一种时空频 TURBO编码方法及装置 Time and space frequency TURBO coding method and device
技术领域 Technical field
本发明涉及通信技术领域, 特别涉及一种用于通信系统的串行和并行 时空频编码方法, 具体的讲是一种时空频 TURBO编码方法及装置。 背景技术 The present invention relates to the field of communication technology, and in particular, to a serial and parallel space-time-frequency coding method for a communication system, and in particular to a time-space-frequency TURBO coding method and device. Background technique
通信系统中, 现有的抗衰落方案主要是采用分集技术, 特别是分集技 术中的时空编码技术, 参见参考文献 [13]、 [14] , 在现有技术中介绍的都 是时空编码, 而没有人涉足于时空频编码领域。 发明内容 In communication systems, the existing anti-fading schemes mainly use diversity technology, especially the space-time coding technology in diversity technology, see references [13], [14]. In the prior art, all are described by space-time coding, and No one is involved in the field of space-time-frequency coding. Summary of the Invention
本发明的目的在于, 提供一种时空频 TURBO编码方法及装置。 事实上, 将时空编码扩展到频域, 可以为时空编码提供更灵活的设计空间, 更大的 提高系统的频谱效率和获得更大的分集增益和编码增益。 An object of the present invention is to provide a time-space-frequency TURBO coding method and device. In fact, extending the space-time coding to the frequency domain can provide more flexible design space for space-time coding, greatly improve the spectral efficiency of the system, and obtain greater diversity gain and coding gain.
本发明的技术方案为: 一种时空频 TURBO 编码方法, 其中包括: 发射 端对输入比特进行级连的时空 TRELLIS 编码; 编码时, 级连的内外码均为 时空 TRELLIS 编码, 且将频率域作为编码的一个维数进行联合编码; 接收 端对接收的信号进行译码。 The technical solution of the present invention is: A spatio-temporal frequency TURBO coding method, which includes: spatio-temporal TRELLIS coding for concatenating input bits at the transmitting end; when coding, the concatenated internal and external codes are both spatio-temporal TRELLIS coding, and the frequency domain is used as One dimension of the coding is jointly coded; the receiving end decodes the received signal.
所述的发射端对输入比特进行级连的时空 TRELLIS 编码是指: 发射端 可对输入比特进行并行级连的时空 TRELLIS 编码, 分别将两个交织前后的 时空 TRELLIS编码信号, 分别调制到两个载波上;在接收端, 可以分别恢复 出两个交织前后的时空 TRELLIS编码信号, 这就构成了时空频 TURBO编码。 The spatio-temporal TRELLIS coding for concatenating input bits by the transmitting end refers to the spatio-temporal TRELLIS coding of the concatenating input bits in parallel at the transmitting end, which respectively modulate two spatio-temporal TRELLIS encoded signals before and after interleaving to two On the carrier; at the receiving end, two space-time TRELLIS coded signals before and after interleaving can be recovered respectively, which constitutes space-time frequency TURBO coding.
所述的发射端对输入比特进行级连的时空 TRELLIS 编码是指: 发射端 可对输入比特进行串行级连的时空 TRELLIS编码,其中:外码经时空 TRELLIS 编码后形成的信息和校验符号, 分别交织后再各自经过时空 TRELLIS编码, 然后在两个载波上调制发射出去。 The spatio-temporal TRELLIS coding for concatenating input bits at the transmitting end refers to the spatio-temporal TRELLIS coding for serial concatenation of the input bits at the transmitting end, where the outer code is the information and check symbol formed after the spatio-temporal TRELLIS encoding After being interleaved, they are respectively subjected to space-time TRELLIS coding, and then modulated and transmitted on two carriers.
所述的发射端对输入比特进行并行级连的时空 TRELLIS 编码还可以 指: 发射端可对输入比特进行能够传出一个信息符号和三个校验符号的并 行级连的时空 TRELLIS 编码, 其中: 采用对两个载波的调制, 可以分别恢 复出两个交织前的时空 TRELLIS编码信号及两个交织后的时空 TRELLIS编 码信号, 然后在两个载波上调制发射出去。 The transmitting space-time TRELLIS encoding of the input bits in parallel concatenation may also refer to: The transmitting end may perform time-space TRELLIS encoding of the input bits in parallel concatenation capable of transmitting one information symbol and three check symbols, where: By adopting modulation of two carriers, two space-time TRELLIS coded signals before interleaving and two space-time TRELLIS coded signals after interleaving can be recovered, and then modulated and transmitted on two carriers.
所述的接收端对接收的信号进行译码是指: 接收端对接收的信号可进
行时空频的 LOG-MAP译码。 The decoding of the received signal by the receiving end means that: the receiving end can advance the received signal Time-space-frequency LOG-MAP decoding.
对于并行级连的时空频 TURBO 编码, 所述的接收端对接收的信号进行 译码是指: 接收端对接收的信号可进行符号级的 LOG- MAP译码; For parallel concatenated space-time-frequency TURBO coding, the receiving terminal decoding the received signal means that the receiving terminal can perform symbol-level LOG-MAP decoding on the received signal;
所述的译码可为迭代译码, 所述的迭代译码应满足下列条件: The decoding may be an iterative decoding, and the iterative decoding shall satisfy the following conditions:
Lc{uk = u(i) 17) = log{p(uk = u(i) | 7)) = log^ p(uk = u(i),ak,ak+l | 7)) = l。g(A exp(aA. (ak ) + βΜ (σΜ ) + fl ' σΜ, yk ))) Lc (u k = u (i) 17) = log (p (u k = u (i) | 7)) = log ^ p (u k = u (i), a k , a k + l | 7) ) = l. g (A exp (a A. (a k ) + β Μ (σ Μ ) + fl 'σ Μ , y k )))
― max— ― _. ― Max-- _.
=h + ( k (ak ) + βΜ (σΜ ) + γ (ak, σΜ , yk ))= h + ( k (a k ) + β Μ (σ Μ ) + γ (a k , σ Μ , y k ))
k k
在上式中, 对 "(0求和, 可以得出: In the above formula, summing "(0, we get:
― ^max _ ― ― ^ Max _ ―
h = log(/ = -2, (σ, ) + βΜ (σί+1 ) + fk (ak , σΜ , yk )) 此时可以进行迭代译码, 计算得到的是全信息^(^= )|?), 迭代时 需要利用边信息 yA.l"(0), 可设先验信息为 。 贝 ij h = log (/ = -2, (σ,) + β Μ (σ ί + 1 ) + f k (a k , σ Μ , y k )) ^ (^ =) |?), The side information y A .l "(0) needs to be used in the iteration, and the prior information can be set to be. Ij
Le{yk | u(i)) = Lc{uk = u{i) | Y)~Lo{uk = u(i)) . Le (y k | u (i)) = Lc (u k = u (i) | Y) ~ Lo (u k = u (i)).
判决时, 采用符号级的最大似然判决准则,
When deciding, the maximum likelihood decision criterion at the symbol level is used.
最大的 作为 k时刻的判决符号输出; The largest is output as a decision symbol at time k;
对于 ^,Α, 初始的状态, 采用对每个 RSC编码器分别加尾比特强迫归 零的方案, 此时: ) 其 值 For the initial state of ^, Α, the tail bit is forced to zero for each RSC encoder. At this time:)
一 1, <yk =° A 1, <y k = °
^ ) = {ο, σΑ 其他值 ^ ) = ( ο, σΑ other values
在译码中如有并行路径的情况, 只要把并行路径当作普通的分支即可, 但需在相关的所有对状态求和的公式中, 还应该加一项对并行分支求和。 If there is a parallel path in the decoding, it is sufficient to treat the parallel path as an ordinary branch, but it is necessary to add an item to sum the parallel branches in all relevant formulas for summing states.
对于串行时空频 TURBO 编码, 所述的接收端对接收的信号进行译码是 指: 接收端对接收的信号可进行比特级 LOG- MAP译码; For serial space-time-frequency TURBO coding, the receiving terminal decoding the received signal refers to: the receiving terminal can perform bit-level LOG-MAP decoding on the received signal;
所述的译码可为迭代译码, 所述迭代译码的步骤为: The decoding may be an iterative decoding, and the steps of the iterative decoding are:
(一)载波 1和载波 2接收的信号分别经过接收匹配滤波器的滤波, 并 分别进行 APP计算, 利用式: (1) The signals received by carrier 1 and carrier 2 are respectively filtered by the receiving matched filter, and APP calculations are performed respectively.
p (uk;0) = p (u = uk I ^ ) = A∑∑ «, (σ, )βΜ (σί+Ι )γ[ (σ, , σΜ , yk ) , 计算得到内码
时空编码的码字 41);,42)''符号级的后验概率, 然后利用公式 p (u k ; 0) = p (u = u k I ^) = A∑∑ «, (σ,) β Μ (σ ί + Ι ) γ [(σ ,, σ Μ , y k ), calculated Internal Code Space-time coded codewords 4 1);, 4 2) '' symbol-level posterior probability, and then use the formula
k。 k.
Pi^ I F) = Π p{bt I 7) , 将其转换为比特级的似然比, 分别得到全后验概率的 Pi ^ IF) = Π p (b t I 7), which is converted to a bit-level likelihood ratio to obtain the full posterior probability, respectively.
1=1 1 = 1
^2) ;J) = 0; ^ 2) ; J) = 0;
(二)然后利用式: j = ι"··Α (2) Then use the formula: j = ι "·· Α
PliukU) = o-o) 。) = ln- ·) = 1;0) j = l,...,n0 Pliu k U) = oo). ) = ln- · ) = 1; 0) j = l, ..., n 0
pk e(ck(j) = 0;O) p k e (c k (j) = 0; O)
计算边信息 ,
; Compute side information, ;
(三)将计算得到的内码的 1个边信息 41)£(" );0), fe(ulu,o) 反交织作为外码^。,^。的先验信息 (cr /); ), (c ) /); ), 其中 上标 "0" 表示外码; (3) 1 side information of the calculated inner code 4 1) £ ("); 0), f e (u l u, o) is deinterleaved as the prior information of the outer code ^., ^. (Cr /);), (C) /);), where the superscript "0" indicates a foreign code;
(四)利用式: ( ; ) = f ^( 4 计算 4( ; ), 在外码的后验似 然信息计算时, 14(^;/)始终为零; 将得到的 ^ k(uk;I), 带入到式 λί (c(y); O) = max l k(ak) + βΜ (σΜ ) + λ, (uk; J) + k (c, (σ, );/)] (4) Use the formula: (;) = f ^ (4 to calculate 4 (;). When the posterior likelihood information of the outer code is calculated, 1 4 (^; /) is always zero; the obtained ^ k (u k ; I), brought into the formula λί (c (y); O) = max l k (a k ) + β Μ (σ Μ ) + λ, (u k ; J) + k (c, (σ,) ; /)]
- max [ ¾ ( ^ ) + βΜ (σΜ ) + k (uk ;Ι) + λ1ζ (ck (ak );/)] 中, 得到 1^(φ·);0), j = l,-,n0\ 然后带入到式:-max [¾ (^) + β Μ (σ Μ ) + k (u k ; Ι) + λ 1ζ (c k (a k ); /)], we get 1 ^ (φ ·); 0), j = l,-, n 0 \ Then bring into the formula:
(c(J');0) = (c( );O)— ( ( );J)中, 计算得到 (c(_/);O), 7=1,-, «0; 利用 式: / I = , 将边信息的似然比转换为符号的后验概率 (c (J '); 0) = (c (); O) — ((); J), (c (_ /); O) is calculated, 7 = 1,-, « 0 ; / I =, convert the likelihood ratio of side information to the posterior probability of the symbol
/=1 / = 1
(五)将 P e ck 0、首先进行串并转换, 得到 0fc;O),/ )e( ;O) , 分别经 过交织得到 , 将此交织后的信息分别作为内码时空频编 码的先验信息
; (E) The P e c k 0, is first serial-parallel conversion, to obtain 0 fc; O), /) e (; O), are obtained after the interleaving, the interleaving information of this were used as inner code coded temporal frequency Prior Information ;
(六)在后续的迭代中, 重复以上所述的 5个步棟; (6) Repeat the 5 steps described above in subsequent iterations;
(七)在最后一次迭代时, 输出外码的 l ("0);O) , j = l,- , 采用下面 的方式进行判决:
u门 = {i,当 >o (7) At the last iteration, output the outer code of l ("0); O), j = l,-, and use the following method to make a decision: u- gate = (i, when> o
" - o,当 A ( >o° 然后回到 1, 进行下一帧的译码。 "-o, when A (> o ° and then return to 1, decode the next frame.
对于串行时空频 TURBO 编码, 所述的接收端对接收的信号进行译码是 指: 接收端对接收的信号可进行符号级的 LOG- MAP译码; For serial space-time-frequency TURBO coding, the receiving terminal decoding the received signal refers to: the receiving terminal can perform symbol-level LOG-MAP decoding on the received signal;
所述的译码可为迭代译码, 所述迭代译码的步骤为: The decoding may be an iterative decoding, and the steps of the iterative decoding are:
此时所有的概率信息都是用符号级的后验概率表示的, 而没有用到比 特级的似然比, 利用式: p k 0、 = ptJ ;。 P u k I 计算得到内码时空 编码的码字 ^'',42)'·符号级的后验概率的边信息 0A;O) , p )e{uk;0) , 此 时不需再计算似然信息, 而是将其反交织作为外码译码器的先验信息 At this time, all the probability information is expressed by the posterior probability at the symbol level, but the bit-likelihood ratio is not used. The formula is: p k 0, = ptJ;. P u k I calculates the space-time coded codeword of the inner code ^ '', 4 2) 'Symbol-level posterior probability side information 0 A ; O), p ) e (u k ; 0), The likelihood information needs to be calculated again, but its de-interleaving is used as the prior information of the outer code decoder
A(c:2)。;/); A (c: 2 ). ; /);
在外码译码时: When decoding in outer code:
¾+ι ) = max( , (σ/£ ) + pk(uk;I) + pk (ck (ak ); J) (σΑ. ) = max(¾+I (ak ) + pk (uk ;I) + pk (ck (ak ); I) ή , ,yk) = hpk (c; I)pk (u; I) ¾ + ι) = max (, (σ / £ ) + p k (u k ; I) + p k (c k (a k ); J) (σ Α .) = Max (¾ + I (a k ) + p k (u k ; I) + p k (c k (a k ); I) price,, y k ) = hp k (c; I) p k (u; I)
计算得到的全信息的后验概率为: The calculated posterior probability of the full information is:
Pt (c; 0) = max [ak (ak ) + βΜ (σΜ ) + pk (uk ;I) + pk (ck (σ, );/)] Pt ( c ; 0) = max [a k (a k ) + β Μ (σ Μ ) + p k (u k ; I) + p k (c k (σ,); /)]
° ( (σ¾)=1 ° ((σ ¾ ) = 1
一人 、 n One person, n
- : ( ( )=0 ) + A+i (σΜ ) + Λ (¾; ) + Pk (ck (σ, ); /) pl(c;0) = pt(c;0)-pk(c;I), 其中在外码的迭代中 始终为零; 将 pk e {ck; O)串并转换并交织分别作为内码时空频编码的先验信息
, 以后为迭代过程。 -: (() = 0) + A + i ( σ Μ) + Λ (¾;) + P k (c k (σ,); /) pl (c; 0) = pt (c; 0) -p k (c; I), which is always zero in the iteration of the outer code; p k e {c k ; O) is serial-parallel converted and interleaved as the prior information of the space-time-frequency encoding of the inner code , Iterative process after.
一种时空频 TURBO编码装置, 其中: 发射端至少包括时空 TRELLIS编 码器和交织器, 所述的时空 TRELLIS 编码器和交织器构成级连的时空 TRELLIS 编码装置; 编码时: 输入比特分别输入时空 TRELLIS编码器和交 织器, 并经编码后输出; 接收端至少包括基于 LOG- MAP译码的时空频译码 装置。 A spatio-temporal frequency TURBO encoding device, wherein: a transmitting end includes at least a spatio-temporal TRELLIS encoder and an interleaver, and the spatio-temporal TRELLIS encoder and the interleaver form a cascaded spatio-temporal TRELLIS encoding device; during encoding: input bits are respectively input into the spatio-temporal TRELLIS An encoder and an interleaver, and output after encoding; the receiving end includes at least a time-space-frequency decoding device based on LOG-MAP decoding.
所述的级连的时空 TRELLIS 编码装置是指: 并行级连的时空 TRELLIS 编码装置, 其由时空 TRELLIS 编码器、 交织器、 调制器、 天线构成; 编码 时: 输入比特分别输入时空 TRELLIS 编码器和交织器, 输入时空 TRELLIS 编码器的比特经编码后输出给调制器, 调制器输出的信号经天线发出; 输
入交织器的比特交织后需输入另外的时空 TRELLIS 编码器, 并经编码后输 出给另外的调制器, 该调制器输出的信号经另外的天线发出。 The cascaded spatio-temporal TRELLIS encoding device refers to: a parallel-cascaded spatio-temporal TRELLIS encoding device, which is composed of a spatio-temporal TRELLIS encoder, an interleaver, a modulator, and an antenna; during encoding: input bits are input into the spatio-temporal TRELLIS encoder and Interleaver, the bits of the input space-time TRELLIS encoder are encoded and output to the modulator, and the signal output by the modulator is sent out via the antenna; The bit interleaved into the interleaver needs to be input into another space-time TRELLIS encoder, and after being encoded, it is output to another modulator, and the signal output by the modulator is sent through another antenna.
所述的级连的时空 TRELLIS 编码装置是指: 串行级连的时空 TRELLIS 编码装置, 其由时空 TRELLIS 编码器、 交织器、 调制器、 天线构成; 编码 时: 输入比特先输入时空 TRELLIS编码器, 该时空 TRELLIS编码器的输出 分别为至少两个交织器的输入, 交织器的输出分别为另外的至少两个时空 TRELLIS 编码器的输入, 该至少两个时空 TRELLIS 编码器的输出分别为至 少两个调制器的输入, 该至少两个调制器输出的信号分别经至少两个天线 发出。 The cascaded spatio-temporal TRELLIS encoding device refers to: a serially cascaded spatio-temporal TRELLIS encoding device, which is composed of a spatio-temporal TRELLIS encoder, an interleaver, a modulator, and an antenna; during encoding: input bits are first input into the spatio-temporal TRELLIS encoder The outputs of the space-time TRELLIS encoder are the inputs of at least two interleavers, the outputs of the interleaver are the inputs of at least two other space-time TRELLIS encoders, and the outputs of the at least two space-time TRELLIS encoders are at least two Input of two modulators, and the signals output by the at least two modulators are respectively sent through at least two antennas.
所述的级连的时空 TRELLIS 编码装置是指: 可传出一个信息符号和三 个校验符号的并行级连的时空 TRELLIS编码装置, 其由时空 TRELLIS编码 器、交织器、调制器、天线构成;编码时:输入比特分别输入一时空 TRELLIS 编码器和一交织器, 输入时空 TRELLIS 编码器的比特经编码后输出给一调 制器, 调制器输出的信号经一天线发出; The cascaded space-time TRELLIS encoding device refers to: a space-time TRELLIS encoding device capable of transmitting one information symbol and three check symbols in parallel, which is composed of a space-time TRELLIS encoder, interleaver, modulator, and antenna ; When encoding: input bits are input to a space-time TRELLIS encoder and an interleaver respectively, the bits of the input space-time TRELLIS encoder are encoded and output to a modulator, and the signal output by the modulator is sent through an antenna;
输入交织器的比特交织后需分别输入第二个时空 TRELLIS编码器和第 二个交织器, 第二个交织器的输出为第三个时空 TRELLIS 编码器的输入, 第二个时空 TRELLIS编码器的输出和第三个时空 TRELLIS编码器的输出共 同为第二个调制器的输入, 第二个调制器输出的信号经第二个天线发出。 After the bits of the input interleaver are interleaved, the second space-time TRELLIS encoder and the second interleaver must be input respectively. The output of the second interleaver is the input of the third space-time TRELLIS encoder. The output and the output of the third space-time TRELLIS encoder are the input of the second modulator, and the signal output by the second modulator is sent by the second antenna.
所述的译码装置可为单速率并行时空频 TURBO译码器; 其由匹配滤波 器、 APP 计算器、 交织器、 反交织器、 判别装置构成; 在译码时: 载波 1 和载波 2接收的信号分别经过接收匹配滤波器 1, 2进入两个 APP计算器, 经基于 LOG- MAP译码算法的迭代译码完成译码。 The decoding device may be a single-rate parallel space-time-frequency TURBO decoder; it is composed of a matched filter, an APP calculator, an interleaver, an deinterleaver, and a discriminating device; when decoding: carrier 1 and carrier 2 receive The signals pass through the receiving matched filters 1, 2 and enter the two APP calculators, and the decoding is completed by iterative decoding based on the LOG-MAP decoding algorithm.
所述的译码装置可为串行时空频 TURBO译码器; 其由匹配滤波器、 APP 计算器、 交织器、 反交织器、 判别装置构成; 在译码时: 载波 1 和载波 2 接收的信号分别经过接收匹配滤波器 1、 2进入两个 APP计算器, 所述的两 个 APP计算器的输出分别为反交织器 1、 2 的输入, 所述的反交织器 1、 2 的输出共同成为第三个 APP计算器的输入, 经基于 LOG- MAP译码算法的迭 代译码完成译码。 The decoding device may be a serial space-time-frequency TURBO decoder; it is composed of a matched filter, an APP calculator, an interleaver, an de-interleaver, and a discriminating device; at the time of decoding: carrier 1 and carrier 2 receive The signals enter the two APP calculators through the receiving matched filters 1 and 2, respectively. The outputs of the two APP calculators are the inputs of the deinterleavers 1 and 2, respectively, and the outputs of the deinterleavers 1 and 2 are common. Become the input of the third APP calculator, and complete the decoding by iterative decoding based on the LOG-MAP decoding algorithm.
本发明的有益效果为: 本发明将时空编码扩展到频域, 可以为时空编 码提供更灵活的设计空间, 更大的提高系统的频谱效率和获得更大的分集 增益和编码增益。
附图说明 The beneficial effects of the present invention are as follows: The present invention extends space-time coding to the frequency domain, which can provide more flexible design space for space-time coding, greatly improve the spectral efficiency of the system and obtain greater diversity gain and coding gain. BRIEF DESCRIPTION OF THE DRAWINGS
图 1为并行时空频编码结构框图; Figure 1 is a block diagram of the parallel space-time-frequency coding structure;
图 2为串行时空频编码结构框图; Figure 2 is a block diagram of the serial space-time-frequency coding structure;
图 3为处理一个信息符号和三个校验符号的并行时空频编码结构框图; 图 4为单速率并行时空频 TURBO译码器结构框图; Figure 3 is a block diagram of a parallel space-time-frequency coding structure that processes one information symbol and three check symbols; Figure 4 is a block diagram of a single-rate parallel space-time-frequency TURBO decoder;
图 5为串行时空频译码器结构框图; FIG. 5 is a structural block diagram of a serial space-time-frequency decoder;
图 6为移动速度为 5km/h时串并行时空频编码的 FER性能曲线; 图 7为移动速度为 60km/h时串并行时空频编码的 FER性能曲线。 Figure 6 shows the FER performance curve of serial-parallel spatiotemporal frequency coding at a moving speed of 5km / h; Figure 7 shows the FER performance curve of serial-parallel spatiotemporal frequency coding at a moving speed of 60km / h.
具体实施方式 detailed description
I .串并行时空频编码结构: I. Serial and parallel space-time-frequency coding structure:
在串并行时空频编码中, 级连的内外码都是时空 TRELLIS 编码。 只是 此时将频率域也作为一个编码的维数考虑进去来联合编码。 In serial-parallel space-time-frequency coding, the concatenated internal and external codes are all space-time TRELLIS codes. However, at this time, the frequency domain is also considered as a coding dimension to jointly code.
并行时空频编码的结构如图 1 所示, 在图中我们可以看出, 采用对两 个载波的调制, 我们可以单独的恢复出两个交织前后的时空编码信号, 因 而, 实际上这就构成了时空频 TURBO 编码。 在后面的译码算法中, 我们可 以看出, 采用上面单速率的编码方案, 可以大大的筒化译码算法。 上面这 种方案, 在单径信道下, 在两个接收天线时, 可以达到 1/2 的编码增益和 8重分集增益。 The structure of parallel space-time-frequency coding is shown in Figure 1. In the figure, we can see that by modulating two carriers, we can recover the two space-time-coded signals before and after interleaving separately. Therefore, this actually constitutes Spatio-temporal frequency TURBO coding. In the later decoding algorithm, we can see that the decoding algorithm can be greatly simplified by using the above single-rate coding scheme. In the above scheme, in a single-path channel, when two receiving antennas are used, a coding gain of 1/2 and an 8-diversity gain can be achieved.
图 2 给出了串行时空频的编码结构, 在串行时空频编码器中, 外码时 空编码后形成的信息和校验符号, 分别交织后再各自经过两个时空编码器, 然后在两个载波上调制发射出去。 图中所示的两个交织应该相同, 或者可 以先将外码的信息和校验和起来, 交织, 然后再分开。 Figure 2 shows the serial space-time-frequency encoding structure. In a serial space-time-frequency encoder, the information and check symbols formed after space-time encoding of the outer code are interleaved and then pass through two space-time encoders. Modulate and transmit on each carrier. The two interleavings shown in the figure should be the same, or the information of the outer code and the checksum can be first interleaved, and then separated.
并行的时空频编码可以有另外一种编码方案, 如图 3 所示, 此时, 类 似于 1/4的 TURBO编码器, 与前面的结构不同, 此时传一个信息符号, 和 3个校验符号, P条低了编码速率, 但是却增加了译码的复杂度。 Parallel space-time-frequency coding can have another coding scheme, as shown in Figure 3. At this time, similar to the 1/4 TURBO encoder, it is different from the previous structure. At this time, an information symbol and 3 checksums are passed. Symbols, P strips lower the coding rate, but increase the complexity of decoding.
II .符号级并行时空频 TURBO编码的 LOG- MAP译码: II. LOG-MAP decoding of symbol-level parallel space-time-frequency TURBO coding:
以上我们给出了串行和并行的时空频编码器, 在这里我们给出针对以 上并行时空频编码器的符号级 LOG- MAP译码算法: Above we have given serial and parallel space-time-frequency encoders. Here we give symbol-level LOG-MAP decoding algorithms for the above parallel space-time-frequency encoders:
在 [8] [9] [10]中给出了, 针对级连码的基于 SIS0模型的通用 MAP译 码算法, 本发明将借鉴其中的 SIS0模型和 MAP译码算法, 并将其应用于时 空频的 LOG- MAP译码算法中。
在下面的推导中, 我们对 TURBO码通用的 LOG- MAP译码算法([15])不 进行详细的推导, 而将主要的篇幅放在符号级的 LOG-MAP译码算法上。 In [8] [9] [10], a general MAP decoding algorithm based on the SIS0 model for concatenated codes is given. The present invention will draw on the SIS0 model and the MAP decoding algorithm and apply it to space-time Frequency in the LOG-MAP decoding algorithm. In the following derivation, we do not make a detailed derivation of the LOG-MAP decoding algorithm ([15]) common to TURBO codes, but put the main space on the symbol-level LOG-MAP decoding algorithm.
在时刻 k, 信息符号的后验概率为: At time k, the posterior probability of the information symbol is:
P(uk = u {
( 1 ) 式中, p uk = u(i), ok , am I Y)表示已知接收信号 +时, uk, σ,, σ,+1的联合概 率分布, 下面的推导与通常的 TURBO码的 LOG- MAP的译码算法相同,我们只 给出结果, 详细的内容可以参考 BCJR的译码算法 [1] [2]。 P (u k = u { (1) where pu k = u (i), o k , a m IY) represents the joint probability distribution of u k , σ,, σ, +1 when the received signal + is known. The following derivation is compared with the ordinary The decoding algorithm of LOG-MAP of TURBO code is the same. We only give the result. For details, please refer to the decoding algorithm of BCJR [1] [2].
在(1) 式中, 应用贝叶斯准则, 可以得到: In formula (1), applying the Bayes rule, we get:
P(uk = u(i), ak , σ4+1 I Ϋ) = hp(uk = u(i), ak , ak+1 ,Ϋ) ( 2 ) 式中的 h满足 P (u k = u (i), a k , σ 4 + 1 I Ϋ) = hp (u k = u (i), a k , a k + 1 , Ϋ) (2) where h satisfies
^ΣΣΣ^ - =" '),σ„σ,+1,7) = 1 ( 3 ) 我们将?分为 f = ( ―, 则有 ^ ΣΣΣ ^-= "'), σ„ σ, +1 , 7) = 1 (3) Will we? Divided into f = (―, then
P{uk = u{i), ak , σΑ.+1 , 7) = p(uk = u(i), ak , σΜ ~,yk>fk +) ( 4 )P (u k = u (i), a k , σ Α . +1 , 7) = p (u k = u (i), a k , σ Μ ~, y k> f k + ) ( 4 )
= P(Yk + I uk =u(i),ak,ak+l,Yk~,yk)p(uk = u(i)k, ak+1 ,yk \ak, Yk~)p(ak , Yk ) 根据 Markov链的性质, 如果已知 σω, 那么 +将不取决于其他参数, 此时有 = P (Y k + I u k = u (i), a k , a k + l , Y k ~, y k ) p (u k = u (i) k , a k + 1 , y k \ a k , Y k ~) p (a k , Y k ) According to the properties of the Markov chain, if σ ω is known, then + will not depend on other parameters.
P(Yk + I uk = u(i),ak,aM,Yk',yk) = p(fk + \ σΜ) P (Y k + I u k = u (i), a k , a M , Y k ', y k ) = p (f k + \ σ Μ )
同理 P(¾ =u(i)k,^k+l,yk \ak,Yk~) = p(uk =u(i)k,aM,yk\ak), Similarly, P (¾ = u (i) k , ^ k + l , y k \ a k , Y k ~) = p (u k = u (i) k , a M , y k \ a k ),
如果我们定义 If we define
rk' , , ) = p(uk = "(!·)' ' yk I k ) rk ',,) = p ( u k = "(! ·)''y k I k )
ak(ak) = p(ak,Y,;) 那么, (2) 式我们可以写成如下的形式: a k (a k ) = p (a k , Y ,;) Then, we can write (2) as follows:
P{uk = "(0, o-k , σΜ I Υ) = hak (ak )β ι (ak+i )γ[ (ak , ak+l, yk ) P {u k = "(0, o- k , σ Μ I Υ) = ha k (a k ) β ι (a k + i ) γ [(a k , a k + l , y k )
在时刻 k, 当输入为^)时, 从状态 σέ转移到 σΑ+1的联 合概率分布, 可以表示为: At time k, when the input is ^), the state proceeds to σ έ σ Α + 1 joint probability distribution can be expressed as:
ή , σ ι ,yk)
= "(?:), ak ' σΜ )p(uk = "(01。k, +1 )ρ(σΜ I σ ( 5 ) 如果我们假设当输入为《(0时, 在 TRELLIS 格图上, 从状态 σΑ转移到 σΜ , 这样, (5) 式可以表示为
ή (^k , σί+ι ,yk) = p(yk I "(0, σλ. )p(uk = )) ( 6 ) 上式中的 i7(3/£ |w(), )表示条件概率, 表示在时刻 k, 在状态 , 当输 入为 M(0时, 接收信号为 Λ的概率 ή, σ ι, y k ) = "( ? :), A k ' σ Μ) p ( u k =" (01. k , +1 ) ρ (σ Μ I σ (5) If we assume that the input is "(0, in the TRELLIS grid On the graph, the state σ Α is shifted to σ Μ , so that (5) can be expressed as ή (^ k, σ ί + ι, y k ) = p (y k I "(0, σ λ .) p (u k =)) (6) i7 (3 / £ | w () ,) Represents the conditional probability, which represents the probability at time k, in the state, when the input is M (0, the received signal is Λ
因为 Λ=# + 所以 P( I u(i),ak) = (πΝ0ΥΜ exp(--L|kc - kSk() Because Λ = # + so P (I u (i), a k ) = (πΝ 0 Υ Μ exp (-L | k c - k S k ()
N, N,
上式中的 M为接收天线的个数。 M in the above formula is the number of receiving antennas.
如果对于每个衰落信道, 有 L个可分离的多径, 则上式将变为: If for each fading channel, there are L separable multipaths, the above formula will become:
P(yk I u(i),ak) -HH¾ ')P (y k I u (i), a k ) -H H ¾ ')
上式中, 带有 /的各个变量表示相应的第 /径信号, 因而只是在计算度 量的时候, 将 L个可分离的多径的度量加起来即可。 In the above formula, each variable with / represents the corresponding / th path signal, so it is only necessary to add L separable multipath metrics when calculating the metric.
采用 LOG- MAP译码算法, 我们定义: Using the LOG-MAP decoding algorithm, we define:
Yk (σ4 , σΑ.+1 ,yk) = \og{yk (ak , σΜ , yk )) 可以得到 Yk (σ 4 , σ Α . +1 , y k ) = \ og (y k (a k , σ Μ , y k )) can be obtained
lQg(∑ exp(¾ (<yk ) + fk (σ, , σΜ , yk ))l Q g (∑ exp (¾ (<y k ) + f k (σ ,, σ Μ , y k ))
t t
« ax( k (ak ) + fk (ak , ak+1 , yk )) 同理, «Ax ( k (a k ) + f k (a k , a k + 1 , y k )) Similarly,
(crk ) = l。g(∑ exp(¾+1 (ak ) + fk {ak , ak+l, yk )) (cr k ) = l. g (∑ exp (¾ +1 (a k ) + f k {a k , a k + l , y k ))
« max(^+1 (ak ) + k (ck , σΜ, yk )) 因为 +1(σ ) , (σέ)在迭代过程中是逐渐增大的, 所以为防止 άΜ(σΜ) , 的溢出, 必须对他们进行归一化, 所谓的归一化只是减 去最大值的过程:«Max (^ +1 (a k ) + k (c k , σ Μ , y k )) Because +1 (σ), (σ έ ) increases gradually during the iteration process, in order to prevent ά Μ ( σ Μ), overflow, they must be normalized, the so-called process is just minus the maximum normalized:
k ' (σλ. ) = ak (ak ) - max ak (ak ) k '(σ λ .) = a k (a k )-max a k (a k )
° °
' ) = (σ, ) - max β, (σ, ) 并且 '(σ 和 A' ( )用于后序的迭代和计算中 , 已经证明 (σΑ)和 ¾( )加减一个因子, 不会对后面的后验概率的计算产生任何的影响, 这 个因子最终会被消掉。
因为在 k+1时刻的状态 σ4+1 , 是在状态 当输入为《(0时的转移状态, 所以在 (1 ) 式的求和中, 我们只需对状态^求和, 此时 (1 ) 式的表示式 变为对数运算之后为: ') = (σ,)-max β, (σ,) and' (σ and A '() are used in subsequent iterations and calculations, it has been proven that (σ Α ) and ¾ () add or subtract a factor, not It will have any effect on the calculation of the posterior probability, and this factor will eventually be eliminated. Because the state σ 4 + 1 at time k + 1 is the transition state when the input is "(0, so in the sum of (1), we only need to sum the state ^, at this time ( 1) The expression of the formula becomes logarithmic operation:
Lc{uk = u(i) I Ϋ) = log(p(uk = u(i) | 7)) = log^ p(uk = u(i),ak,ak+l 1 7)) Lc (u k = u (i) I Ϋ) = log (p (u k = u (i) | 7)) = log ^ p (u k = u (i), a k , a k + l 1 7 ))
= l°g(^∑ ex (¾ (σΑ. ) + ¾+1 (σ,+1 ) + γ[ pk , ak+l, yk ))) ( η )= l ° g (^ ∑ ex (¾ (σ Α .) + ¾ +1 (σ, +1 ) + γ [p k , a k + l , y k ))) (η)
― max — 一 _ . ― Max — a _.
= h + ( k (σ/£ ) + βΜ (ak+l ) + fk (σ/£, σΜ , yk )) 在上式中, 我们对 "(0求和, 可以求出 = h + ( k (σ / £ ) + β Μ (a k + l ) + f k (σ / £ , σ Μ , y k )) In the above formula, we can sum up "(0, we can find
- ^-, max _ — _. -^-, max _ — _.
Λ = log(/z) = -2^ {ak (ak ) + βΜ (σΑ+1 ) + γ (σ, , σΜ , yk )) Λ = log (/ z) = -2 ^ (a k (a k ) + β Μ (σ Α + 1 ) + γ (σ ,, σ Μ , y k ))
'· Gk '· G k
此时, 我们可以进行迭代译码, 在 (7 ) 式中, 计算得到的是全信息 Lc{uk = u(i) I Ϋ) , 迭代时需要利用边信息 Le(yk \ u(i)) , 我们假设先验信息为 Lo{uk = u(i)) At this time, we can perform iterative decoding. In formula (7), the full information Lc {u k = u (i) I Ϋ) is obtained. The side information Le (y k \ u (i )), We assume that the prior information is Lo (u k = u (i))
Le(yk | u(i)) = Lc(uk = u(i) | Y) ~Lo{uk = u(i)) Le (y k | u (i)) = Lc (u k = u (i) | Y) ~ Lo (u k = u (i))
判决时, 我们采用符号级的最大似然判决准则, 即取后验概率 p{uk = } | )最大的"(0作为 k时刻的判决符号输出。 When making a decision, we use the maximum likelihood decision criterion at the symbol level, that is, take the largest posterior probability p {u k =} |) "(0 as the decision symbol output at time k.
对于 初始的状态, 与传统的 TURBO码类似, 由于我们采用的是 递归卷积码, 所以采用传统的加零尾比特的方法, 不能保证系统的状态归 零。 在 TURBO码中, 如果没有 TRELLIS的状态终止, 系统的性能会大大的 恶化 [3] , [4] , [5]。 为简化系统的设计, 我们在交织时, 只是对信息比特 进行交织, 每个 RSC 编码器具有自己的尾比特, 虽然这会带来性能的损失 [6] [7] , 但是会筒化系统的设计。 在我们的方案中, 每个 RSC的状态终止, 采用强迫的状态归零法, 因为我们知道每个 RSC 编码器结束时的状态, 所 以我们可以通过增加尾比特, 强迫状态归零。 本发明采用的这种算法并非 最优, 最优的 TURBO码的状态终止, 需要在交织前加尾比特 [1] [3], 这样 可以保证 TURBO具有最大的自由距离, 因为这时, 等效于, 我们知道了整 个 TURBO码的 TRELLIS格图, 是对 TURBO码整个格图的状态终止。 这种稍 复杂的 TURBO码的状态归零算法,需要和交织器结合起来一起设计, 可以保 证达到最优的性能, 具体的算法可以参考以上的参考文献。 The initial state is similar to the traditional TURBO code. Because we use recursive convolutional codes, the traditional method of adding zero tail bits cannot guarantee that the state of the system returns to zero. In the TURBO code, if no TRELLIS state is terminated, the system performance will be greatly deteriorated [3], [4], [5]. In order to simplify the design of the system, we only interleave the information bits when interleaving. Each RSC encoder has its own tail bit. Although this will bring a loss of performance [6] [7], it will simplify the system. design. In our scheme, the state of each RSC is terminated, and the forced state return method is adopted, because we know the state at the end of each RSC encoder, so we can force the state to zero by adding a tail bit. The algorithm used in the present invention is not optimal. The state of the optimal TURBO code terminates, and tail bits [1] [3] need to be added before interleaving. This can ensure that TURBO has the maximum free distance, because at this time, the equivalent Therefore, we know the TRELLIS trellis of the entire TURBO code, which is the termination of the state of the entire trur code of the TURBO code. This slightly complex state return algorithm for TURBO codes needs to be designed together with the interleaver to ensure optimal performance. For specific algorithms, please refer to the above references.
在本发明中, 我们采用对每个 RSC 编码器分别加尾比特强迫归零的方
案, 此时 In the present invention, we use the method of Case, at this time
一 。k = 0 One. k = 0
Ω<3(σί') = {0, σ/£为其他值 Ω <3 ( σί ') = { 0, σ / £ are other values
1, σ4 = 01, σ 4 = 0
^^^ ^,^^为其他值 ^^^ ^, ^^ are other values
图 4 给出了单速率并行时空频的译码结构, 在译码器中有并行路径的 情况, 我们只要把并行路径当作普通的分支即可, 此时上面所有对状态求 和的公式中, 还应该加一项对并行分支求和即可。 此时, 其他所有的公式 都不需要做任何改动, 即可应用到具有并行路径的情况。 Figure 4 shows the decoding structure of single-rate parallel spatiotemporal frequency. In the case of a parallel path in the decoder, we only need to treat the parallel path as a common branch. At this time, in all the above formulas for summing states, You should also add a sum of parallel branches. At this time, all other formulas can be applied to the case with parallel paths without any changes.
III.符号级串行时空频 TURBO编码的 LOG- MAP译码: III. LOG-MAP decoding of symbol-level serial space-time-frequency TURBO coding:
以上我们介绍了并行的时空频的 LOG- MAP译码算法, 在这理我们介绍 串行的时空频 TURBO编码的 LOG-MAP译码算法。 串行的时空频 TURBO编码 的译码过程相对于并行的时空 TURBO 编码的译码而言, 要略微复杂, 因为 在译码过程中, 需要反复的在比特的似然比和符号的后验概率的计算之间 互相转换。 但是串行的时空频编码的性能要优于并行的时空频编码, 从后 面给出的仿真, 可以看出这一点。 Above, we introduced the parallel space-time-frequency LOG-MAP decoding algorithm. Here we introduce the serial space-time-frequency TURBO coding LOG-MAP decoding algorithm. The decoding process of serial spatiotemporal TURBO coding is slightly more complicated than the decoding of parallel spatiotemporal TURBO coding, because during the decoding process, the likelihood ratio of the bits and the posterior probability of the symbol need to be repeated. Conversions between calculations. However, the performance of serial space-time-frequency coding is better than parallel space-time-frequency coding. This can be seen from the simulations given below.
符号级的串行级连 TURBO 码的译码算法已经在 [11] [12] [13]被介 绍。 由于时空 TURBO 编码是基于符号的编码, 所以传统的观念认为其译码 能采用基于符号级的 LOG- MAP译码算法。 The symbol-level serial concatenated TURBO code decoding algorithm has been introduced in [11] [12] [13]. Since the spatio-temporal TURBO coding is symbol-based coding, the traditional concept is that its decoding can use the symbol-based LOG-MAP decoding algorithm.
在这里, 破除了传统观念的框架, 首先给出一种比特级的串行级连时 if-频 TURBO码的译码算法, 这种译码算法比符号级译码算法要复杂, 引入 这个算法的意图在于, 说明时空频编码既可以采用符号级的 LOG- MAP译码 算法, 也可以采用比特级的 LOG- MAP译码算法。 而且根据这种译码算法, 可以很容易的推导出符号级的译码算法。 下面将根据这个算法给出符号级 的 LOG-MAP译码算法。 Here, to break the framework of traditional concepts, a bit-level serial concatenated time-frequency TURBO code decoding algorithm is first given. This decoding algorithm is more complicated than the symbol-level decoding algorithm. This algorithm is introduced. The intention is to explain that the spatio-temporal frequency coding can use either a symbol-level LOG-MAP decoding algorithm or a bit-level LOG-MAP decoding algorithm. And based on this decoding algorithm, a symbol-level decoding algorithm can be easily derived. The symbol-level LOG-MAP decoding algorithm will be given below based on this algorithm.
在这种译码算法中, 我们会用到比特级的似然比的计算, 而且还会 计算符号级的后蹌概率, 所以下面先给出已知符号的后验概率, 求比特级 的似然比的过程, 以及已知比特级的似然比, 计算符号的后验概率的过程。 In this decoding algorithm, we will use the bit-level likelihood ratio calculation, and also calculate the symbol-level posterior probability, so the posterior probability of a given symbol is given below to find the bit-level likelihood. The process of likelihood ratio, and the likelihood ratio of a known bit level, the process of calculating the posterior probability of a symbol.
已知符号的后验概率, 求比特级的似然比: Given the posterior probability of the symbol, find the bit-level likelihood ratio:
假设, 符号集为 S- ,^,...,^} , 而且每个符号 都是由 k。个比特组 成, 记为 {&,^,..., :。}。 同时设符号级 S 中的 N 个符号的后验概率记为 P = (pi,p2,-,pN) = {p(sl \Y),p(s2 \Y),-,p(sN \Y)} , 下面给出符号中第 /个比特
的似然比 l, , λ _ Ρ , = Y) _ 善 ( 9 ) Assume that the symbol set is S-, ^, ..., ^}, and each symbol is composed of k. It consists of bits, denoted as {&, ^, ...,:. }. Also set the posterior probability of N symbols in symbol level S as P = (p i , p 2 ,-, p N ) = (p (s l \ Y), p (s 2 \ Y),-, p (s N \ Y)}, the following gives the / bit in the symbol Likelihood ratio l,, λ _ P, = Y) _ good (9)
'"^=017)" ∑p(Si \Y) 上式中, e{b/ =1}表示符号中第 /个比特为 1的所有符号的集合。 已知符号的后 -脸 f既率, 求比特级的似然比: '"^ = 017)" Σp ( Si \ Y) In the above formula, e {b / = 1} represents the set of all symbols whose symbol / bit is 1. Knowing the back-face f rate of the symbol, find the bit-level likelihood ratio:
符号集合及其元素的定义与相面的相同, 假设我们知道第 /个比特的 后验概率的似然 首先求出:
The definition of the symbol set and its elements is the same as that of the face. Suppose we know the likelihood of the posterior probability of the / bit, first find
1 + βλ' 1 + β λ '
1 1
p(bl =0\Y) = p (b l = 0 \ Y) =
\ + ελ· \ + ε λ ·
我们可以将上两式写成一个通用的用 bt表示的式子 , p(b, I Y) = 6 We can write the above two formulas as a general formula expressed by b t , p (b, IY) = 6
1 + βλ' 1 + β λ '
若符号 由 kQ个比特组成, 记为 {b , .·., } 贝' J: p(si \Y) = flp(bl \Y) (10) If the symbol is composed of k Q bits, it is denoted as {b, ...,} '' J: p (s i \ Y) = flp (b l \ Y) (10)
1=1 1 = 1
(9), (10) 两式给出了有符号的后验概率计算比特级的似然比和由比 特的似然比计算符号的后验概率的算法, 这在后面的计算中都将被用到。 (9), (10) Both formulas give the signed posterior probability calculation algorithm of bit-level likelihood ratio and the posterior probability of symbol by bit likelihood ratio algorithm, which will be used in the subsequent calculations. Used.
时空频 TURBO编码的比特级的 LOG- MAP译码算法: Space-time-frequency TURBO-encoded bit-level LOG-MAP decoding algorithm:
在这种译码算法中, 我们在接收端知道的是符号级的后验概率, 而在 外码的计算和迭代中, 用到的都是比特级的似然比。 In this decoding algorithm, we know the symbol-level posterior probability at the receiving end, and in the calculation and iteration of the outer code, we use the bit-likelihood ratio.
首先, 为下面推导的方便, 我们先定义一些符号表示, 如果没有特殊 的说明, 这些符号的意义就是下面介绍的含义: First, for the convenience of the following derivation, we first define some symbolic representations. If there is no special description, the meaning of these symbols is the meaning introduced below:
u: 表示输入的信息比特 (w(l), M(2),..., u(k0 ) u: input information bits (w (l), M (2), ..., u (k 0 )
c: 表示信息编码后的码字(c(l),c(2),...,c(i。) c: code words (c (l), c (2), ..., c (i.))
/: 表示译码器的输入信息 /: Represents the input information of the decoder
0: 表示译码器的输出信息(包括全信息和边信息) 0: indicates the output information of the decoder (including full information and side information)
u(j): 表示输入信息比特 u的第: j个比特 u (j): input information bit u: jth bit
c(j): 表示编码后码字 c的第 j个比特
表示后验概率的似然比c (j): represents the j-th bit of the codeword c after encoding Likelihood ratio representing posterior probability
Pk AQ: 表示全后验概率P k A Q: represents the full posterior probability
(): 表示全似然比信息 (): Indicates full likelihood ratio information
与上面的推导类似, 在时刻 k, 信息符号的后验概率为: Similar to the above derivation, at time k, the posterior probability of the information symbol is:
(11) 上式中的 属于符号集^ (11) in the above formula belongs to the symbol set ^
我们可以将等式右面求和符号的式子展开写成 , Μ ,γ[ 的表达式, 即: ^(" 所以: We can expand the expression of the summation symbol on the right side of the equation into the expression of Μ , γ [, that is: ^ ("So:
Ρί (
k , σΜ ,yk) ( 12 ) 上面式中的 , βΜ , 的含义与表达式与上面的相同。 关键在于 ^的计 算。 Ρί ( k , σ Μ , y k ) ( 12 ) In the above formula, β Μ , has the same meaning and expression as above. The key lies in the calculation of ^.
在比特级的 LOG- MAP译码算法中, 迭代中既用到比特级的似然比, 还 要用到上面的公式计算的符号级的后验概率, 然后利用前面给出的符号级 的后验概率到比特级的似然比的转换公式(9)进行转换。 In the bit-level LOG-MAP decoding algorithm, both the bit-level likelihood ratio and the symbol-level posterior probability calculated by the above formula are used in the iteration, and then the symbol-level posterior given above is used. The conversion formula (9) from the test probability to the bit-level likelihood ratio is converted.
在上面, 我们得到 On top, we get
ή ' σΜ, ) = Pu (Λ I " = i , σ*, σΜ )pk (u^uk \ k, σΜ )Pk (σλ-+ι \ k) ( 13 ) 如果我们假设当输入符号为 ^时, 在 TRELLIS格图上, 从状态 σλ转移 到 +1, 这样, 上式可以表示为 Price ' σ Μ,) = Pu (Λ I "= i , σ *, σ Μ) p k ( u ^ u k \ k, σ Μ) Pk ( σ λ- + ι \ k) ( 13 ) If we assume When the input symbol is ^, on the TRELLIS trellis, transition from the state σ λ to +1 . In this way, the above formula can be expressed as
rk l^k,ak+l,yk) = pk(yk \ uk,ak)pk{u = uk) ( 14) 上式中的 ρ(Λ 1^, )表示条件概率, 表示在时刻 k, 在状态 crt, 当输 入为 ^时, 接收信号为 Λ的概率 r k l ^ k , a k + l , y k ) = p k (y k \ u k , a k ) p k (u = u k ) (14) ρ ( Λ 1 ^,) in the above formula represents Conditional probability, which represents the probability that the received signal is Λ at time k and in the state cr t when the input is ^
因为 所以 (Λ I = exp(- ||j ||2) ( 15 ) 上式中的 M为接收天线的个数。 Because (Λ I = exp (-|| j || 2 ) (15), M is the number of receiving antennas.
同时, 和/? 々迭代表达式如下 At the same time, and /? 々 iteration expression is as follows
«ί+ι ) * max( , (σ, ) + fk (σ, , σω ,yk)) (16)
k (σ4 ) «匪 (¾+1 (σΑ ) + , ( Α , σΜ ,yk)) ( 17 ) 利用 (12 ) 式, 我们可以求出符号的全后验概率 ^( ;O) , 则边信息 可以求得为: « Ί + ι) * max (, (σ,) + f k (σ ,, σ ω , y k )) (16) k (σ 4 ) «Band (¾ +1 (σ Α ) +, ( Α , σ M , y k )) (17) Using the formula (12), we can find the full posterior probability of the symbol ^ (; O ), Then the side information can be obtained as:
pt(uk;0) = pt(uk;0)-Pk(uk;I) ( 17-2 ) 然后利用 [9]式, 我们可以求出边信息的似然比 pt (u k ; 0) = pt (u k ; 0) -Pk (u k ; I) (17-2) Then using [9], we can find the likelihood ratio of the side information
对每一个载波, 我们都可以计算 (^( );0), 如果设两个载波计算得到 的边信息的似然比分别为: For each carrier, we can calculate (^ (); 0). If we set the likelihood ratios of the side information calculated by the two carriers as:
和 42)e (" ■); 0) And 4 2) e ("■); 0)
如果记外码输出给内码时空编码器 1, 和内码时空编码器 1 的码字分 别记为 c ° , c )0 , 则实际上 ( 1)e (uk U); o) , 2)e («, ω; o) ) 分别为内码译码 器 1, 2计算得到的 ^。,^)。的概率信息。 If the mnemonic code is output to the inner code space-time encoder 1, and the codewords of the inner code space-time encoder 1 are respectively denoted as c °, c ) 0 , then ( 1) e (u k U); o), 2 ) e («, ω; o)) are ^ calculated by the inner code decoders 1, 2 respectively. , ^). Probability information.
在外码的整个译码过程, 都是利用似然比进行运算, 所以, 下面给出, 利用似然比推导出的后验概率似然比的表达式。 In the entire decoding process of the outer code, the likelihood ratio is used for calculation. Therefore, the expression of the posterior probability likelihood ratio derived from the likelihood ratio is given below.
此时, 除; ^( ,σί+1,Λ)的计算需要从符号级的条件概率表示转换为比 特级的似然比的表示外, 其他的计算与前面的基本相同。 At this time, except that the calculation of ^ (, σ ί + 1 , Λ ) needs to be converted from the symbol-level conditional probability representation to the bit-level likelihood ratio representation, other calculations are basically the same as the previous ones.
在前面, 已经给出 In the previous, it has been given
ϊ[ (o-k , ,yk) = pk (yk \uk,ak)p(uk =u) = pk (yk | ck )Pk (u; I) ϊ [(o- k ,, y k ) = p k (y k \ u k , a k ) p (u k = u) = p k (y k | c k ) Pk (u; I)
上式中的 )表示, 状态为 0"t, 当输入为^时的输出。 () In the above formula indicates that the state is 0 " t , and the output when the input is ^.
P(ck) P ( c k)
上式中 所以^( , +1,^) = ( »; ) ( 20) 利用符号的后验概率与似然比的转换关系式 ( 9 ), 将用似然比表示 的后验概率带入到 (20) 式中, 我们可以得到 In the above formula, ^ (, +1 , ^) = (»;) ( 20 ) Use the conversion relationship between the posterior probability of the symbol and the likelihood ratio (9) to bring the posterior probability expressed by the likelihood ratio into In (20), we can get
(u(j); 0)= max [ ak (ak ) + (σ,+1 ) + γ[ (ak , a ,yk)]
= 、, ) + d (σ"ι ) + 4 ½;Ή* (ck (ak );/)] 一 、 η [ ) + d (σ"ι ) + ·(¾;/) + Κ (σΑ );/)] (u (j); 0) = max [a k (a k ) + (σ, +1 ) + γ [(a k , a, y k )] =,,) + D ( σ "ι) + 4 ½; Ή * (c k (a k ); /)] I, η [) + d ( σ " ι) + · (¾; / ) + Κ ( σ Α ); /)]
j = ,-, (2D 上式中, j =,-, (2D In the above formula,
同理, 我们得到 Similarly, we get
Ak{c{j);0)= max [ ak (ak ) + ¾+1 (σ/£+1 ) + ^ (Μ,; /) + λ!( (ck (ak ); /) ] A k (c (j); 0) = max [a k (a k ) + ¾ +1 (σ / £ + 1 ) + ^ ( Μ ,; /) + λ ! ( (C k (a k ); /)]
人-: cA(J)( A)-l Person-: c A (J) ( A ) -l
一 max, n [ ¾ (σ^ ) + βΜ (σΑ+1 ) + ^ (Μ^; /) + ^ ( k (ak ); /) ] j' = l, ···,"。 (24) 式中 Α的迭代表达式变为: -Max, n [¾ (σ ^) + β Μ (σ Α + 1 ) + ^ (Μ ^; /) + ^ ( k (a k ); /)] j '= l, ···, ". (24) The iterative expression of A in the formula becomes:
ak+l (σΜ ) = max(a, (ak ) + k(uk;I) + Xk (ck (ak ); I) k (^k ) = max(¾+1 (ak ) + k(uk;I) + Xk (ck (ak ); /) 同时, 迭代过程中的边信息计算过程如下: a k + l (σ Μ ) = max (a, (a k ) + k (u k ; I) + X k (c k (a k ); I) k (^ k) = max (¾ +1 ( a k ) + k (u k ; I) + X k (c k (a k ); /) At the same time, the side information calculation process in the iteration process is as follows:
(uU); 0) = XA k (uU); 0) - k (uk (7); /) j = l,-,k0 ( 25 ) λ (c(j); O) = A k (c(j); O) - Xk (ck (;); ) j = h-,n0 ( 26 ) 下面我们对照图 5, 来讲解串行时空频译码的过程。 串行时空频 TURBO 编码的 LOG- MAP译码过程为: (uU); 0) = X A k (uU); 0) -k (u k (7); /) j = l,-, k 0 (25) λ (c (j); O) = A k (c (j); O)-X k (c k (;);) j = h-, n 0 (26) Let us explain the process of serial space-time-frequency decoding with reference to FIG. 5. The LOG-MAP decoding process of serial space-time-frequency TURBO coding is:
(一)载波 1和载波 2接收的信号分别经过接收匹配滤波器 1, 2进入两 个 APP计算模块, 利用 ( 12 )式, 计算得到内码时空编码器 1和 2的码字 c ',c 符号级的后-险¾¾率, 然后利用 (10) 式, 将其转换为比特级的似然 比, 分别得到全后验概率的似然比 (" ■); 0), _/ = 1,··'Α, 初始迭代时, 令 = 0,
Q. (1) The signals received by carrier 1 and carrier 2 enter the two APP calculation modules through the receiving matched filters 1, 2 respectively, and use formula (12) to calculate the codewords c ', c of the inner code space-time encoders 1 and 2. The symbol-level posterior-risk ratio is then converted to the bit-level likelihood ratio using the formula (10) to obtain the likelihood ratio of the full posterior probability ("■); 0), _ / = 1, respectively. · 'Α, at initial iteration, let = 0, Q.
(二)然后利用 (18), (19)计算边信息 41)2( );O), d );O)。 (三)将第 2 步中计算得到的内码的 2 个边信息 41)£ (J');O) ,
(2) Then use (18), (19) to calculate the side information 4 1) 2 (); O), d); O). (3) The 2 side information of the inner code calculated in step 2 4 1) £ (J '); O),
反交织作为外码 ^。^^。的先验信息 ^(^。( ; ), c 0u);D , 其中
上标 "0" 表示外码。 Deinterleaving is used as the outer code ^. ^^. A priori information ^ (^. (;), C 0 u); D, where The superscript "0" indicates a foreign code.
(四)利用 ( 23) 式计算 ( ;/), 在外码的后验似然信息计算时, Ak(uk;I)始终为零; 将得到的 4( ;/), (uk;I) , 带入到 (24) 中, 得到 λ ]);0) , j = l,-,n0 ', 然后带入到 ( 26 ) 中, 计算得到 Ι Φ·);0), = l,— ,w。; 利用 [10]式, 将边信息的似然比转换为符号的后验概率 ( ;O)(4) Use formula (23) to calculate (; /). When calculating the posterior likelihood information of the outer code, A k (u k; I) is always zero; 4 (; /), (u k ; I), brought into (24) to get λ]); 0), j = l,-, n 0 ', and then brought into (26), calculated to get Ι Φ ·); 0), = l , —, W. ; Use [10] to convert the likelihood ratio of side information to the posterior probability of the symbol (; O)
(五)将^ Ct;O)首先进行串并转换, 得到 , 分别经 过交织器 1, 2, 得到; ^)2( ;O), (¾;O), 将此交织后的信息分别作为内 码时空频编码器 1, 2的先验信息;
(5) ^ Ct ; O) is first serial-to-parallel converted to get, respectively, obtained through the interleaver 1, 2, ^ ) 2 (; O), ( ¾ ; O), and use the interleaved information as internal Prior information of code space-time frequency encoders 1, 2;
(六)在后续的迭代中, 重复 2- 5步。 (VI) in a subsequent iteration, repeating 2--5 steps.
(七)在最后一次迭代时, 输出外码的 j = l ",k0, 采用下面 的方式进行判决 k
(7) At the last iteration, output the outer code of j = l ", k 0 , and use the following method to make a decision k
然后回到 1, 进行下一帧的译码。 Then return to 1, and decode the next frame.
由比特级的 LOG- MAP译码算法推导符号级的 LOG-MAP译码算法: 在上面给出了一种比特级和符号级混合的 LOG-MAP译码算法, 在这小 节, 根据这种算法推导基于符号级的 LOG-MAP译码算法。 此时所有的概率 信息都是用符号级的后验概率表示的, 而没有用到比特级的似然比。 The symbol-level LOG-MAP decoding algorithm is derived from the bit-level LOG-MAP decoding algorithm: A bit-level and symbol-level mixed LOG-MAP decoding algorithm is given above. In this section, this algorithm is derived Symbol-level LOG-MAP decoding algorithm. At this time, all the probability information is expressed by the posterior probability at the symbol level, but no bit-level likelihood ratio is used.
利用 (17-2)计算得到内码时空编码器 1 和 2 的码字 符号级 的后验概率的边信息 4;0), p )e{uk-0), 此时不需再计算似然信息。 而 是将其反交织作为外码译码器的先验信息 , Pk(c )0- )。 Use (17-2) to calculate the side information 4 of the posterior probability of the code character number level of the inner code spatio-temporal encoders 1 and 2; 0), p ) e (u k -0).然 信息。 Natural information. Instead, the inverse interleaving is used as the prior information of the outer code decoder, Pk (c ) 0- ).
在外码译码器中, In the outer code decoder,
¾·+ι ) = max(ak (σ, ) + pk {uk; I) + pk (ck (ak );/) ¾ · + ι) = max (a k (σ,) + p k {u k ; I) + p k (c k (a k ); /)
。 .
(σΑ. ) = max(¾+1 (ak ) + Pk (uk; J) + pk (ck (ak ); /) (σ Α .) = max (¾ +1 (a k ) + Pk (u k ; J) + p k (c k (a k ); /)
ή , σ ,yk) = hpk (c; I)pk (u;I) Price (σ, y k ) = hp k (c; I) p k (u; I)
计算得到的全信息的后验概率为 (在外码的迭代中 ρ^,'Γ)始终为 零) The calculated posterior probability of the full information is (in the iteration of the outer code, ρ ^, 'Γ) is always zero)
Pt (c; O) = m^c [ ak (ak ) + ¾+1 (σΜ ) + pk (uk ;I) + pk (ck (ak ); /) ] Pt ( c ; O) = m ^ c [a k (a k ) + ¾ +1 (σ Μ ) + p k (u k ; I) + p k (c k (a k ); /)]
- max [ ak (σ, ) + βΜ (σ,+1 ) + pk (uk ;I) + pk (c, (σ, ); I) σ*¾·Ο)(σ·*:)=0
pt(c;0) = pt(c;0)-pk(c;l) -max [a k (σ,) + β Μ (σ, +1 ) + p k (u k ; I) + p k (c, (σ,); I) σ * ¾ · Ο) (σ · * :) = 0 pt (c; 0) = pt (c; 0) -p k (c; l)
将 ^ ( ;O)串并转换并交织分别作为内码时空频编码器 1, 2的先验 信息
;I) , 以后的迭代过程与小节 3中的相同。 ^ (; O) serial-to-parallel conversion and interleaving are used as a priori information of the inner code space-time frequency encoders 1, 2 ; I), the subsequent iteration process is the same as in section 3.
IV. 仿真结果和结论分析: IV. Simulation results and conclusion analysis:
下面给出采用上面所述的并行和串行时空频编译码算法时的仿真结 果, 主要的仿真参数列在表 1 中, 在下面的仿真结果中, 已经考虑了时空 频编码后的总功率与没有时空频编码的功率相同。 The simulation results when using the parallel and serial space-time-frequency coding and decoding algorithms described above are given below. The main simulation parameters are listed in Table 1. In the following simulation results, the total power and space-time-frequency coding have been considered. The power without space-time frequency coding is the same.
下面两图分别了并行时空频编码与串行时空频编码与传统的时空分组 编码 [13] [14] (扩展到两个频率上进行编码)的性能比较。 The following two figures respectively show the performance comparison between parallel space-time-frequency coding and serial space-time-frequency coding and traditional space-time block coding [13] [14] (extended to two frequencies for coding).
在图 6中, 给出了移动速度为 5km/h时串并行时空频编码的 FER性能 曲线。 从图中, 可以看出, 在高信噪比时, 串行的时空频编码性能最优。 在 10E- 4的 FER条件下, 串行的时空频编码比并行的好大约 0.5dB, 比传 统的时空分组编码的性能好大约 4dB, 而且随着信噪比的提高, 这种差距 会越来越大。在 10E-5的 FER时, 串行的时空频编码比并行的好大约 1.5dB。 In Figure 6, the FER performance curve of serial-parallel space-time-frequency coding at 5km / h is given. From the figure, it can be seen that at high SNR, the performance of serial space-time-frequency coding is optimal. Under the FER condition of 10E-4, the serial space-time frequency coding is about 0.5dB better than parallel, and the performance is about 4dB better than the traditional space-time block coding, and this gap will grow as the signal-to-noise ratio increases. Bigger. At 10E-5 FER, the serial space-time frequency coding is about 1.5dB better than parallel.
在图 7中, 给出了移动速度为 60km/h时串并行时空频编码的 FER性能
曲线。 从图中可以看出, 在高信噪比时, 仍然是串行的时空频编码性能最 优。 在 10E-4的 FER条件下, 串行的时空频编码比并行的好大约 0. 5dB, 比传统的时空分组编码的性能好大约 1. 5dB。 在 10E- 5 的 FER时, 串行的 时空频编码比并行的好大约 IdB, 比比传统的时空分组编码好大约 2. 5dB。 In Figure 7, the FER performance of serial-parallel space-time-frequency coding at a moving speed of 60 km / h is given. Curve. It can be seen from the figure that at high SNR, the serial space-time-frequency coding performance is still optimal. 5dB。 Under 10E-4 FER conditions, serial space-time frequency coding is better than parallel about 0.5dB, better than the traditional space-time block coding performance is about 1.5dB. 5dB。 10E-5 FER, serial space-time frequency coding is better than parallel about IdB, better than the traditional space-time block coding about 2.5dB.
在前面的分析中, 可以知道串行的时空频译码, 相对于并行的译码 而言, 略微复杂, 但是相对于传统的时空分组编码要复杂的多。 仿真结果 证明了, 串行的时空频增益是最大的, 因此性能的改善是以译码复杂度的 增加为代价的。 In the previous analysis, we can know that serial time-space-frequency decoding is slightly more complicated than parallel decoding, but it is much more complicated than traditional space-time block coding. The simulation results show that the serial space-time-frequency gain is the largest, so the performance improvement comes at the cost of increased decoding complexity.
本发明将时空编码扩展到频域, 可以为时空编码提供更灵活的设计空 间, 更大的提高系统的频语效率和获得更大的分集增益和编码增益。 The present invention extends space-time coding to the frequency domain, which can provide more flexible design space for space-time coding, greatly improve the frequency speech efficiency of the system, and obtain greater diversity gain and coding gain.
以上具体实施方式仅用于说明本发明, 而非用于限定本发明。 本发明涉及的参考文献如下: The above specific implementations are only used to illustrate the present invention, but not intended to limit the present invention. The references involved in the present invention are as follows:
[l].L.R.Ba l,J.Cocke,F.Jelinek, and J.Raviv, "Optimal decoding of linear codes for minimizing symbol error rate," IEEE Transactions on Information Theory, pp.284- 287,Mar.l974. [l] .LRBa l, J.Cocke, F.Jelinek, and J.Raviv, "Optimal decoding of linear codes for minimizing symbol error rate," IEEE Transactions on Information Theory, pp.284- 287, Mar.l974.
[2] .P.Robertson and T.Worz, "Bandwidth-efficient turbo trellis-coded modulation using punctured component codes," IEEE Journal on Selected Areas in Communications, vol.l6,pp.206-218,Feb.l998. [2] .P. Robertson and T. Worz, "Bandwidth-efficient turbo trellis-coded modulation using punctured component codes," IEEE Journal on Selected Areas in Communications, vol.16, pp.206-218, Feb.l998.
[3].Johan Hokfelt, Ove Edfors, and Torleiv Maseng, "On the Theory and Performance of Trellis Termination Methods for Turbo Codes", IEEE JSAC, vol.19, No.5, May 2001. [3]. Johan Hokfelt, Ove Edfors, and Torleiv Maseng, "On the Theory and Performance of Trellis Termination Methods for Turbo Codes", IEEE JSAC, vol.19, No.5, May 2001.
[4].A.S.Barbulescu and S.S.Pietrobon, "Termination for turbo encoders," in 17th Biennial Symp. Communications, Kingston, ON, Canada, May 1994,PP.389-392. [4] .AS Barbulescu and SSPietrobon, "Termination for turbo encoders," in 17 th Biennial Symp. Communications, Kingston, ON, Canada, May 1994, PP. 389-392.
[5].M.Hattori, J.Murayama, and RJ.McElice, "Pseudo-random and self-terniinating and interleavers for turbo codes," in Information Theory Workshop, San Diego, CA, Feb.1998.pp.9-10. [5] .M. Hattori, J.Murayama, and RJ.McElice, "Pseudo-random and self-terniinating and interleavers for turbo codes," in Information Theory Workshop, San Diego, CA, Feb. 1998.pp. 9- 10.
[6].Jung.P., and NABHAN.M, " Dependence of the error performance of turbo codes on the interleaver structure in short frame transmission systems,", Electron. Lett., 1994,30,pp.287-288. [6] .Jung.P., And NABHAN.M, "Dependence of the error performance of turbo codes on the interleaver structure in short frame transmission systems,", Electron. Lett., 1994, 30, pp.287-288.
[7]Jung.p, and NABHAN.M, " Performance evaluation of turbo codes for short frame transmission systems," Electron丄 ett., 1994,30,pp.l 11-113. [7] Jung.p, and NABHAN.M, "Performance evaluation of turbo codes for short frame transmission systems," Electron 丄 ett., 1994, 30, pp.l 11-113.
[8]. J.Hagenauer, P.Robertson, and L.Papke, "Iterative(Turbo) decoding of systematic convolutional codes with the MAP and SOVA algorithms," in Proceedings of the ITG Conference,VDE— Verlag, Oct.1994. [8]. J. Hagenauer, P. Robertson, and L. Papke, "Iterative (Turbo) decoding of systematic convolutional codes with the MAP and SOVA algorithms," in Proceedings of the ITG Conference, VDE— Verlag, Oct. 1994.
[9]. S.Benedetto, G.Montorsi, D.Divsalar, and F.Pollara, " A soft-input soft-output maximum a posteriori (MAP) to decode parallel and serial concatenated codes," in The Telecommunications and Data Acquisition Progress Report 42-127, pp.1-20, Jet Propulsion Laboratory, Nov.1996.
[10]. S.Benedetto, G.Montorsi, D.Divsalar, and F.Pollara, "A soft-input soft -output APP module for iterative decoding of concatenated codes," IEEE Communications Letters, vol.1, pp.22-243 Jan.1997. [9]. S. Benedetto, G. Montorsi, D. Divsalar, and F. Pollara, "A soft-input soft-output maximum a posteriori (MAP) to decode parallel and serial concatenated codes," in The Telecommunications and Data Acquisition Progress Report 42-127, pp.1-20, Jet Propulsion Laboratory, Nov. 1996. [10]. S. Benedetto, G. Montorsi, D. Divsalar, and F. Pollara, "A soft-input soft -output APP module for iterative decoding of concatenated codes," IEEE Communications Letters, vol.1, pp.22 -24 3 Jan. 1997.
[11]. S.Benedetto, D.Divsalar, G.Montorsi, and F.Pollara, "Serial concatenated trellis coded modulation with iterative decoding: Design and Performance," in IEEE Communications Theory Mini Conference, pp.38-43,1997. [11]. S. Benedetto, D. Divsalar, G. Montorsi, and F. Pollara, "Serial concatenated trellis coded modulation with iterative decoding: Design and Performance," in IEEE Communications Theory Mini Conference, pp.38-43, 1997 .
• [12].S.Benedetto? D.Divsalar, G.Montorsi, and F.Pollara, "Serial concatenated trellis coded modulation with iterative decoding: Design and Performance," in International Symposium on Information Theory, (Ulm, Germany.), IEEE, Juiie-July?1997. • [12] .S . Benedetto ? D. Divsalar, G. Montorsi, and F. Pollara, "Serial concatenated trellis coded modulation with iterative decoding: Design and Performance," in International Symposium on Information Theory, (Ulm, Germany.) , IEEE, Juiie-July ? 1997.
[13].V.Tarokh5N.Seshadri3A.R.Calderbank5 "Space-Time Codes for High Data Rate Wireless Communication: Performance Criterion and Code Construction", IEEE Trans. Inforai.Theory?vol.44,No.2? March 1998. [13] .V.Tarokh 5 N.Seshadri 3 ARCalderbank 5 "Space-Time Codes for High Data Rate Wireless Communication: Performance Criterion and Code Construction", IEEE Trans. Inforai.Theory ? Vol.44, No. 2 ? March 1998 .
[14]. S.Alamouti, "A simple transmit diversity technique for wireless communications/' IEEE JSAC, vol 14, No,8,1451-1458,OcU998. [14]. S. Alamouti, "A simple transmit diversity technique for wireless communications / 'IEEE JSAC, vol 14, No, 8,1451-1458, OcU998.
[ 15] . J.Hagenauer3Elke Offer, and Lutz Papke, "Iterative Decoding of Binary Block and Convolutional Codes", IEEE Trans on Information Theory, Vol,42,No.2 March 1996.
[15]. J. Hagenauer 3 Elke Offer, and Lutz Papke, "Iterative Decoding of Binary Block and Convolutional Codes", IEEE Trans on Information Theory, Vol, 42, No. 2 March 1996.