 개시가 후술되는 용어들에 한정되는 것은 아니며, 동등한 기술적 의미를 가지는 다른 용어가 사용될 수 있다.
또한, 본 개시는, 일부 통신 규격(예: 3GPP(3rd Generation Partnership Project))에서 사용되는 용어들을 이용하여 다양한 실시 예들을 설명하지만, 이는 설명을 위한 예시일 뿐이다. 본 개시의 다양한 실시 예들은, 다른 통신 시스템에서도, 용이하게 변형되어 적용될 수 있다.
극부호는 2008년 E. Arikan 에 의해 제안된 오류 정정 부호로 낮은 부호화/복잡도 성능을 가지면서도 모든 B-DMCs(binary discrete memoryless channel) 에서 데이터 전송 한계인 채널 용량(channel capacity)을 달성하는 것이 증명된 최초의 오류 정정 부호이다. 극부호는 다른 채널 용량 근접 부호(capacity-approaching codes)인 터보 코드(turbo code), LDPC(low-density parity-check) 코드 대비 짧은 길이의 부호를 전송할 때 오류-정정 성능 및 복호 복잡도 상 이점이 있다. 이러한 장점으로 인해 2017년 현재 5세대 이동통신 를 위해 진행 중인 3GPP NR 표준화에서 짧은 길이의 제어 정보를 전송하는 용도로 극부호의 사용을 결정하였다.
본 개시는 통신 시스템에 사용되는 극부호의 부호화 및 복호화 시스템을 고려한다. 기본적인 극부호의 복호 방법은 순차적 제거(SC) 복호이다. SC 복호기는 부호화 입력 비트 시퀀스를 한 비트씩 순차적으로 복호해가며, 어떤 비트를 복호함에 있어서 그 이전까지 복호된 비트 결과를 이용한다. 성능 향상을 위해 사용되는 SCL(SC-list) 복호, SCS(SC-stack) 복호, SCF(SC-flip) 복호 등도 기본적으로는 순차적인 복호 방식을 사용한다. 이러한 순차적인 복호 동작의 특성으로 인해 SC 기반의 복호기들(SC-based decoders)은 병렬화 구조(parallelized architecture), 파이프라인 구조(pipeline architecture) 로 구현하는 것이 적합하지 않으며, 병렬 프로세싱에 적합한 Turbo 부호와 LDPC 부호의 복호에 필요한 클락 사이클(clock cycle) 이 더 크다는 단점이 있다. 즉, polar code의 SC 기반 복호기는 복호 처리율(decoding throughput) 은 일반적으로 낮다.
또한 성능 향상을 위해 사용되는 SCL 복호기의 경우 복잡도가 높다는 문제점이 있다. SCL 복호기는 사전에 설정된 리스트(list) 수만큼의 후보군을 유지해가며 각 후보군에 대해서 독립적인 복호를 진행한다. 매 비트의 복호 과정에서 각 리스트에 대한 복호는 독립적으로 이루어지며, 복호 결과를 바탕으로 확률이 높은 리스트를 유지하고 낮은 리스트를 제외한다. 따라서 SCL 복호기가 수행하는 연산의 수와 시간은 리스트의 수에 비례하여 증가하게 된다.
본 발명은 다양한 부호 파라미터(부호 길이, 부호율 등) 와 설정을 지원하는 극부호 시스템의 SC, SCL 복호기 등의 복호 복잡도 및 지연을 감소시키는 것을 목적으로 한다. 본 발명이 적용되는 일 실시 예로는 3GPP NR 제어 신호 전송을 위해 사용되는 극부호 시스템이 있다. 3GPP NR 극부호 시스템은 길이 12 부터 1706 까지의 비트를 부호화하며, 다양한 부호율을 지원한다. 본 발명은 이와 같이 다양한 부호 길이와 부호율을 지원하는 극부호 시스템에서 사용되는 SC, SCL 복호기의 복잡도 및 지연을 효과적으로 감소시키는 것을 목적으로 한다. 또한 본 발명은 상기 목적을 달성을 위해 수반하게 되는 추가적인 하드웨어, 연산량, 연산 시간 등의 비용(cost) 을 최소화하는 것을 목적으로 한다. 일례로, 종래에는 부호 파라미터에 따른 polar code의 구조를 사전에 분석하고, 분석된 결과에 맞춰 해당 부호 전용의(dedicated) 간략화된(simplified) SCL 복호기를 구성하는 방법이 제안된 바 있다. 이와 같이 부호 구조를 사전에 분석하고 전용의 연산 블록을 추가하는 방법은 부호 구조가 고정되어 있거나(static) 사용하는 부호 파라미터의 종류가 적은 경우에는 효과적일 수 있다. 하지만 만약 지원해야 하는 부호 파라미터의 수가 많은 경우(가령 3GPP NR 과 같이 다양한 부호 길이와 부호율을 지원하는 경우), 상기 간략화된 SCL 복호기를 구성하는 것이 비효율적일 수 있다. 주어지는 부호 파라미터를 기반으로 간략화된 SCL 복호기를 구성하기 위해 추가되는 부호 분석 과정과 연산은 복호기의 간략화라는 본래의 목적과 다르게 전반적인 연산량과 연산 시간을 오히려 증가시킬 수 있다. 본 발명은 상기와 같은 종래 기술이 다양한 파라미터를 지원하는 통신 시스템에 적합하지 않거나 효율적이지 않은 문제를 해결한다. 즉, 본 발명은 다양한 부호 파라미터와 부호 구조를 지원함에 있어서, 추가로 실시하게 되는 SCL 본래 복호 동작 외의 연산과 이에 의해 유발되는 지연을 최소화하는 것을 주 목적으로 한다.
도 1은 본 개시의 다양한 실시 예들에 따른 무선 통신 시스템을 도시한다. 도 1은 무선 통신 시스템에서 무선 채널을 이용하는 노드(node)들의 일부로서, 송신단(transmitting device)(110)과 수신단(receiving device)(120)를 예시한다. 도 1은 하나의 송신단(110) 및 하나의 수신단(120)를 도시하나, 복수의 송신단 또는 복수의 수신단을 포함할 수 있다. 또한, 설명의 편의를 위하여, 본 개시에서 송신단(110)과 수신단(120)이 별개의 개체인 것으로 설명하나, 송신단(110)과 수신단(120)의 기능은 서로 바뀔 수 있다. 예를 들어, 셀룰러 통신 시스템의 상향링크의 경우, 송신단(110)는 단말, 수신단(120)는 기지국이 될 수 있다. 하향링크의 경우, 송신단(110)는 기지국, 수신단(120)이 단말이 될 수 있다. 일부 실시 예들에서, 송신단(110)는, 정보 비트들의 극부호를 이용하여 부호화를 수행한 후, 송신할 수 있다. 수신단(120)는 송신단(110)에서 송신된 부호화된 정보를 극부호를 이용하여 복호함으로써 정보 비트들을 복원할 수 있다.
도 2는 본 개시의 다양한 실시 예들에 따른 무선 통신 시스템에서 수신단 장치의 구성의 예를 도시한다. 즉, 도(2에 예시된 구성은 수신단(120)의 구성으로서 이해될 수 있다. 이하 사용되는 '...부', '...기' 등의 용어는 적어도 하나의 기능이나 동작을 처리하는 단위를 의미하며, 이는 하드웨어나 소프트웨어, 또는, 하드웨어 및 소프트웨어의 결합으로 구현될 수 있다.
도 2를 참고하면, 수신단(120)는 통신부(210), 저장부(220), 제어부(230)를 포함할 수 있다.
통신부(210)는 무선 채널을 통해 신호를 송수신하기 위한 기능들을 수행할 수 있다. 예를 들어, 통신부(210)는 시스템의 물리 계층 규격에 따라 기저대역 신호 및 비트열 간 변환 기능을 수행할 수 있다. 예를 들어, 데이터 송신 시, 통신부(210)는 송신 비트열을 부호화 및 변조함으로써 복소 심벌들을 생성할 수 있다. 또한, 데이터 수신 시, 통신부(210)는 기저대역 신호를 복조 및 복호화를 통해 수신 비트열을 복원할 수 있다. 또한, 통신부(210)는 기저대역 신호를 RF(radio frequency) 대역 신호로 상향변환한 후 안테나를 통해 송신하고, 안테나를 통해 수신되는 RF 대역 신호를 기저대역 신호로 하향변환할 수 있다.
이를 위해, 통신부(210)는 송신 필터, 수신 필터, 증폭기, 믹서(mixer), 오실레이터(oscillator), DAC(digital to analog convertor), ADC(analog to digital convertor) 등을 포함할 수 있다. 또한, 통신부(210)는 다수의 송수신 경로(path)들을 포함할 수 있다. 나아가, 통신부(210)는 다수의 안테나 요소들(antenna elements)로 구성된 적어도 하나의 안테나 어레이(antenna array)를 포함할 수 있다. 하드웨어의 측면에서, 통신부(310)는 디지털 유닛(digital unit) 및 아날로그 유닛(analog unit)으로 구성될 수 있으며, 아날로그 유닛은 동작 전력, 동작 주파수 등에 따라 다수의 서브 유닛(sub-unit)들로 구성될 수 있다.
통신부(210)는 상술한 바와 같이 신호를 송신 및 수신한다. 이에 따라, 통신부(210)는 '송신부', '수신부' 또는 '송수신부(transceiver)'로 지칭될 수 있다. 또한, 이하 설명에서, 무선 채널을 통해 수행되는 송신 및 수신은 통신부(210)에 의해 상술한 바와 같은 처리가 수행되는 것을 포함하는 의미로 사용된다. 또한, 통신부(210)는 백홀(backhaul) 망을 통해 연결된 다른 네트워크 개체와의 통신을 위한 백홀 통신부를 포함할 수 있다.
통신부(210)는 본 개시의 다양한 실시 예들에 따른 복호화를 수행하기 위해 복호화기(decoder)(미도시)를 포함할 수 있다. 복호화기는 극부호에 기반하여 복호 동작을 수행할 수 있다.
저장부(220)는 수신단(120)의 동작을 위한 기본 프로그램, 응용 프로그램, 설정 정보 등의 데이터를 저장할 수 있다. 저장부(220)는 휘발성 메모리, 비휘발성 메모리 또는 휘발성 메모리와 비휘발성 메모리의 조합으로 구성될 수 있다. 그리고, 저장부(220)는 제어부(230)의 요청에 따라 저장된 데이터를 제공할 수 있다.
제어부(230)는 수신단(120)의 전반적인 동작들을 제어할 수 있다. 예를 들어, 제어부(230)는 통신부(220)를 통해 신호를 송신 및 수신할 수 있다. 또한, 제어부(230)는 저장부(230)에 데이터를 기록하거나 읽을 수 있다. 이를 위해, 제어부(230)는 적어도 하나의 프로세서 또는 마이크로(micro) 프로세서를 포함하거나, 또는, 프로세서의 일부일 수 있다. 다시 말해, 제어부(230)는 통신부(220)에 포함된 각 구성들의 동작을 제어할 수 있다. 예를 들어, 제어부(230)는 수신단(120)이 후술하는 다양한 실시 예들에 따른 동작들을 수행하도록 제어할 수 있다.
도 3은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 극부호의 부호화 과정의 일 예를 나타낸다. 도 3에 예시된 동작들 중, 시스템의 요구 조건 및 특징에 따라 과정 중 일부가 생략될 수 있고, 혹은 별도의 동작이 추가될 수 있다. 도 3에서, 송신단이 전송하고자 하는 정보 비트(information bit) 개수는 A, 부호화 후 채널을 통해 전송되는 부호어 비트(codeword bit) 개수는 E라 표현된다. 송신단에서 수행되는 부호화의 각 단계는 다음과 같다.
1) 정보 비트 시퀀스 생성(Information bit sequence generation)
전송하고자 하는 길이 A 의 정보 비트 시퀀스 b={b0,b1,...,bA-1}가 주어진다. 정보 비트 시퀀스는 전송하고자 하는 전체 정보의 일부, 즉 세그먼트(segment)일 수 있다.
2) 외부 부호화(Outer code)
연접 외부 인코더(concatenated outer endoer)(302)에서, 정보 비트 시퀀스 b는 성능 향상을 위해 먼저 외부 부호(concatenated outer code)로 부호화된다. 이러한 외부 부호는 보통 극부호의 SCL(SC-list) 복호와 같이 다수의 부호어 후보군들을 고려하여 복호를 진행하는 복호기의 성능을 높이기 위해 사용된다. SCL 복호는 이하 수신단 동작에서 자세하게 설명하기로 한다. 이러한 목적으로 극부호 에 연접되어 사용되는 외부 부호로는 CRC(cyclic redundancy check) 부호와 같은 오류 검출 부호나 BCH 부호, PC(parity check), 컨볼루션(convolutional) 부호 등 오류 정정 부호가 있다. 외부 부호는 하나만 사용될 수 있고, 혹은 둘 이상의 외부 부호가 복합적으로 사용될 수도 있다. 예를 들어, 3GPP NR에 정의된 극부호 중 상향링크(uplink) 제어 정보를 위해 사용되는 극부호는 정보 비트의 길이에 따라 정보 비트 길이가 20비트 이상이면 11 비트 CRC 부호와 함께 사용되고, 정보 비트 길이가 19비트 이하이면 3비트 PC 부호와 6비트 CRC 부호와 함께 사용된다. 이러한 외부 부호화는 보통 시스테메틱(systematic) 부호로서, 입력된 비트 시퀀스에 패리티 비트를 추가한다. 하나 이상의 외부 부호에 의해 생성된 전체 패리티 비트의 길이를 B 이라고 하고, 외부 부호화에 의해 생성된 부호어의 길이를 K=A+B 이라고 한다. 그리고 외부 부호화의 결과로 생성된 비트 시퀀스는
Figure PCTKR2020018557-appb-I000001
라고 한다. 외부 부호화는 극 부호화(polar coding) 자체에 필수적인 동작은 아니기 때문에, 외부 부호화는 생략될 수 있고, 이 경우 B=0 이고 b'=b 이다.
3) 부채널 할당(Sub-channel allocation)
부채널 할당기(304)에서, 비트 시퀀스 b'는 극부호 부호화를 위해 길이 N의 비트 시퀀스 u={u0,u1,...,uN-1}에 매핑된다. 여기서, N은 모부호(mother polar code)의 크기로 2의 거듭 제곱수이며, K 보다는 큰 값 중 사전에 결정된 기준에 의해 결정된다. 비트 시퀀스 u는 극부호 부호화기의 입력 비트 시퀀스(encoder input bit sequence)이며, 사전에 정해진 방법 및 기준에 따라 b'의 비트가 u에 매핑된다(304). 부호화 입력 비트 시퀀스 u의 각 비트는 이후 송신단과 수신단의 동작에 의한 채널 분화(channel polarization)에 의해 서로 다른 품질의 가상 채널인 부채널(split channel, sub-channel)을 통과하는 것처럼 해석될 수 있다. 이때, 각 부채널을 합성 채널(synthetic channel)이라고 일컫기도 한다. 따라서, 상기 b'의 비트들을 우수한 품질의 부채널을 통과하는 u의 비트들에 매핑하기 위해서 각 부채널의 채널 용량(symmetric capacity), Bhatacharayya 파라미터, 밀도 진화(density evolution)의 결과 등을 이용한다. 또한, 이 과정에서는 이후 수행되는 부호율-조정(rate-matching) 동작이 고려된다. 이러한 특징 때문에, b'를 u에 매핑하는 과정은 부채널 할당 과정으로도 지칭된다. 이 과정에서, b'가 매핑되는 부채널에 해당하는 u의 비트는 비-동결(unfrozen) 비트라고 지칭되며, 나머지 부채널에 해당하는 u의 비트는 동결(frozen) 비트라 지칭된다. 명칭과 같이, 동결 비트는 그 값이 고정되는데, 일반적으로 그 값은 0 이다.
4) 생성 행렬 곱셈(Generator Matrix Multiplication)
생성 행렬 곱셈부(306)에서, 길이 N의 부호화 입력 비트 시퀀스 u는 극부호의 생성 행렬(generator matrix) G와 곱해지며, 이에 따라 동일한 길이 N의 부호화 출력 비트 시퀀스 x가 생성된다. 최초 Arikan에 의해 극부호가 제안되었을 때, 생성 행렬 G는 [수학식 1] 과 같이 정의되었다.
Figure PCTKR2020018557-appb-I000002
위의 식에서
Figure PCTKR2020018557-appb-I000003
이며, 위 첨자
Figure PCTKR2020018557-appb-I000004
연산은 n회의 크로네커 곱(Kronecker power)을 의미한다. 예를 들어,
Figure PCTKR2020018557-appb-I000005
이며,
Figure PCTKR2020018557-appb-I000006
이다. 그리고, BN은 크기 N×N 비트-역전 퍼뮤테이션(bit-reversal permutation) 행렬이다. 예를 들어 길이 8인 벡터{a0,a1,a2,a3,a4,a5,a6,a7} 와 8×8 비트-역전 퍼뮤테이션(bit-reversal permutation) 행렬 B8이 곱해져 인덱스가 비트-역전 퍼뮤테이션(bit-reversal permutation) 된 {a0,a4,a2,a6,a1,a5,a3,a7} 가 얻어진다. 하지만, 다른 실시 예에 따라, BN을 제외한 단순한 형태의 [수학식 2]와 같은 생성 행렬이 사용될 수 있다.
Figure PCTKR2020018557-appb-I000007
이하 설명에서, 별도의 언급이 없으면
Figure PCTKR2020018557-appb-I000008
으로 정의된 생성 행렬의 사용이 가정된다. 이와 같은 가정으로 설명된 내용은 비트-역전 퍼뮤테이션(bit-reversal permutation) 동작을 바탕으로
Figure PCTKR2020018557-appb-I000009
으로 정의된 생성 행렬을 사용한 극부호로 쉽게 변경되어 설명될 수 있음에 유의하여야 한다.
5) 부호율-조정(Rate-matching)
부호율 조정부(308)에서, 생성된 부호화 출력 비트 시퀀스 x로부터 전송하고자 하는 길이 E의 비트 시퀀스를 생성하는 동작이 수행된다. 이와 같이 부호율 조정을 통해 얻어지는 전송 비트 시퀀스는 c로 표현된다. 부호율-조정에 따라, 극부호의 성능 향상을 위해 부호화 출력 비트 시퀀스 x는 재조정 수 있다. 예를 들어, 3GPP NR 극 부호화 시스템에서, 부호화 출력 비트 시퀀스 x는 32개의 서브-블록(sub-block) 단위로 인터리빙되고, 순환 버퍼(circular buffer)에 저장되고, 순차적으로 추출됨으로써 길이 E의 부호어 시퀀스가 생성된다. 만약 부호어의 길이 E가 극부호 모부호의 크기 E보다 작다면, 천공(puncturing) 혹은 단축(shortening) 중 하나의 동작이 수행될 수 있다. 부호화 출력 비트 시퀀스 x의 일부 비트가 천공되면, 부호화 입력 비트 시퀀스 u가 겪는 부채널의 일부가 불능(incapable)이 되며, 부채널 할당 과정은 이러한 불능 비트(incapable bit)를 고려하여 이뤄진다. 만약 부호화 출력 비트 시퀀스 x의 일부 비트를 단축하고자 하는 경우, 부호화 입력 비트 시퀀스 u의 일부 비트 또한 단축되어야 하며, 부채널 할당 과정은 이러한 단축 비트(shortening bit)를 고려하여 수행된다. 반면, 부호어의 길이 E가 극부호 모부호의 크기 E보다 크다면, 반복(repetition)이 수행될 수 있다.
도 4는 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 극부호의 복호화 과정의 일 예를 나타낸다. 도 4에 예시된 동작들 중 시스템의 요구 조건 등에 따라 다른 동작이 추가될 수도 있고 혹은 포함된 동작이 생략될 수도 있다.
1) 복조 LLR 생성(Demodulated LLR generation)
수신된 신호를 복조(demodulation) 하여 전송된 비트 c에 대응되는 확률 정보가 생성된다. 확률 정보는 확률 벡터, LR(likelihood ratir), LLR(log-likelihood ratio) 등의 값으로 주어진다. 간결한 설명을 위해 아래에서는 별도의 언급이 없다면 LLR을 고려한다. 하지만, 본 발명은 LLR 기반의 복호 동작에 한정되지 아니한다. 전송 비트 시퀀스 c에 대응되는 LLR 시퀀스는 l={l0,l1,...,lE-1}라고 지칭된다.
2) 부호율-역조정(Rate-dematching)
부호율 역조정부(402)에서, 길이 E의 LLR 시퀀스 l는 길이 N의 극부호 복호화기(404)에 입력되도록 송신단의 부호율-조정의 역과정을 거친다. 송신단의 부호율 조정단에서 천공(puncturing) 이 발생했다면 해당 비트에 대한 LLR 값은 0으로 결정되며, 단축(shortening) 이 발생했다면 해당 비트에 대한 LLR 값은 비트 값 0에 대응하는 LLR 값의 최대값으로 결정된다. 만약 특정 비트에 대해서 반복이 발생했다면 대응되는 LLR 값을 모두 더해서(combining) 해당 비트에 대한 LLR 값이 결정된다. 이와 같은 과정을 통해서 결정된 길이 N의 LLR 시퀀스를 l'={l'0,l'1,...,l'N-1}로 지칭된다.
3) 외부 부호화를 이용한 SC 복호(Outer Code aided SC-based decoding)
극부호 복호화기(404)에서, 길이 N 의 LLR 시퀀스 l'가 계산 혹은 결정되면 이를 바탕으로 SC-기반의 복호 동작이 수행된다. SC-기반의 복호로는 일반적인 SC 복호, SCL(SC-list) 복호, SCS(SC-stack) 복호 등이 있다. SC-기반의 복호는 부호화 입력 시퀀스의 각 비트를 인덱스 순서에 따라 한 비트씩 순차적으로 복호한다. 이러한 복호기들은 부호화 입력 비트 시퀀스의 인덱스 값 순서로, 즉, u0,u1,...,uN-1 순서로 각 비트에 대한 복호를 수행한다. 구체적으로 i 번째 비트, ui 에 대한 복호는 아래의 절차에 의해 이뤄진다.
① 수신 신호와 앞서 복호된 비트들의 추정값
Figure PCTKR2020018557-appb-I000010
을 기반으로 ui 값의 0과 1에 대한 확률 기반 지표(metric) 계산한다.
② 계산한 확률 기반 지표를 기반으로 비트 값
Figure PCTKR2020018557-appb-I000011
추정(estimation) 한다.
③ 다음 비트의 복호를 위해 추정값
Figure PCTKR2020018557-appb-I000012
를 순차적 제거(successive cancellation) 기반으로 복호기에 반영한다.
상술한 바와 같이 각 비트에 대한 복호는 이전까지 복호가 되어 추정된 비트 값을 기반으로 이뤄진다. 가령, 비트 ui의 복호 시에는 이전에 복호된 비트(u0,…,ui-1) 에 대한 추정 값인
Figure PCTKR2020018557-appb-I000013
과 그에 대한 확률 정보 혹은 확률 정보에 준하는 값의 누적 값 등이 이용된다. 여기서, 각 부분적인 비트 시퀀스
Figure PCTKR2020018557-appb-I000014
는 리스트(list) 혹은 경로(path)라 지칭되며, 각 경로에 대한 복호를 수행해오면서 계산한 누적된 확률 정보 혹은 확률 정보에 준하는 값은 PM(path-metric)이라 지칭된다. SCL 복호기는 비트 ui의 복호 시 설정된 리스트 크기 L개만큼의 리스트
Figure PCTKR2020018557-appb-I000015
를 유지해가며 복호를 진행하는 방식이다. 가령, SCL 복호기는 지금까지 복호에서 유지하고 있는 리스트
Figure PCTKR2020018557-appb-I000016
를 바탕으로 ui 값 0과 1에 대한 확률 정보를 계산한다. 그리고, 각 ui의 확률 정보 계산 시 고려한 각 리스트
Figure PCTKR2020018557-appb-I000017
에 대한 PM 에 ui 비트 값에 0, 1에 대한 확률 정보 혹은 확률 정보에 준하는 값을 업데이트하며, 총 2L개의 리스트
Figure PCTKR2020018557-appb-I000018
에 대한 PM 을 계산한다. 만약, ui가 동결 비트(frozen bit)인 경우, 계산된 PM과 상관없이 송/수신단이 서로 약속한 결정된 비트값으로 추정값
Figure PCTKR2020018557-appb-I000019
이 결정된다. 만약, ui가 비동결 비트(unfrozen bit) 중 정보 비트(information bit)에 해당하면, 총 2L개의 리스트 중에 PM 값을 바탕으로 확률이 높다고 판단한 L개의 리스트
Figure PCTKR2020018557-appb-I000020
가 선택된다. SCL 복호기는 이와 같은 방식으로 각 비트의 복호마다 총 L개의 리스트를 유지하며 복호를 진행한다. 그리고, 모든 비트에 대한 복호가 완료되면, 최종적으로 얻은 L개의 리스트들, 즉 후보 비트 시퀀스들 중에 가장 확률이 높은 부호어가 PM 기반으로 선택된다. 만약 L이 1로 설정되어 있으면 SCL 복호기는 기본적인 SC 복호기와 동일하게 동작한다. SCL 복호 중간 혹은 완료 후, 연접된 외부 부호화를 활용하여 오류 정정 성능을 향상시킬 수 있다. 가령 CRC 부호가 연접되어 사용된 경우, 복호 후 얻은 L개의 리스트 중에서 CRC 부호의 제약 조건을 만족하면서 가장 확률이 확률이 높은 부호어를 최종 복호 결과로 추정한다.
4) 메시지 비트 추출(Message bit extraction)
메시지 비트 추출부(405)에서, 추정된 부호화 입력 비트 시퀀스
Figure PCTKR2020018557-appb-I000021
로부터 정해진 위치의 메시지 비트가 획득된다. 최종적으로 추정된 메시지 비트 시퀀스는
Figure PCTKR2020018557-appb-I000022
로 표기된다.
도 3을 참고하여 설명된 부호화 과정 중 생성 행렬 곱셈 과정과 도 4를 참고하여 설명된 복호화의 과정 중 SC 기반의 복호 과정은 극부호에 대응하는 그래프를 이용하여 설명될 수 있다.
도 5는 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 극부호에 대응하는 그래프의 일 예를 도시한다. 도 5는 크기가 8인 극부호에 대한 이진 그래프(bipartite graph)예시한다.
도 5를 참고하면, 이진 그래프는 원으로 표시되는 변수 노드들(variable nodes)(511), 정사각형으로 표시되는 검사 노드들(check nodes)(512), 그리고 이 노드들을 연결하는 연결선들(edges)(513)을 포함한다. 부호화 과정에서 변수 노드는 개개의 비트를 나타내며, 복호화 과정에서 각 변수 노드에서는 해당 비트에 대한 LLR 등의 확률 정보를 나타낸다. 검사 노드는 연결된 이웃 변수 노드들의 선형 제약(linear constraint)을 나타내며, 하나의 검사 노드에 이웃한 변수 노드에 대응되는 모든 비트의 이진-합(modulo-2 sum)은 0이다. 길이 N의 극부호 그래프는 log2N+1 개의 단계(stage)로 구성될 수 있다. 기호 t가 그래프에서 단계의 인덱스를 나타내며, 그래프의 가장 왼쪽의 변수 노드들에 대한 단계 인덱스는 t=0, 그래프의 가장 오른쪽의 변수 노드들에 대한 단계 인덱스는 t=log2N 이다. 극부호 그래프의 특징들 중 하나는 전체 그래프가 Z 모양을 갖는 2×2 기본적인 요소(531)들의 연결들로 구성된다는 것이다. 전체 극부호 그래프에서 기본적인 요소들 사이의 연결은 [수학식 2]의 생성 행렬로부터 결정된다.
도 3을 참고하여 설명된 부호화 과정 중 생성 행렬 곱셈 과정은 도 5의 그래프 상에서 이해될 수 있다. 도 5의 그래프 상에서 가장 왼쪽 N개의 변수 노드들(521)은 도 3의 부호화 과정의 길이 N의 부호화 입력 비트 시퀀스 u에 대응하며, 가장 오른쪽 N개의 변수 노드들(522)은 길이 N의 부호화 출력 비트 시퀀스 x에 대응한다. 부호화 과정은 가장 왼쪽에 있는 부호화 입력 비트 시퀀스 u로부터 가장 오른쪽에 있는 부호화 출력 비트 시퀀스 x를 검사 노드의 관계식을 만족하도록 계산하는 과정을 나타낸다. 즉, 부호화 과정은 도 5에 예시된 그래프의 왼쪽으로부터 오른쪽으로 연결된 변수 노드들과 검사 노드의 관계에 따라 비트값을 단계적으로 업데이트하는 동작으로 이해될 수 있다.
도 4를 참고하여 설명된 복호화 과정 중 SC 기반의 복호 과정 또한 도 5의 그래프 상에서 신뢰-전파(belief-propagation)로 이해될 수 있다. 복호를 위한 첫 단계로, 부호율-역조정된 N개의 내재적(intrinsic) LLR 등 소프트 값(soft-value) 확률 정보가 그래프의 가장 오른쪽의 N개의 변수 노드들(521)에 입력된다. 여기서, 간결한 설명을 위해 소프트 값(soft-value) 확률 정보를 LLR로 가정하나, 이러한 가정이 본 발명의 적용 범위를 제한하지는 아니한다. 상기와 같이 내재적(intrinsic) LLR 을 설정한 뒤, 정해진 절차에 따라 SC 기반의 복호가 수행된다. SC 기반 복호의 가장 중요한 특징은 그래프의 가장 왼쪽 N개의 변수 노드에 해당하는 부호화 입력 비트 시퀀스 u=(u0,u1,...,uN-1) 가 비트 인덱스에 따라 한 비트씩 순차적으로 복호된다는 점이다. 부호화 입력 비트 시퀀스 u의 각 비트에 대한 순차적인 복호는 해당 비트에 대한 사후 LLR(a-posteriori LLR, AP-LLR)을 계산하고, 계산된 LLR을 바탕으로 비트 값을 추정하는 과정이다. 각 비트의 AP-LLR은 그래프의 가장 오른쪽 편에 주어진 내재적(intrinsic) LLR 값으로부터 계산되는데, 이 과정은 그래프 오른편에서 왼편으로, 각 단계(stage) 별 변수 노드에 대한 LLR 값을 순차적으로 계산하는 것으로 이루어진다. 즉, 초기 내재적(intrinsic) LLR은 t=log2N 단계에 주어지며, 이로부터 t=log2N-1, log2N-2, ... 순서로 각 단계의 일부 변수 노드에 대한 LLR 값이 계산된다. 이러한 과정을 통해 t=0 단계의 각 변수 노드에 대한 LLR 값, 즉, 하나의 부호화 입력 비트에 대한 AP-LLR 이 계산된다.
극부호 그래프 상에서의 상세한 LLR 계산 과정은 다음과 같다. 모든 LLR 값의 계산은 그래프를 구성하는 Z 모양의 기본 요소(531) 상에서 이루어진다. 도 6는 본 개시의 다양한 실시 예에 따른 통신 시스템에서 LR 계산 과정을 도시한다. Z 모양의 왼쪽 두 변수 노드에 해당하는 비트 a=(a1,a2) 와 오른쪽 두 변수 노드에 해당하는 비트 b=(b1,b2)의 2×2 polar coding을 묘사하며, 이를 수학식으로 표현하면 아래와 같다.
Figure PCTKR2020018557-appb-I000023
비트 a1,a2,b1,b2에 대응하는 LLR 값을 각각
Figure PCTKR2020018557-appb-I000024
라하면, 각 기본 요소에 대한 복호에서 오른쪽 두 변수 노드에 해당하는
Figure PCTKR2020018557-appb-I000025
Figure PCTKR2020018557-appb-I000026
가 주어지며, 단계에 따라
Figure PCTKR2020018557-appb-I000027
Figure PCTKR2020018557-appb-I000028
중 하나의 LLR 값이 계산된다. 만약 a1,a2 중 어느 하나의 비트 값도 현재 추정되지 아니하면, 도 6의 좌측(610)과 같이 이하 [수학식 4]의 연산을 통해
Figure PCTKR2020018557-appb-I000029
이 계산될 수 있다.
Figure PCTKR2020018557-appb-I000030
위의 식은 비트들의 관계에 따라 정확하게 LLR 값을 계산하는 방식이며, 구현을 위해 이하 [수학식 5]와 같이 근사화될 수 있다.
Figure PCTKR2020018557-appb-I000031
[수학식 5]에서, 함수 sgn(x)는 x의 부호(sign)를 출력하는 함수로, x의 값이 0보다 크면 1을, 0보다 작으면 -1을 출력한다. [수학식 4] 혹은 [수학식 5]와 같이 a1의 LLR을 계산하는 함수는 f 함수라 지칭된다. 복호의 진행에 따라 비트 a1의 추정 값
Figure PCTKR2020018557-appb-I000032
이 얻어지면 도 6의 우측(620)과 같이 이하 [수학식 6]을 통해
Figure PCTKR2020018557-appb-I000033
를 계산한다.
Figure PCTKR2020018557-appb-I000034
[수학식 6]과 같이 a2의 LLR 을 계산하는 함수는 g 함수라 지칭되 수 있다. 앞서 기술한 바와 같이, SC 기반의 복호에서 비트 인덱스에 따라 순차적으로 비트의 복호 및 추정이 이루어지며, 따라서 a2에 대한 복호는 a1의 복호 후에 수행된다. 상술한 바와 같이 계산된
Figure PCTKR2020018557-appb-I000035
Figure PCTKR2020018557-appb-I000036
는 그래프 상에서 왼편으로 연결된 다른 Z 모양의 기본 요소로 전달된다. 이와 같은 방식으로 그래프의 가장 왼편의 변수 노드, 즉, 단계 t=0 의 부호화 입력 비트의 LLR 값이 얻어지면, 이를 바탕으로 비트 값이 추정된다. 만약 해당 비트가 동결 비트이면 LLR 값에 상관없이 송신단과 수신단이 서로 약속한 값이 대입되며, 비동결 비트이면 업데이트된 LLR 값을 경판정(hard-decision)한 값이 대입된다. 추정된 비트는 그래프의 왼쪽으로부터 오른쪽으로 현재 복호 단계에서 가능한 범위의 변수 노드로 전달되어, [수학식 6]의 과정과 같이 다른 변수 노드의 LLR 계산에 이용된다.
도 7은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 복호 이진 트리(decoding binary tree)의 일 예를 도시한다. 도 7은 앞서 설명한 SC 기반의 복호기가 수행하는 연산의 종류와 절차를 효율적으로 설명하기 위해 널리 활용되는 복호 이진 트리를 예시한 것으로, 크기 8의 극부호에 대한 복호에 대한 예이다. 이 이진 트리는 SC 복호기가 수행하는 연산의 종류를 확인하고, 각 연산이 어떠한 순서로 이루어지는지를 이해하는데 유용하다. 이진 트리에서 깊이(depth)는 도 5의 극부호 그래프의 단계(stage)와 대응한다. 이진 트리의 가장 아래쪽 리프(leaf) 노드는 단계 t=0 의 변수 노드에 대응하며, 루트(root) 노드는 단계 t=log2N의 변수 노드들에 대응한다. 따라서, 이진 트리의 깊이를 d 라고 할 때, 대응하는 극부호 그래프의 단계와의 관계는 d=log2N-t가 된다. 이진 트리에서 깊이 d의 노드들은 극부호 그래프의 단계 t=log2N-d 의 변수 노드들에 대응한다. 이진 트리에서 깊이 d의 한 노드가 나타내는 극부호 그래프의 단계 t=log2N-d 의 변수 노드들의 수는 N/2d 개가 된다. 따라서 깊이 d=0 의 루트 노드는 극부호 그래프의 단계 t=log2N 의 모든 N개의 변수 노드들에 대응한다. 또한, 깊이 d=log2N 의 리프 노드는 도 7에 도시된 것과 같이 극부호 그래프의 단계 t=0 의
Figure PCTKR2020018557-appb-I000037
개의 변수 노드에 대응된다. 즉, 이진 트리에서 i 번째 리프 노드는 i 번째 부호화 입력 비트 ui를 나타낸다.
SC 기반의 복호 절차는 이진 트리에서 루트 노트로부터 각 리프 노드를 깊이-우선 탐색(depth-first search, DFS) 방식으로 방문(visit) 또는 탐색(search)하는 것으로 이해할 수 있다. 각 노드를 DFS 방식으로 방문하는 과정에서, 트리의 왼쪽 노드를 방문할 때는 대응하는 변수 노드들의 LLR 값은 [수학식 4] 혹은 [수학식 5]의 f 함수에 의해 계산되며, 트리의 오른쪽 노드를 방문할 때 대응하는 변수 노드들의 LLR 값은 [수학식 6]의 g함수에 의해 계산된다. 예를 들어, SC 복호의 첫 번째 동작으로서, 깊이 d=0의 루트 노드로부터 깊이 d=1 의 왼쪽 노드를 방문하는 것은, 단계 t=log2N-0=log2N의 N개의 변수 노드에 있는 내재적(intrinsic) LLR을 가지고, 단계 t=log2N-1 의 N/2개의 변수 노드에 대한 LLR 을 f 함수로 계산하는 것으로 이해될 수 있다. 이와 같이, 상기 이진 트리에서의 DFS 방식의 방문에 의해, 부모 노드(parent node)가 나타내는 변수 노드들의 LLR들을 이용하여 자식 노드(child node)가 나타내는 절반의 변수 노드들 LLR 값을 노드의 위치에 따라 f 함수 혹은 g 함수로 계산하게 된다. DFS 방식으로 상기 과정을 진행하여 만약 i번째 리프 노드의 LLR 값이 계산되면, 이 값이 해당 i 번째 부호화 입력 비트 ui 에 대한 AP-LLR이 된다. 이를 바탕으로
Figure PCTKR2020018557-appb-I000038
를 추정하면, 추정된 비트 값은 DFS 방식의 역-추적(traceback) 순서에 의해 다시 부모 노드에 대응하는 변수 노드들로 전달되며, 해당 부모 노드의 오른편 자식 노드를 방문할 때 각 Z 모양의 기본 요소에서 [수학식 6]과 같은 g 함수 연산을 가능케 한다.
이진 트리를 바탕으로 설명된 복호 과정으로부터 i 번째 부호화 입력 비트 ui의 복호를 위해서 수행되는 극부호 그래프 상에서의 단계 별 연산의 종류와 횟수를 확인할 수 있다. 예를 들어, 길이 8의 극부호 에서 u0에 대한 복호를 할 때, 극부호 그래프의 단계 2, 1, 0 에서는 각각 f,f,f 연산이 각각 4회, 2회, 1회 수행된다. 또 다른 예로, u6 에 대한 복호 시, 극부호 그래프의 단계 2, 1, 0 에서는 g,g,f 연산이 4회, 2회, 1회 수행된다. 이를 일반화하면, ui의 복호를 위해서 수행되는 연산의 단계 별 종류는 비트 인덱스 i의 이진 표현(binary representation)으로부터 확인될 수 있다. 비트 인덱스 i 를 2진수 i=(in-1 in-2 ... i0)2 로 표기할 경우, it 는 곧 극부호 그래프에서 단계 t 에서 수행하는 연산의 종류를 나타내게 된다. 만약 it=0 이면, ui의 복호 과정 중 단계 t 에서는 f 연산이 수행되며, 만약 it=1 이면, ui의 복호 과정 중 단계 t 에서는 g 연산이 수행된다.
또한 상기 이진 트리의 DFS 방식을 이용한 설명으로부터, SC 복호기 동작의 연산 절차와 순서를 확인할 수 있다. SC 복호기가 길이 N의 극부호의 i 번째 비트 ui 를 복호하는 상황을 고려하자. 복호기가 수행하는 두 종류의 f 함수와 g 함수는 동일한 복잡도와 지연으로 구현할 수 있다고 가정하며, f 함수와 g 함수 구분 없이 총 수행하는 모든 함수 연산의 횟수가 확인될 수 있다. 만약 특별한 방법을 사용하지 않고, 길이 N의 극부호 에서 한 부호화 입력 비트를 복호할 때 필요한 f 함수와 g 함수의 연산 총 횟수는 다음과 같이 계산될 수 있다. 앞서 살펴보았던 바와 같이, 단계 t 에서 LLR 값이 계산되는 변수 노드의 개수, 즉 총 연산의 횟수는
Figure PCTKR2020018557-appb-I000039
이다. 각 비트의 복호를 위해서, 단계 t=log2N-1 으로부터 t=0 로의 연산이 발생하며, 단계 t 에서 변수 노드
Figure PCTKR2020018557-appb-I000040
개의 LLR에 대한 연산이 각각 이루어지므로, 총 연산의 횟수는 이하 [수학식 7]과 같이 계산될 수 있다.
Figure PCTKR2020018557-appb-I000041
총 N개의 부호화 입력 비트들에 대해서 전술한 연산이 수행되므로, 극부호의 SC 복호에 필요한 f 함수와 g 함수 연산의 총 수는 N(N-1) 이다. 하지만, 극부호의 구조를 적절하게 활용하면, 각 부호화 입력 비트를 복호하기 위해 항상 N-1 회의 연산을 수행할 필요는 없다. 예를 들어, 길이 8의 극부호에서 u4와 u5를 복호하는 것을 고려하면, 비트 u4의 복호 시 단계 2, 1, 0 에서 각각 g,f,f 함수 연산이 발생하고, u5 의 복호 시 단계 2, 1, 0 에서 각각 g,f,g 함수 연산이 수행된다. 비트 u4와 u5의 복호에서 수행하는 단계 2와 단계 1에서의 함수 종류와 입력 대상이 모두 동일하기 때문에, 순차적인 복호 과정에서 비트 u5의 복호 시 해당 연산을 다시 수행할 필요가 없다. 즉, 비트 u5의 복호 시 단계 t=0 에서의 g 함수 연산만이 요구된다. 이와 같이, 각 부호화 입력 비트의 복호 시 필요한 연산만 수행될 경우, 전체 수행되는 연산의 총 횟수는 Nlog2N 가 된다. 즉, SC 복호기는 효율적으로 구현할 경우, 복잡도는 O(N2) 가 아니라 O(Nlog2N)로 구현될 수 있다.
상술한 과정을 이진 트리 상에서의 DFS를 이용하여 설명한다. DFS 방식으로 노드들을 방문하는 과정에서, 복호기는 특정 리프 노드에 도달하면, 다시 부모 노드로 역-추적(traceback)하고, 부모 노드의 오른쪽에 있는 노드를 방문했는지 여부가 확인된다. 부모 노드의 오른쪽에 있는 노드를 방문하지 않았다면, 복호기는 해당 노드를 방문하게 되고, 그렇지 않았다면 상위의 부모 노드로 역-추적(traceback)한다. 이진 트리 상에서 이러한 절차는 앞서 설명한 효율적으로 구현된 SC 복호의 동작을 정확하게 묘사한다. 도 7의 예에서, 비트 u4 에 대응하는 리프 노드를 방문한 뒤에, 단계 t=1(깊이 t=2)의 부모 노드를 들러 바로 오른편에 있는 단계 t=0(깊이 t=3)의 비트 u5 에 대응하는 리프 노드를 찾게 된다. 이는 곧 비트 u4 의 복호 후에 바로 단계 t=0 의 연산을 수행하는 것을 묘사한다. 이때 수행되는 연산의 횟수는 단계 t=0 의 노드가 나타내는 변수 노드의 수
Figure PCTKR2020018557-appb-I000042
개이며, 비트 u5 에 대응하는 노드가 부모 노드의 오른편 노드이기 때문에 연산의 종류는 g 함수이다. 이후, 비트 u6 의 복호 시 DFS의 절차에 따라 단계 t=1 의 연산이 먼저 수행된다.
SCL 복호는 다음과 같이 수행된다. SCL 복호는 각 부호화 입력 비트에 대한 LLR 값 계산 및 추정 단계에서 사전에 정한 L 개의 후보군을 유지하는 방법이다. SCL 복호기는 비트 ui에 대한 복호 시, 앞선 비트들 u0,u1,...,ui-1의 복호 과정에서 얻어진 L 개의
Figure PCTKR2020018557-appb-I000043
부분(partial) 추정 비트 시퀀스를 고려한다. 각 부분 추정 비트 시퀀스 후보군은 리스트(list) 혹은 경로(path)라고 지칭되며, 각 리스트에 대해서 확률 정보인 경로 메트릭(path-metric, PM)이 각 비트 복호 시마다 업데이트된다. PM은 각 리스트에 대한 확률 정보로서, LLR 기반으로 구현된 SCL 복호기의 경우, 리스트의 PM이 낮을수록 해당 리스트의 확률이 높은 것을 의미한다. SCL 복호기는 복호 시작 전에 각 리스트의 PM을 사전에 정한 값으로 초기화하는데, 초기 값은 보통 0이다. SCL 복호기는 각 리스트
Figure PCTKR2020018557-appb-I000044
에 대해서 ui 의 LLR 등의 확률 정보를 계산하고, ui가 각각 0과 1일 경우에 대한 LLR 등의 확률을 계산하여 2L 개의 리스트
Figure PCTKR2020018557-appb-I000045
,
Figure PCTKR2020018557-appb-I000046
에 대한 PM을 업데이트한다. 업데이트된 2L 개의 리스트
Figure PCTKR2020018557-appb-I000047
중 PM 을 바탕으로 L 개의 리스트를 선택하고 나머지는 폐기한다(pruning). 이러한 과정은 비트 인덱스에 따라 순차적으로 수행된다. 최초 u0 를 복호하는 과정에서 고려되는 리스트의 개수는 1개이며, 복호하는 비트의 인덱스가 작은 복호 초반의 경우에는 고려하는 리스트 개수는 L 개보다 작다. 비동결 비트를 한 비트씩 복호할 때마다 리스트의 개수는 2배가 되며, 복호가 진행되면서 앞서 설명한 바와 같이 L 개의 리스트들이 고려된다. 각 리스트에 대한 복호는 서로 다른 부분 추정 비트 시퀀스
Figure PCTKR2020018557-appb-I000048
를 독립적으로 고려하여 진행되므로, SCL 복호기가 수행하는 연산량은 SC 복호기의 연산량의 대략 L 배가 된다.
도 8은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 이진 결정 트리(binary decision tree)의 예를 도시한다. 도 8은 길이 8의 극부호에 대해 SCL 복호기가 순차적 비트를 추정하는 과정을 묘사한 이진 결정 트리의 일 실시 예다. 도 8의 이진 결정 트리는 앞서 살펴본 도 5의 극부호 그래프, 도 6의 복호 이진 트리와는 다르다. SCL 복호는 SC 복호와 마찬가지로 비트 인덱스 순으로 부호화 입력 비트를 복호한다. SCL 복호의 이러한 동작은 도 8의 이진 결정 트리에서 루트 노드로부터 시작하여 발생할 수 있는 경로 중 최대 L 개를 유지하며 한 깊이씩 순차적으로 내려가는 동작으로 이해될 수 있다. 도 8의 예는 SCL 복호기의 리스트 크기 L=4 로 결정된 상황을 가정하며, 매 깊이마다 최대 4 개의 경로가 유지된다. 또한, 도 8의 예는 길이 8 극부호의 부호화 입력 비트 중 {u0,u1,u2,u4} 가 동결 비트이고, 나머지 비트 {u3,u5,u6,u7} 가 비동결 비트로 결정된 상황을 가정한다.
비트 u0,u1,u2 는 동결 비트이기 때문에, 계산된 LLR 의 값에 상관없이 비트 값이 0으로 결정된다. 비트 u0,u1,u2 가 모두 0이라는 가정 하에, 비동결 비트 u3 에 대한 LLR 값을 계산한다. 도 8에 표기된
Figure PCTKR2020018557-appb-I000049
은 부호화 입력 비트 ui 가 값
Figure PCTKR2020018557-appb-I000050
을 가질 경우에 대한 확률값으로 BM(branch metric) 라고 일컬으며, LLR 값으로부터 얻어진다. PM은 각 비트에 대한 BM을 함께 고려하여 얻어진다. 도 8의 SCL 복호기는 최대 L=4의 리스트를 고려할 수 있기 때문에, u3 에 대한 복호 후 두 리스트들 (u0,u1,u2,u3)=(0,0,0,0), (u0,u1,u2,u3)=(0,0,0,1)는 도 8에 도시된 것과 같이 모두 유지될 수 있다. 두 리스트들에 대한 PM 은 각 비트 값에 대해 계산한 BM을 바탕으로 업데이트된다. 그리고, 비트 u4 는 동결 비트이므로, 두 리스트들은 (u0,u1,u2,u3,u4)=(0,0,0,0,0), (u0,u1,u2,u3,u4)=(0,0,0,1,0) 으로 업데이트되며, 리스트들 각각에 대한 PM도 업데이트 된다. 이때, 비록 u4 가 동결 비트이고 값이 0으로 공통이지만, 각 리스트에서 가정하는 (u0,u1,u2,u3) 의 값이 다르므로, 서로 다른 LLR, BM이 계산될 수 있다. 이러한 절차에 의해, 비동결 비트 u5 에 대한 복호가 수행된다. 도 8의 예에서, 비동결 비트 u6 에 대한 SCL 복호를 수행해야 비로소 L=4 보다 많은 8개의 후보군이 생성되는데, 이때 계산된 PM 을 바탕으로 L=4 개의 리스트가 선택되고, 나머지 4개는 버려진다. 이와 같은 절차에 의해, 모든 비트에 대한 복호를 마무리하면, 총 L=4 개의 리스트들이 만들어지는데, 복호기는 이 중 PM 을 바탕으로 확률이 가장 높다고 판단한 하나의 리스트를 선택 및 출력한다. 만약 CRC 부호 등의 외부 부호화가 함께 연접되었다면, 리스트 중 CRC 부호의 조건을 만족하면서 PM 기반으로 확률이 가장 높다고 판단한 리스트가 선택 및 출력된다.
다양한 실시 예들에 따른 복호 기술은 SCL 복호기를 성능 손실 없이 낮은 복잡도와 지연을 가질 수 있다. 이를 위해, SCL 복호기가 수행하는 동작을 보다 자세하게 분석하여 복잡도/지연 관점에서 개선할 수 있는 부분을 확인하도록 한다. 도 8의 이진 결정 트리를 참고하며, 일 실시 예에 따른 복호 과정을 설명하면 다음과 같다.
각 비트 ui, i=0,...N-1에 대해서, 추정 값에 따라 계산하는 BM 은 이진 결정 트리에서 최대 L 개의 경로 또는 리스트를 선택하는 과정에서 고려된다. 도 8에 도시된 바와 같이, 첫 비동결 비트 이전의 동결 비트들을 위해 계산한 BM은 이후의 모든 PM들에 동일한 값으로서 포함되며, 이는 서로 다른 PM들 사이에 공통으로 포함된 일종의 상수와 같이 작용한다. 따라서, SCL 복호기에서 첫번째 비동결 비트 이전의 동결 비트들을 위해 계산한 BM은 이후 순차적인 복호를 진행하면서 L 개의 리스트를 선택함에 있어서 아무런 영향을 주지 않는다. 비동결 비트 인덱스 중 가장 작은 값을 qmin 라 하면, 부호화 입력 비트
Figure PCTKR2020018557-appb-I000051
는 동결 비트들이며, 전술한 내용에 근거하여 해당 비트들에 대한 PM을 업데이트할 필요가 없으며, 나아가, 해당 비트들에 대해서 복호 자체를 수행할 필요가 없다.
부호화 과정에서, 부호율 조정을 위해 단축(shortening) 동작이 수행될 수 있다. 오류 정정 부호에 대한 단축은 부호의 길이(예: 출력 비트 개수) 및 차원(예: 입력 비트 개수)를 동일한 크기만큼 줄이는 동작이다. 단축된 극부호(shortened polar code)에서 단축된 비트의 값은 동결 비트와 동일하게 고정되며, 일반적으로 0이다. 일부 문헌에서, 단축된 비트(shortened bits) 및 동결 비트를 구분하지 않고 표기하기도 하지만, 단축된 비트 미 동결 비트는 다른 목적에 의해 사용된다.
극부호의 구조적 특성으로 인해, 단축을 할 비트들은 이진 지배(binary domination) 규칙에 의해 결정될 수 있다. 길이 N 의 극부호에서 J 비트를 단축하는 것을 고려하면, 단축되는 비트들의 인덱스 집합을
Figure PCTKR2020018557-appb-I000052
라고 할 때,
Figure PCTKR2020018557-appb-I000053
는 이진 지배의 특성에 따라 보통 모집합 {0,…,N-1} 에서 큰 값을 갖는 인덱스들로 구성된다. 큰 인덱스를 갖는 비트들이 단축된다는 것은, 곧 순차적인 복호 순서 상 뒤쪽의 비트들이 단축됨을 의미한다.
다양한 실시 예들에 따른 복호 기술의 일 동기는 단축된 극부호에서 가장 마지막 비동결 비트 이후에 위치한 단축된 비트(shortened bit)들에 대해서 PM을 업데이트 할 필요가 없다는 것이다. 단축된 비트의 값은 보통 0으로 고정되므로, 정보를 포함하지 아니한다. 따라서, 복호기는 해당 비트에 대한 복호를 수행할 필요가 없다. 또한, 극부호의 비트 생성의 인과성(causality)으로 인해, 복호 순서 상 뒤쪽에 있는 단축된 비트가 앞쪽에 위치한 다른 비트들의 생성 및 복호에 영향을 주지 아니한다. 단축된 극부호에서 비동결 비트 인덱스 중 가장 큰 값을 qmax 라고 하면, 부호화 입력 비트
Figure PCTKR2020018557-appb-I000054
는 단축된 비트들이며, 앞서 설명한 것처럼 단축된 비트들에 대해서 PM을 업데이트할 필요가 없다. 즉, 단축된 비트들에 대한 복호는 수행되지 아니할 수 있다.
전술한 동결 비트들 및 단축된 비트들에 근거하여 수행되는 복호 동작, 즉, 수신단의 동작은 이하 도 9 또는 이하 도 10과 같다.
도 9는 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 수신단의 복호화를 위한 흐름도를 도시한다. 도 9는 수신단(120)의 동작 방법을 예시한다.
도 9를 참고하면, 901 단계에서, 수신단은 데이터를 수신한다. 데이터는 극부호를 이용하여 부호화된 복수의 비트들을 포함한다. 수신단이 단말인 경우, 수신단은 데이터의 수신에 앞서 DCI(downlink control information)을 수신할 수 있다. 수신단이 기지국인 경우, 수신단은 데이터의 수신에 앞서 UCI(uplink control information) 를 송신할 수 있다.
903 단계에서, 수신단은 복호가 불필요한 전단의 적어도 하나의 비트를 확인한다. 일 실시 예에 따라, 수신단은 DCI 또는 UCI에 포함되는 정보에 기반하여 적어도 하나의 비트를 확인할 수 있다.
905 단계에서, 수신단은 나머지 비트들 중 적어도 일부에 대한 복호를 수행한다. 예를 들어, 복호가 불필요한 적어도 하나의 비트가 복수의 비트들의 후단에도 존재하는 경우, 수신단은 복호 동작을 복수의 비트들의 마지막 비트 전에 종료할 수 있다.
도 10은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 수신단의 복호화를 위한 다른 흐름도를 도시한다. 도 10은 수신단(120)의 동작 방법을 예시한다.
도 10을 참고하면, 1001 단계에서, 수신단은 극 부호의 입력 비트 개수 K 및 출력 비트 개수 E를 확인한다. 다시 말해, 수신단은 주어지는 부호 파라미터인 부호의 입력 비트 개수에 대응하는 코드 차원(code dimension) 및 출력 비트 개수에 대응하는 코드 길이(code length)를 확인한다.
1003 단계에서, 수신단은 사용할 극부호의 부호를 구성한다. 예를 들어, 수신단은 부호 파라미터에 기반하여 사용할 극부호의 모부호(mother code), 부호율-조정(rate matching) 방식 등을 결정한다.
1005 단계에서, 수신단은 부호화의 시작 비트 인덱스 s 및 종료 비트 인덱스 e를 결정한다. 수신단은 결정된 부호 파라미터, 부호 설정 등에 기반하여 SCL 복호기를 시작할 비트 인덱스 s 및 종료할 비트 인덱스 e를 결정할 수 있다. 부호 파라미터, 설정에 기반하여 시작/종료 비트 인덱스 s 및 e를 결정함에 있어서, 다양한 방법들이 활용될 수 있다. 일 실시 예에 따라, s 및 e는 부호 파라미터, 설정과 관련된 수학식에 의해 결정될 수 있다. 다른 실시 예에 따라, s 및 e는 부호 파라미터, 설정과 관련된 참조 테이블에 의해 결정될 수 있다. 또한, 또 다른 실시 예에 따라, s 및 e는 부호 파라미터, 설정과 관련된 적절한 함수에 의해 결정될 수 있다. 또 다른 실시 예에 따라, s 및 e는 머신 러닝, 딥러닝, 강화학습 중 적어도 하나에 기반하여 결정될 수 있다. 결정된 시작 인덱스 및 종료 비트 인덱스에 기반하여, SCL 복호가 수행된다.
1007 단계에서, 수신단은 비트 us로부터 복호를 시작한다. 다양한 실시 예들에 따른 SCL 복호는 종래의 SCL 복호와 달리 us에 대한 LLR을 계산하는 것으로 시작된다. 즉, us의 복호에 앞서, 어떠한 연산도 수행되지 아니할 수 있다. 따라서, us의 LLR 계산을 위해서 활용할 수 있는 사전에 계산된 결과가 없기 때문에, us의 복호에 필요한 모든 연산이 수행될 수 있다. 즉, us의 LLR 계산 시 채널을 통해 주어지는 단계 log2N 의 내재적(intrinsic) LLR을 바탕으로, 각 단계 t 에서
Figure PCTKR2020018557-appb-I000055
개의 변수 노드에 대한 LLR 값이 순차적으로 계산된다. 이 경우, 전체 연산의 횟수는 [수학식 7]과 같이 N-1 회가 된다.
1009 단계에서, 수신단은 비트 ue까지 순차적으로 복호를 수행한다. 즉, 수신단은 비트 us+1,...,ue 에 대한 복호를 수행한다. 이때, 각 비트에 대한 복호 시, 해당 비트보다 앞서 복호된 비트들에 대해 수행된 연산의 결과가 활용될 수 있다.
1011 단계에서, 수신단은 추정 부호화 입력 비트를 선택하고, 출력한다. 즉, ue에 대한 복호가 완료되면, 수신단은 SCL 복호를 종료하고, 리스트들 중 하나를 출력한다. 이때 출력되는 리스트는 PM, 외부 부호화에 대한 복호화 결과 중 적어도 하나에 기반하여 선택될 수 있다.
도 10과 같은 방식을 통해, 수신단의 SCL 복호기는 us 이전 적어도 하나의 비트 및 ue 이후 적어도 하나의 비트에 대한 복호를 수행하지 아니하고, 생략할 수 있으며, 이를 통해 복호에 필요한 연산량과 지연이 감소된다.
상기의 동작에서 s=0 혹은 s=N/2 인 경우에는, 본 발명의 동작이 아니어도 N-1 회의 연산이 수행되지만, 극부호 시퀀스의 이진 지배(binary domination)에 의한 부분 순서 특성 상 s=0, s=N/2 이 될 수 없음에 유의하여 한다. 이러한 점에서, 전술한 복호 동작은 종래의 기법과 차별성을 가진다.
다양한 실시 예들에 따라, 시작/종료 비트 인덱스 s 및 e는 다음과 같이 결정될 수 있다.
일 실시 예에 따르면, 복호기는 주어지는 부호 파라미터와 설정에 따라 구성되는 극부호의 가장 작은 비동결 비트 인덱스 qmin 과 가장 큰 비동결 비트 인덱스 qmax 를 확인하고, 이에 따라 각각 s=qmin, e=qmax 를 설정할 수 있다. 이러한 설정에서 만약 극부호의 부호율-조정(rate-matching) 방법이 단축(shortening) 이 아닌 천공(puncturing), 반복(repetition), 확장(extension) 등인 경우에는 e=qmax=N-1 로 고정할 수 있다. 이는 단축 외의 부호율-조정 방법을 사용할 경우, 부호화 입력 비트 중에 가장 신뢰도가 높은 비트가 uN-1 이기 때문에, 이 비트는 반드시 비동결 비트로 선택된다는 특징을 사전에 고려한 것이다. 이와 같이 부호 파라미터에 따라 결정되는 qmin 및 qmax를 각각 시작/종료 비트 인덱스로 활용하면, 가장 많은 비트에 대한 복호를 스킵할 수 있으며, 가장 큰 복잡도 및 지연 저감 효과를 얻을 수 있다.
다른 실시 예에 따르면, 복호기는 주어지는 부호 파라미터와 설정에 따라 구성되는 극부호의 가장 작은 비동결 비트 인덱스 qmin 과 가장 큰 비동결 비트 인덱스 qmax를 확인하고, 이에 따라 각각 0≤s≤qmin, qmax≤e≤N-1 을 만족하는 범위에서 s 및 e를 결정할 수 있다. 이 실시 예에 의해 구성된 복호기는 모든 부호 파라미터마다 시작/종료 비트 인덱스 를 s=qmin, e=qmax 로 설정하는 실시 예의 복호기보다는 더 적은 비트에 대한 복호를 스킵한다. 따라서, 앞서 살펴본 실시 예보다는 높은 복호 복잡도 및 지연을 보인다. 하지만, 복호기가 부호 파라미터마다 s 및 e를 결정하는데 사용하는 연산량, 메모리 등이 더 적다는 장점이 있다.
도 11은 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 시작 비트 인덱스의 분포에 대한 일 예를 도시한다. 도 11은 3GPP NR 하향링크를 위한 극부호 시스템에서, 부호화 입력 비트 수 A 및 부호율-조정 후 출력 비트 수 E 에 따라 구성된 극부호의 첫 번째 비동결 비트 인덱스, 즉, qmin 을 예시한다. 도 11에서, qmin의 값이 명암으로 표시되어 있으며, 밝을수록 그 값이 크고, 어두울수록 그 값이 작다. 도 11에 나타난 바와 같이, 일정 범위의 (A,E) 조합에서는 qmin값이 동일하다. 또한, 동일한 모부호 크기 N을 사용하게 되는 유사한 파라미터 (A,E) 조합에서의 qmin 값 또한 비슷하다. 더불어, 부호율 조정 방법으로 단축이 사용될 경우에는 높은 인덱스의 비트들인 단축되면서, 상대적으로 낮은 인덱스의 비트가 비동결 비트가 되기 때문에 qmin 이 작아진다. 이러한 특징들을 활용하면, 부호 파라미터 (A,E), 모부호 크기 N, 사용하는 부호율-조정 방법 등을 바탕으로 시작 비트 인덱스 s를 효율적으로 결정하는 것이 가능하다.
도 12 및 도 13은 (A,E) 조합에 따른 qmin 값을 바탕으로 복호를 시작한 비트 인덱스 0≤s≤qmin 를 결정하는 일 실시 예들을 나타낸다. 앞서 살펴본 바와 같이 극부호 복호의 복잡도와 처리 시간은 모부호의 크기 N에 비례한다. 모부호의 크기 N이 크면, 복잡도 및 처리 시간이 크기 때문에, 결정되는 N에 따라 적응적으로 시작 비트 인덱스 s를 결정할 수 있다. 다음은 도 12에 도시된 것과 같이 하향링크를 위한 극부호 시스템에서 시작 비트 인덱스 s를 결정하는 일 예를 나타낸다. 하기의 실시 예는 편의 상 가장 간단한 시스템에 대한 것으로, 실제 시스템은 더 다양한 분기에 의해 구성될 수 있음에 유의하여야 한다.
(1) 만약 N<512 이면, s=0 으로 설정한다. 3GPP NR 하향링크 극부호 시스템에서 최대 모부호의 크기는 N=512 이며, 이 경우가 가장 복잡도와 처리 시간이 크다. N이 256 이하이면 복잡도와 처리 시간이 최대보다 절반 이하가 되므로, 이 이러한 경우에는 s=0 으로 설정할 수 있다. 물론 N이 256 이하인 경우에 대해서도 qmin을 참고하여 s 값을 0보다 크도록 결정할 수 있다.
(2) 만약 N=512, A≥80 이고 단축이 아닌 다른 부호율-조정 방법이 사용되었다면, 도 9에 표시된 것과 같이 해당 영역에서 가장 작은 qmin 인 127 을 s 로 설정한다.
(3) 만약 N=512, A<80이고 단축이 아닌 다른 부호율-조정 방법이 사용되었다면, 도 12에 표시된 것과 같이 해당 영역에서 가장 작은 qmin 인 223 을 s 로 설정한다.
다음은 도 13에 도시된 것과 같이 하향링크를 위한 극부호 시스템에서 시작 비트 인덱스 s를 결정하는 또 다른 일 예를 나타낸다.
(1) 만약 N<256 이면, s=0 으로 설정한다.
(2) 만약 N=512, A≥80이고 단축이 아닌 다른 부호율-조정 방법이 사용되었다면, 도 13에 도시된 것과 같이 해당 영역에서 가장 작은 qmin 인 127 을 s 로 설정한다.
(3) 만약 N=512, A<80이고 단축이 아닌 다른 부호율-조정 방법이 사용되었다면, 도 10에 표시된 것과 같이 해당 영역에서 가장 작은 qmin 인 223 을 s 로 설정한다.
(4) 만약 N=256, A>38 이고 단축이 아닌 다른 부호율-조정 방법이 사용되었다면, 도 10에 표시된 것과 같이 해당 영역에서 가장 작은 qmin 인 31 을 s 로 설정한다.
(5) 만약 N=256, A≤38 이고 단축이 아닌 다른 부호율-조정 방법이 사용되었다면, 도 10에 표시된 것과 같이 해당 영역에서 가장 작은 qmin 인 95 를 s 로 설정한다.
상기의 예제들과 같이 일정 범위의 A (A,E) 조합에 대해서 가장 작은 qmin 을 s 로 설정함으로써 복호를 스케줄링하는데 필요한 복잡도, 메모리, 지연 등을 감소시킬 수 있다. 동일한 값의 s 를 사용할 (A,E) 조합의 범위는 시스템을 설계할 때 요구 사항과 특징 등을 고려하여 미리 결정될 수 있다. 만약 복호 복잡도와 지연의 감소를 목표로 하면 더 세세하게 작은 (A,E) 범위에 따라 s를 설정할 수 있다. 만약 복호 복잡도와 처리 시간의 감소보다 간단한 복호 스케줄링을 목표로 하면 큰 (A,E) 조합의 범위를 기반으로 s를 설정할 수 있다.
부호 파라미터 (A,E) 에 따른 qmin 값 분포를 바탕으로 s 를 결정하는 것과 마찬가지로 부호 파라미터 (A,E) 에 따른 qmax 값 분포를 바탕으로 qmax≤e≤N-1 범위에서 e 가 결정할 수 있다. 이 때 앞서 언급한 바와 같이 단축이 아닌 천공, 확장, 반복 등의 부호율-조정 방법이 사용된 경우, 혹은 E 가 2의 제곱수라 어떠한 부호율-조정 방법도 사용되지 않은 경우, qmax=N-1 이 자연스러운 결정임을 고려하여 e 값의 설정이 보다 단순화될 수 있다. 즉, 부호율-조정 방법이 단축인 경우에 대해서만 qmax 를 고려하여 qmax≤e≤N-1 범위에서 e 를 결정하고, 그렇지 않은 경우에는 e=qmax=N-1 로 고정할 수 있다. 도 14는 3GPP NR 하향링크를 위한 극부호 시스템에서, 부호화 입력 비트 수 A 와 부호율-조정 후 출력 비트 수 E 에 따라 구성된 극부호의 마지막 비동결 비트 인덱스, 즉, qmax 을 도시한다. 도 14는 극부호의 구성이 단축을 사용하는 경우에 대해서만 qmax 를 도시한다. 이와 같이, 사전에 기-결정된 qmax의 정보를 바탕으로, qmax≤e≤N-1 범위에서 e 를 결정한다.
s 및 e의 설정을 위해 qmin 및 qmax분포를 확인하는 것은 복호기 혹은 복호기의 사전 프로세서의 설계와 관련된 것이다. 이러한 동작은 복호 동작과 관련이 없음에 유의하여야 한다. 즉, 이러한 동작은 시스템 설계 시 1회의 설계(one-time design)에 해당하는 부분이며, 복호 동작의 복잡도 및 처리 시간을 크게 상승시키지 아니할 것이다.
종래에도 SC 및 SCL 복호기의 복잡도와 지연을 감소시키기 위한 방법들이 제안된 바 있다. 대표적인 방법이 간략화된 SC(simplified SC, SSC) 복호 기법, 간략화된 SSCL(simplified SCL, SSCL) 복호 기법이다. 이러한 방법들은 부호 파라미터에 따라 도 5에 보인 복호 이진 트리를 수정하여 일부 노드들을 합쳐 특수 노드로 만들고, 특수 노드를 별도의 방법으로 효율적으로 처리함으로써 전체적으로 수행되는 연산을 감소시킨다. 대표적인 특수 노드로는 rate-0 노드, rate-1 노드, 반복(repetition) 노드, 단일 패리티 검사(single parity-check, SPC) 노드 등이 있다.
SSC 복호기 또는 SSCL 복호기를 구현하기 위해서, 부호 파라미터에 따라 복호 이진 트리의 분석 및 수정과 같이 복호기의 동작을 스케줄링하는 동작의 선행이 요구된다. 만약 시스템에서 지원해야 하는 부호 파라미터의 조합 개수가 적은 경우, 사전에 스케줄링된 정보를 메모리 등에 저장함으로써, 효율적인 구현이 가능하다. 하지만, 3GPP NR 극부호 시스템과 같이 무수한 부호 파라미터 조합들을 고려하는 경우, 사전에 스케줄링된 정보를 메모리에 저장하는 것이 어려우며, 복호기는 각 부호 파라미터마다 사전에 복호 이진 트리를 분석하고, 스케줄링을 진행해야 한다. 또한, 특수 노드들(예: rate-0, rate-1, repetition, SPC 노드 등) 을 처리하기 위해서 별도의 함수, 프로세스, 메모리 등이 필요하다. 특수 노드의 종류는 동일하더라도, 하나의 특수 노드가 포함하는 자식 노드들의 개수에 따라 서로 다른 함수, 프로세스, 메모리가 필요할 수도 있다. 그리고, 특수 노드에서는 LLR과 PM 계산을 위해 수행하는 연산은 간략화되는 것이지, 연산 자체가 되는 것은 아니다.
다양한 실시 예들에 따른 복호기는 전술한 SSC 복호기 또는 SSCL 복호기에 비하여 간단하게 구현될 수 있다. 도 15는 본 개시의 다양한 실시 예들에 따른 통신 시스템에서 복호기의 구성을 도시한다. 도 15를 참고하면, 복호기는 디코딩 제어기(1520) 및 SCL 디코더(1570)를 포함한다. 디코딩 제어기(1520)은 전술한 실시 예들 중 하나에 따라 시작 비트 인덱스 또는 종료 비트 인덱스를 결정한다. SCL 복호기(1570)는 디코딩 제어기(1520)에 의해 결정된 시작 비트 인덱스 또는 종료 비트 인덱스에 기반하여 복호 동작을 수행한다.
다양한 실시 예들에 따른 복호기에게 요구되는 복호에 대한 스케줄링은 SSC, SSCL 복호보다 간단하다. 본 개시의 다양한 실시 예들에 따른 복호 동작은 복호 이진 트리의 분석을 요구하지 아니하며, 단순히 시작 비트 인덱스 및 종료 비트 인덱스를 결정하는 것만을 요구한다. 나아가, 본 개시의 다양한 실시 예들에 따른 복호기의 경우, 앞서 살펴본 발명의 일 실시 예에 따라, 약간의 성능 손해를 감수하면서 복호기를 스케줄링 하는 복잡도, 메모리 등을 더 감소시키는 것도 가능하다. 또한, 본 개시의 다양한 실시 예들에 따른 복호 방식은 특수 노드를 고려하지 아니하므로, 특수 노드의 종류와 크기에 따라 별도의 함수, 프로세스, 메모리 등을 구성하는 것이 요구되지 아니한다. 마지막으로, SSC 복호 또는 SSCL 복호의 경우, 각 특수 노드에 대해서 PM 등이 계산되지만, 본 개시의 다양한 실시 예들에 따른 복호기는 일부 비트에 대해서 연산 자체를 수행하지 않는다.
도 16 및 17은 본 개시의 다양한 실시 예들에 따른 복호 기법의 성능을 도시한다. 도 16 및 17은 3GPP NR 상향링크 극부호 시스템을 고려한다. 도 16 및 도 17은 극부호 입력 비트 수 A 및 부호율-조정 출력 비트 수 E 에 따른 일반적인 SCL 복호 대비 본 개시의 다양한 실시 예들에 따른 복호 기법의 지연 저하율(latency reduction ratio) 및 복잡도 저하율(complexity reduction ratio)을 각각 보여준다. 각 파라미터에서의 저하율은 일반적인 SCL 복호기와 본 발명의 SCL 복호기가 수행하는 실제 연산 수와 단계를 계산하여 얻어진 것이다. 여기서, 저하율 0은 복잡도 및 지연 저하가 없다는 의미이고, 저하율 1은 모든 연산의 생략됨을 의미한다. 즉, 저하율이 크면 클수록 복잡도 저하율 및 지연 저하율이 크다 의미한다. 이 저하율은 각 도면에서 명암으로 표시되며, 밝을수록 그 값이 크고, 어두울수록 그 값이 작다. 각 도면에 나타난 바와 같이 A 가 작은 경우, 본 발명에 의한 복호 복잡도와 지연 저감이 최대 40% 수준까지 커진다. 3GPP NR 상향링크를 통해 전송되는 UCI(uplink control information) 가 일반적으로 100 비트 이하임을 고려하면, 대부분의 UCI 전송 상황에서 상당한 복호 복잡도 및 지연의 안정적인 저감을 기대할 수 있다.
다양한 실시 예들에 따르면, 극부호 부호화된 부호어의 SCL 복호기 등으로 복호함에 있어, 불필요한 연산을 생략함으로써, 필요한 연산량, 연산 시간이 감소된다. 특히, 전술한 바와 같이, 복호 과정의 단순화를 위해 부호의 구성을 사전에 분석하거나, 분석된 결과에 따라 특정 요소마다 적응적으로 연산을 수행하는 블록이 요구되지 아니한다. 이를 통해, 극부호의 복호 단순화를 위해 추가로 사용/구성하는 연산 및 하드웨어가 최소화된다. 또한, 본 개시의 다양한 실시 예들에 따른 복호 기법은 넓은 범위의 부호 차원(code dimension), 부호 길이(code length), 부호율(code rate) 로 부호화된 부호어의 복호를 빈틈없이 포괄적으로 지원할 수 있다. 본 개시의 다양한 실시 예들에 따르면, 낮은 부호율이 사용되는 경우, 약 절반 이상의 연산이 생략되고, 이로 인해 복호 복잡도 및 지연이 약 40% 가량 낮아질 수 있다.
본 개시의 청구항 또는 명세서에 기재된 실시 예들에 따른 방법들은 하드웨어, 소프트웨어, 또는 하드웨어와 소프트웨어의 조합의 형태로 구현될(implemented) 수 있다.
소프트웨어로 구현하는 경우, 하나 이상의 프로그램(소프트웨어 모듈)을 저장하는 컴퓨터 판독 가능 저장 매체가 제공될 수 있다. 컴퓨터 판독 가능 저장 매체에 저장되는 하나 이상의 프로그램은, 전자 장치(device) 내의 하나 이상의 프로세서에 의해 실행 가능하도록 구성된다(configured for execution). 하나 이상의 프로그램은, 전자 장치로 하여금 본 개시의 청구항 또는 명세서에 기재된 실시 예들에 따른 방법들을 실행하게 하는 명령어(instructions)를 포함한다.
이러한 프로그램(소프트웨어 모듈, 소프트웨어)은 랜덤 액세스 메모리 (random access memory), 플래시(flash) 메모리를 포함하는 불휘발성(non-volatile) 메모리, 롬(read only memory, ROM), 전기적 삭제가능 프로그램가능 롬(electrically erasable programmable read only memory, EEPROM), 자기 디스크 저장 장치(magnetic disc storage device), 컴팩트 디스크 롬(compact disc-ROM, CD-ROM), 디지털 다목적 디스크(digital versatile discs, DVDs) 또는 다른 형태의 광학 저장 장치, 마그네틱 카세트(magnetic cassette)에 저장될 수 있다. 또는, 이들의 일부 또는 전부의 조합으로 구성된 메모리에 저장될 수 있다. 또한, 각각의 구성 메모리는 다수 개 포함될 수도 있다.
또한, 프로그램은 인터넷(Internet), 인트라넷(Intranet), LAN(local area network), WAN(wide area network), 또는 SAN(storage area network)과 같은 통신 네트워크, 또는 이들의 조합으로 구성된 통신 네트워크를 통하여 접근(access)할 수 있는 부착 가능한(attachable) 저장 장치(storage device)에 저장될 수 있다. 이러한 저장 장치는 외부 포트를 통하여 본 개시의 실시 예를 수행하는 장치에 접속할 수 있다. 또한, 통신 네트워크상의 별도의 저장장치가 본 개시의 실시 예를 수행하는 장치에 접속할 수도 있다.
상술한 본 개시의 구체적인 실시 예들에서, 개시에 포함되는 구성 요소는 제시된 구체적인 실시 예에 따라 단수 또는 복수로 표현되었다. 그러나, 단수 또는 복수의 표현은 설명의 편의를 위해 제시한 상황에 적합하게 선택된 것으로서, 본 개시가 단수 또는 복수의 구성 요소에 제한되는 것은 아니며, 복수로 표현된 구성 요소라 하더라도 단수로 구성되거나, 단수로 표현된 구성 요소라 하더라도 복수로 구성될 수 있다.
한편 본 개시의 상세한 설명에서는 구체적인 실시 예에 관해 설명하였으나, 본 개시의 범위에서 벗어나지 않는 한도 내에서 여러 가지 변형이 가능함은 물론이다. 그러므로 본 개시의 범위는 설명된 실시 예에 국한되어 정해져서는 아니 되며 후술하는 특허청구의 범위뿐만 아니라 이 특허청구의 범위와 균등한 것들에 의해 정해져야 한다.
본 개시(disclosure)는 일반적으로 통신 시스템에서 자원 할당에 관한 것으로, 보다 구체적으로 통신 시스템에서 극부호(polar code)의 복호화를 위한 장치 및 방법에 관한 것이다.

Claims (15)

  1. 통신 시스템에서 수신 장치의 동작 방법에 있어서,
    극부호(polar)를 이용하여 부호화된, 복수의 비트들을 포함하는 데이터를 수신하는 과정과,
    상기 복수의 비트들 중 복호 연산이 불필요한 적어도 하나의 비트를 확인하는 과정과,
    상기 적어도 하나의 비트를 제외한 나머지 비트들 중 적어도 일부에 대한 복호를 수행하는 과정을 포함하는 방법.
  2. 청구항 1에 있어서,
    상기 복호 연산이 불필요한 적어도 하나의 비트는, 극부호의 입력 비트 개수 및 출력 비트 개수에 기반하여 확인되는 방법.
  3. 청구항 1에 있어서,
    상기 복호 연산이 불필요한 적어도 하나의 비트는, 극부호의 입력 비트 개수 및 출력 비트 개수의 조합이 속하는 범위에 대응하는 인덱스 값을 검색함으로써 확인되는 방법.
  4. 청구항 3에 있어서,
    상기 범위에 대응하는 인덱스 값은, 상기 범위에 속하는 상기 입력 비트 개수 및 상기 출력 비트 개수의 조합들의 첫번째 비동결(unfrozen) 비트의 인덱스들 중 최소 값인 방법.
  5. 청구항 1에 있어서,
    상기 복호 연산이 불필요한 적어도 하나의 비트는, 극부호의 입력 비트 개수 및 출력 비트 개수에 기반하여 결정되는 상기 극부호에 관련된 부호 설정에 기반하여 결정되며,
    상기 부호 설정은, 모부호(mother code)의 크기, 부호율 조정(rate matching) 방식 중 적어도 하나를 포함하는 방법.
  6. 청구항 1에 있어서,
    상기 복호 연산이 불필요한 적어도 하나의 비트는, 복호에 대한 시작 비트 인덱스 및 복호에 대한 종료 비트 인덱스 중 적어도 하나에 의해 특정되는 방법.
  7. 청구항 1에 있어서,
    상기 복호 연산이 불필요한 적어도 하나의 비트는, 상기 복수의 비트들 중 전단에 위치한 적어도 하나의 동결(frozne) 비트 및 상기 복수의 비트들 중 후단에 위치한 적어도 하나의 단축된(shortened) 비트 중 적어도 하나를 포함하는 방법.
  8. 청구항 1에 있어서,
    상기 적어도 일부에 대한 복호를 수행하는 과정은,
    상기 적어도 일부 중 가장 앞에 위치한 비트를 복호하기 위해, 상기 복수의 비트들의 개수 보다 1 작은 횟수의 LLR(log likelihood ratio) 계산 연산을 수행하는 과정을 포함하는 방법.
  9. 청구항 1에 있어서,
    상기 데이터의 수신에 앞서 DCI(downlink control information)을 수신하는 과정을 더 포함하며,
    상기 복호 연산이 불필요한 적어도 하나의 비트는, 상기 DCI에 포함되는 정보에 기반하여 확인되는 방법.
  10. 청구항 1에 있어서,
    상기 데이터의 수신에 앞서 UCI(uplink control information)을 송신하는 과정을 더 포함하며,
    상기 복호 연산이 불필요한 적어도 하나의 비트는, 상기 UCI에 포함되는 정보에 기반하여 확인되는 방법.
  11. 통신 시스템에서 수신 장치에 있어서,
    송수신기와,
    상기 송수신기와 연결된 적어도 하나의 프로세서를 포함하며,
    상기 적어도 하나의 프로세서는,
    극부호(polar)를 이용하여 부호화된, 복수의 비트들을 포함하는 데이터를 수신하고,
    상기 복수의 비트들 중 복호 연산이 불필요한 적어도 하나의 비트를 확인하고,
    상기 적어도 하나의 비트를 제외한 나머지 비트들 중 적어도 일부에 대한 복호를 수행하는 수신 장치.
  12. 청구항 11에 있어서,
    상기 복호 연산이 불필요한 적어도 하나의 비트는, 극부호의 입력 비트 개수 및 출력 비트 개수에 기반하여 확인되는 수신 장치.
  13. 청구항 11에 있어서,
    상기 복호 연산이 불필요한 적어도 하나의 비트는, 극부호의 입력 비트 개수 및 출력 비트 개수의 조합이 속하는 범위에 대응하는 인덱스 값을 검색함으로써 확인되는 수신 장치.
  14. 청구항 13에 있어서,
    상기 범위에 대응하는 인덱스 값은, 상기 범위에 속하는 상기 입력 비트 개수 및 상기 출력 비트 개수의 조합들의 첫번째 비동결(unfrozen) 비트의 인덱스들 중 최소 값인 수신 장치.
  15. 청구항 11에 있어서,
    상기 복호 연산이 불필요한 적어도 하나의 비트는, 극부호의 입력 비트 개수 및 출력 비트 개수에 기반하여 결정되는 상기 극부호에 관련된 부호 설정에 기반하여 결정되며,
    상기 부호 설정은, 모부호(mother code)의 크기, 부호율 조정(rate matching) 방식 중 적어도 하나를 포함하는 수신 장치.
PCT/KR2020/018557 2019-12-24 2020-12-17 통신 시스템에서 극부호의 복호화 장치 및 방법 WO2021132983A1 (ko)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US17/788,526 US20230034299A1 (en) 2019-12-24 2020-12-17 Device and method for decoding polar code in communication system
EP20904850.3A EP4075672A4 (en) 2019-12-24 2020-12-17 DEVICE AND METHOD FOR POLAR CODE DECODING IN A COMMUNICATION SYSTEM

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
KR10-2019-0174395 2019-12-24
KR1020190174395A KR20210081952A (ko) 2019-12-24 2019-12-24 통신 시스템에서 극부호의 복호화 장치 및 방법

Publications (1)

Publication Number Publication Date
WO2021132983A1 true WO2021132983A1 (ko) 2021-07-01

Family

ID=76574465

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2020/018557 WO2021132983A1 (ko) 2019-12-24 2020-12-17 통신 시스템에서 극부호의 복호화 장치 및 방법

Country Status (4)

Country Link
US (1) US20230034299A1 (ko)
EP (1) EP4075672A4 (ko)
KR (1) KR20210081952A (ko)
WO (1) WO2021132983A1 (ko)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018231924A1 (en) * 2017-06-14 2018-12-20 Idac Holdings, Inc. Two-stage scrambling for polar coded pdcch transmission
WO2019011555A1 (en) * 2017-07-10 2019-01-17 Accelercomm Ltd GENERATION OF PARALLEL PARALLEL BITS FOR BLOCKS FOR POLAR CODES
KR101951663B1 (ko) * 2012-12-14 2019-02-25 삼성전자주식회사 Crc 부호와 극 부호에 의한 부호화 방법 및 장치
KR20190034665A (ko) * 2016-08-12 2019-04-02 알까뗄 루슨트 폴라 코드들의 코딩 및 디코딩 방법 및 장치

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018161274A1 (en) * 2017-03-08 2018-09-13 Qualcomm Incorporated Polar coding design for performance and latency
WO2019227276A1 (en) * 2018-05-28 2019-12-05 Qualcomm Incorporated Polar code construction for incremental redundancy

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101951663B1 (ko) * 2012-12-14 2019-02-25 삼성전자주식회사 Crc 부호와 극 부호에 의한 부호화 방법 및 장치
KR20190034665A (ko) * 2016-08-12 2019-04-02 알까뗄 루슨트 폴라 코드들의 코딩 및 디코딩 방법 및 장치
WO2018231924A1 (en) * 2017-06-14 2018-12-20 Idac Holdings, Inc. Two-stage scrambling for polar coded pdcch transmission
WO2019011555A1 (en) * 2017-07-10 2019-01-17 Accelercomm Ltd GENERATION OF PARALLEL PARALLEL BITS FOR BLOCKS FOR POLAR CODES

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
NOKIA, NOKIA SHANGHAI BELL: "Bit-interleaving for polar codes", 3GPP DRAFT; R1-1716786_BIT-INTERLEAVING FOR POLAR CODES_UPDATED, 3RD GENERATION PARTNERSHIP PROJECT (3GPP), MOBILE COMPETENCE CENTRE ; 650, ROUTE DES LUCIOLES ; F-06921 SOPHIA-ANTIPOLIS CEDEX ; FRANCE, vol. RAN WG1, no. Nagoya, Japan; 20170918 - 20170921, 19 September 2017 (2017-09-19), Mobile Competence Centre ; 650, route des Lucioles ; F-06921 Sophia-Antipolis Cedex ; France, XP051353852 *
See also references of EP4075672A4 *

Also Published As

Publication number Publication date
KR20210081952A (ko) 2021-07-02
EP4075672A1 (en) 2022-10-19
EP4075672A4 (en) 2023-01-18
US20230034299A1 (en) 2023-02-02

Similar Documents

Publication Publication Date Title
JP6817452B2 (ja) レートマッチング方法、符号化装置、および通信装置
US10498490B2 (en) Coding and decoding of a polar code concatenated with interleaving with an outer systematic code
US9831895B2 (en) System and method for a message passing algorithm
CN109257148B (zh) 一种基于高斯近似阈值判断的极化码bp译码方法
KR20190038933A (ko) Qc ldpc 코드의 레이트 매칭 방법 및 이를 위한 장치
JP2022174079A (ja) 符号化方法、復号方法、装置、および装置
EP2599228B1 (en) Decoding techniques for tail-biting codes
KR20020018643A (ko) 고속 map 디코딩을 위한 방법 및 시스템
WO2020226304A1 (en) Apparatus and method for decoding of low-density parity check codes in wireless communication system
CN110383727B (zh) 用于ldpc码的分层解码方法及其装置
WO2015000100A1 (zh) 实现Turbo均衡补偿的方法以及Turbo均衡器和系统
US10819372B2 (en) Method for dividing transport block of LDPC code and apparatus therefor
WO2018079987A1 (ko) Ldpc 코드의 운송블록 분할 방법 및 이를 위한 장치
US10312945B2 (en) Channel coding method using concatenated code structure and apparatus therefor
WO2021132983A1 (ko) 통신 시스템에서 극부호의 복호화 장치 및 방법
US11018696B2 (en) Soft output decoding of polar codes
US20210194509A1 (en) Information decoder for polar codes
WO2021107746A1 (ko) 통신 및 방송 시스템에서 극부호의 복호화 방법 및 장치
KR102338852B1 (ko) 무선 통신 시스템에서 신호를 복호하기 위한 장치 및 방법
CN112737600B (zh) 译码方法和译码器
EP3720023B1 (en) Polar code decoding method and communication device
KR101835341B1 (ko) Sc-ldpc 코드의 쌍방향 슬라이딩 윈도우 복호 방법 및 이를 위한 장치
CN113162633B (zh) 极化码的译码方法及装置、译码器、设备、存储介质
Chhabra et al. An efficient high performance turbo code implementation in Rayleigh fading channel
GB2394627A (en) Communication unit and method of decoding

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20904850

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2020904850

Country of ref document: EP

Effective date: 20220711

NENP Non-entry into the national phase

Ref country code: DE