CN101488759B - Decoding method for MIMO OFDM system low density correcting code - Google Patents

Abstract

The invention provides a decoding method for low-density check code of multi-input multi-output orthogonal frequency division multiplexing system. On the basis of the log-likelihood ratio belief propagation principle, the method is characterized in that log-likelihood ratio belief propagation decoding of the low-density check code is realized by a coordinate rotation digital computer structure, a shifter, and correlated operation of an adder, a multiplier and a divider. The decoding method further realizes multiplexing of the coordinate rotation digital computer structure by a selector so that the decoding structure of the low-density check code is simpler.

Description

The interpretation method of MIMO OFDM system low density correcting code

Technical field

The present invention relates to channel decoding method in the modern digital communication; Especially a kind of interpretation method of MIMO OFDM system low density correcting code; Be the realization of LDPC (low-density check) sign indicating number LLRBP (log-likelihood ratio belief propagation) algorithm, wherein comprise and utilize simple circuit configuration to realize complicated logarithmic function, hyperbolic tangent function and inverse function thereof.

Background technology

The space multiplexing technique of MIMO (multiple-input and multiple-output) system can realize high speed data transfer, but presents serious frequency optionally the time when channel, and the mimo system transmission performances will sharply descend.OFDM (OFDM) system can resist frequency selective fading effectively through the method that channel distribution is become a plurality of orthogonal sub-channels.So, mimo system and ofdm system combined carry out transfer of data and can improve spectrum efficiency greatly.Based on this, the scheme that mimo system combines with ofdm system becomes the research focus of broadband wireless communications.Under the multipath transmission environment; Although most subcarriers can carry out correct detection all the time in the ofdm system; But the performance of system still can receive influence and the rapid deterioration of deep fade because of the parton carrier wave; A kind of important resolution is that error correction coding is applied in the system, becomes first-selection as the LDPC sign indicating number of the research focus of error correction coding because of its excellent performance near shannon limit.The MIMO-OFDM systematic schematic diagram is as shown in Figure 1, and involved in the present invention is exactly the interpretation method of dotted portion.

Summary of the invention

The interpretation method that the purpose of this invention is to provide a kind of MIMO OFDM system low density correcting code; This interpretation method is on the basis based on the log-likelihood ratio belief propagation algorithm; Utilization CORDIC (CORDIC) has calculated needed logarithmic function, hyperbolic tangent function and inverse function thereof in the decoding, is a kind of advantages of simplicity and high efficiency LDPC interpretation method.

The inventive method is achieved in that

The interpretation method of MIMO OFDM system low density correcting code; Based on log-likelihood ratio belief propagation principle; The log-likelihood ratio belief propagation comprise initialization, variable node to check-node transmit information, check-node transmits information, decoding judgement to variable node; It is characterized in that: the hyperbolic tangent function from check-node to variable node and the inverse function thereof that transmit information for initialized logarithmic function and; Utilize the CORDIC structure; The angular discretization of the rotation that above-mentioned function is related to is a fixed value, utilizes these fixed values progressively to deflect into the angle of required rotation then, obtains the difference equation of CORDIC structure;

Wherein, In the calculating of the logarithmic function during initialization; Channel information is exported behind divider successively and is obtained a vector through multiplier, adder and subtracter; Be the CORDIC structure under the vector pattern of initial vector with this vector, be defined as first kind of CORDIC structure, the output result of this structure is the check-node initialization information;

When check-node transmits information to variable node; Earlier variable node information is moved to right through shift unit; Forming a vector, is coordinate rotation digital computer algorithm under the rotary mode of initial vector with this vector again, is defined as second kind of CORDIC structure; Output forms a vector again through multiplier through divider with the component in this structure output vector; Be coordinate rotation digital computer algorithm structure under the vector pattern of initial vector with this vector, be defined as the third CORDIC structure, the component in this structure output vector is moved to left through shift unit promptly obtains the information that check-node passes to variable node.

Specifically according to the following steps:

At first realize the initialization of the check-node information of log-likelihood ratio belief propagation algorithm through first kind of CORDIC structure, adder, subtracter, multiplier and divider:

${\mathrm{\Λ}}_c\left(c_n\right)=\ln \frac{P(x_n=1|y_n)}{P(x_n=0|y_n)};$

Realize that through adder variable node transmits information to check-node in the loe-density parity-check code then:

${\mathrm{\Λ}}_{n\→m}^{\left(k\right)}\left(c_n\right)={\mathrm{\Λ}}_c\left(c_n\right)+\underset{u\∈M\left(n\right)-\left\{m\right\}}{\mathrm{\Σ}}{L}_{n\←u}^{(k-1)}\left(c_n\right);$

Then utilize second kind to realize that with the third CORDIC structure, shift unit, multiplier and divider check-node transmits information to variable node:

$L_{n\←m}^{\left(k\right)}\left(c_n\right)=2\arctan h\left(\underset{u\∈N\left(m\right)-\left\{n\right\}}{\mathrm{\Π}}\tanh \frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2}\right);$

Utilize adder and selector to realize the decoding judgement at last:

${\mathrm{\Λ}}^{\left(k\right)}\left(c_n\right)={\mathrm{\Λ}}_c\left(c_n\right)+\underset{u\∈M\left(n\right)}{\mathrm{\Σ}}{L}_{n\←u}^{\left(k\right)}\left(c_n\right)$

${\hat{c}}_n=\left\{\begin{array}{cc}0& {\mathrm{\Λ}}^{\left(k\right)}\left(c_n\right)>0\\ 1& {\mathrm{\Λ}}^{\left(k\right)}\left(c_n\right)\≤0\end{array}\right.;$

And judge whether the code word after the judgement is correct, promptly $H{\hat{c}}^T=0$ Whether set up, if correctly then iteration end, otherwise continue iteration, up to maximum iteration time.

Initial method is following:

(1) with channel information, i.e. channel transition probability P (x _n=1|y _n) and P (x _n=0|y _n) be output as through after the divider

(2) with channel transition probability ratio

Vector through obtaining after multiplier, adder and the subtracter does $({\left[\frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}\right]}^2+1,{\left[\frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}\right]}^2-\mathrm{1,0});$

(3) with vector $({\left[\frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}\right]}^2+1,{\left[\frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}\right]}^2-\mathrm{1,0})$ As the initial vector of coordinate rotation digital computer algorithm vector pattern, form first kind of CORDIC structure, component z in the output vector of this structure _nBe

Variable node is existing method to check-node transmission information, realizes through following steps:

With Λ _c(c _n), L _{N ← u} ^(k-1)(c _n), (u ∈ M (n)-and m}) through obtaining the information Λ of variable node after the adder to the check-node transmission _c(c _n), promptly realize

${\mathrm{\Λ}}_{n\→m}^{\left(k\right)}\left(c_n\right)={\mathrm{\Λ}}_c\left(c_n\right)+\underset{u\∈M\left(n\right)-\left\{m\right\}}{\mathrm{\Σ}}{L}_{n\←u}^{(k-1)}\left(c_n\right)$

Check-node transmits information to variable node and realizes by following steps:

(1) with variable node information Λ _{U → m} ^(k)(c _u) move to right through shift unit and to realize multiply by

Promptly realize

Thereby form vector

(2) with

Be the initial vector of coordinate rotation digital computer algorithm rotary mode structure, obtain second kind of CORDIC structure, the component in the output vector in this structure $x_n=A_n\mathrm{Cosh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2},$ $y_n=A_n\mathrm{Sinh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2};$

(3) x above _n, y _nObtain through after the divider $\frac{y_n}{x_n}=\frac{A_n\mathrm{Sinh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2}}{A_n\mathrm{Cosh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2}}=\mathrm{Tanh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2};$

(4)

Obtain through behind the multiplier again $\underset{u\∈N\left(m\right)-\left\{n\right\}}{\mathrm{\Π}}\mathrm{Tanh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2},$ Form vector $(1,\underset{u\∈N\left(m\right)-\left\{n\right\}}{\mathrm{\Π}}\mathrm{Tanh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2},0);$

(5) with $(1,\underset{u\∈N\left(m\right)-\left\{n\right\}}{\mathrm{\Π}}\mathrm{Tanh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2},0)$ Be the initial vector of coordinate rotation digital computer algorithm vector pattern structure, obtain the third CORDIC structure, component in the output vector of this structure $z_n=\mathrm{Arctan}h\left(\underset{u\∈N\left(m\right)-\left\{n\right\}}{\mathrm{\Π}}\mathrm{Tanh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2}\right);$

(6) will $z_n=\mathrm{Arctan}h\left(\underset{u\∈N\left(m\right)-\left\{n\right\}}{\mathrm{\Π}}\mathrm{Tanh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2}\right)$ The process shift unit moves to left and obtains the information that check-node passes to variable node $L_{n\←m}^{\left(k\right)}\left(c_n\right)=2\mathrm{Arctan}h\left(\underset{u\∈N\left(m\right)-\left\{n\right\}}{\mathrm{\Π}}\mathrm{Tanh}\frac{{\mathrm{\Λ}}_{u\→m}^{\left(k\right)}\left(c_u\right)}{2}\right).$

Utilize adder and selector to realize the decoding judgement at last:

${\mathrm{\Λ}}^{\left(k\right)}\left(c_n\right)={\mathrm{\Λ}}_c\left(c_n\right)+\underset{u\∈M\left(n\right)}{\mathrm{\Σ}}{L}_{n\←u}^{\left(k\right)}\left(c_n\right)$

${\hat{c}}_n=\left\{\begin{array}{cc}0& {\mathrm{\Λ}}^{\left(k\right)}\left(c_n\right)\≤0\\ 1& {\mathrm{\Λ}}^{\left(k\right)}\left(c_n\right)\≤0\end{array}\right.;$

And judge whether the code word after the judgement is correct, promptly ${H\hat{c}}^T=0$ Whether set up, if correctly then iteration end, otherwise continue iteration, up to maximum iteration time.

Advantage of the present invention and remarkable result: displacement and plus and minus calculation are only used because the CORDIC structure is very simple in (1), so the present invention has adopted the CORDIC structure, make that the complexity of LDPC sign indicating number LLR BP interpretation method is lower; (2) when utilizing CORDIC Structure Calculation logarithmic function, hyperbolic tangent function and inverse function, utilized the multiplex technique of circuit structure, made that the complexity of LDPC interpretation method is lower; (3) the improvement algorithm of many simplification is arranged based on CORDIC, the complexity of interpretation method of the present invention can also further reduce thereupon; (4) this interpretation method is the realization of LLR BP decoding principle, because LLR BP decoding principle is the optimum soft-decision algorithm of LDPC code performance at present, so the performance of this interpretation method is the highest.

Description of drawings

Fig. 1 is MIMO-OFDM system transmitting terminal and receiving terminal block diagram;

Fig. 2 is the flow chart of the LLR BP decoding algorithm of LDPC sign indicating number;

Fig. 3 is the multiplex circuit of two kinds of mode of operations of CORDIC structure of the present invention;

Fig. 4 is a LDPC sign indicating number decoding architecture of the present invention.

Embodiment

At first set forth the principle of the inventive method:

LLR BP algorithm is that best performance soft of LDPC sign indicating number declared decoding algorithm, and the flow chart of LLR BP decoding algorithm is as shown in Figure 2, n=1 wherein, and 2 ..., N; M=1,2 ..., M, N are the number of LDPC sign indicating number variable node, M is the number of LDPC code check node, N _StopMaximum iteration time for LLR BP decoding.

In LLR BP decoding algorithm; Used logarithmic function during initialization; When giving variable node, check-node transfer reliability information used hyperbolic tangent function and inverse function thereof; Traditional implementation method to these functions is Taylor expansion or look-up table, and it is also complicated so just to exist bigger error, hardware to realize.Here utilize the CORDIC structure to realize these functional operation, because CORDIC is simple in structure, precision is higher, so can well realize LLR BP decoding algorithm.

The thought of CORDIC structure is to be fixed value with the angular discretization of rotating earlier, utilizes these fixed values progressively to deflect into the angle of required rotation then.The difference equation of realizing the CORDIC structure of hyperbolic functions is:

$\left\{\begin{array}{c}{x}_{i+1}=x_i-{\mathrm{\σ}}_i{y}_i{2}^{-i}\\ y_{i+1}=y_i-{\mathrm{\σ}}_i{x}_i{2}^{-i}\\ z_{i+1}=z_i-{\mathrm{\σ}}_i\mathrm{arctanh}\left(2^{-i}\right)\end{array}\right.$

The CORDIC structure has two kinds of mode of operations, i.e. rotary mode and vector pattern.As far as rotary mode, σ _i=sign (z _i), the iteration result of difference equation is:

$\left\{\begin{array}{c}{x}_n=A_{n-1}[x_0\mathrm{<figure-callout\; id="0"\; label="cosh\; z"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">cosh</figure-callout>}{z}_{}{}_0+{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">y</figure-callout>}}_0\sinh z_0]\\ y_n=A_{n-1}[{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">y</figure-callout>}}_0\mathrm{<figure-callout\; id="0"\; label="cosh\; z"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">cosh</figure-callout>}{z}_{}{}_0+x_0\sinh z_0]\\ z_n=0\\ A_n=\underset{i=0}{\overset{n-1}{\mathrm{\&Pi;}}}\sqrt{1-2^{-2i}}\&ap;0.8\end{array}\right.$

For vector pattern, σ _i=-sign (y _i), the iteration result of difference equation is:

$\left\{\begin{array}{c}{x}_n=A_{n-1}\sqrt{x_0^2-{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">y</figure-callout>}}_0^2}\\ y_n=0\\ z_n={\mathrm{<figure-callout\; id="0"\; label="z"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">z</figure-callout>}}_0+\mathrm{arctanh}\frac{y_0}{x_0}\\ A_n=\underset{i=0}{\overset{n-1}{\mathrm{\&Pi;}}}\sqrt{1-2^{-2i}}\&ap;0.8\end{array}\right.$

Two kinds of mode of operations according to the CORDIC structure are known, can carry out multiplexing to these two kinds of mode of operations through selector.Streamline implementation structure after multiplexing is as shown in Figure 3, wherein is that dotted portion is for multiplexing and selector that on the basis of traditional CORDIC pipeline organization, increase.

Iteration result according to top rotary mode and vector pattern can know, as far as rotary mode, if initial value is made as (x ₀, y ₀, z ₀)=(0,1, z ₀), then

$\frac{y_n}{x_n}=\frac{A_n\sinh z_0}{A_{\mathrm{<figure-callout\; id="0"\; label="n\; cosh\; z"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">n</figure-callout>}}\cosh z_{}{}_0}={\mathrm{<figure-callout\; id="0"\; label="tanh\; z"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">tanh</figure-callout>}z}_0$

The CORDIC structure of remembering this specific initial value is added divider to realize tanh z ₀Structure be CORDIC-2 (be aforementioned second kind of CORDIC structure and add divider); As far as vector pattern, initial value is made as (x ₀, y ₀, z ₀)=(1, y ₀, 0), z then _n=arc tanh y ₀, the CORDIC structure of remembering this specific initial value is CORDIC-3 (being aforementioned the third CORDIC structure); Analysis above similar is in vector pattern, if initial value is made as (x ₀, y ₀, z ₀)=(a ²+ 1, a ²-1,0), z then _n=ln a.The CORDIC structure of remembering this specific initial value is CORDIC-1 (being aforementioned first kind of CORDIC structure).So; CORDIC-2 and CORDIC-3 combine and add multiplier and shift unit (shift unit moves to left and realizes multiply by 2; Shift unit moves to right and realizes multiply by

) can realize that LLR BP algorithm check-node transfer reliability information is to variable node, promptly

$L_{n\&LeftArrow;m}^{\left(k\right)}\left(c_n\right)=2\arctan h\left(\underset{u\&Element;N\left(m\right)-\left\{n\right\}}{\mathrm{\&Pi;}}\tanh \frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2}\right)$

CORDIC-1 can be used for realizing Λ in the computing of LLR BP algorithm initialization _c(c _n) calculating, promptly

${\mathrm{\&Lambda;}}_c\left(c_n\right)=\ln \frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}$

Based on top analysis, ldpc decoder structure (variable node transfer reliability information is given the structure of check-node part because of simply omitting) as shown in Figure 4.

LLR BP algorithm and accompanying drawing above the following reference describe.

At first being initialization, mainly is Λ here _c(c _n) calculating, promptly

${\mathrm{\&Lambda;}}_c\left(c_n\right)=\ln \frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}$

Calculate Λ _c(c _n) having used CORDIC-1, the mode of operation of CORDIC-1 is a vector pattern, can in Fig. 3, make CORDIC be operated under the vector pattern through selector, and the initial value that CORDIC-1 is set is:

$({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">y</figure-callout>}}_0,z_0)=({\left[\frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}\right]}^2+1,{\left[\frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}\right]}^2-\mathrm{1,0})$

Realize Λ _c(c _n) the circuit implementation structure shown in Fig. 4 (the first half), the channel channel transition probability information of self-channel is delivered among the CORDIC-1 at first in the future, calculates Λ _c(c _n) as the initialization of LDPC sign indicating number decoding.

Be that variable node transfer reliability information is given check-node then, promptly realize

${\mathrm{\&Lambda;}}_{n\&RightArrow;m}^{\left(k\right)}\left(c_n\right)={\mathrm{\&Lambda;}}_c\left(c_n\right)+\underset{u\&Element;M\left(n\right)-\left\{m\right\}}{\mathrm{\&Sigma;}}{L}_{n\&LeftArrow;u}^{(k-1)}\left(c_n\right)$

This step is simple relatively, only needs adder to get final product, because of it realizes simple in structure so in Fig. 4, omit.

Then be a most important step, check-node transfer reliability information is given variable node, promptly realizes

$L_{n\&LeftArrow;m}^{\left(k\right)}\left(c_n\right)=2\arctan h\left(\underset{u\&Element;N\left(m\right)-\left\{n\right\}}{\mathrm{\&Pi;}}\tanh \frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2}\right)$

Realize that needed hardware circuit is CORDIC-2, initial value that a realization multiply by

CORDIC-2 does to need shift unit to move to left

$({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">y</figure-callout>}}_0,z_0)=(\mathrm{0,1},\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2})$

Utilize CORDIC-2 to realize Afterwards, also to realize with multiplier $\underset{u\&Element;N\left(m\right)-\left\{n\right\}}{\mathrm{\&Pi;}}\mathrm{Tanh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2},$ The initial value of final CORDIC-3 does

$({\mathrm{<figure-callout\; id="0"\; label="x"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">x</figure-callout>}}_0,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H200910024440XE00021.png,H200910024440XE00031.png"\; state="\{\{state\}\}">y</figure-callout>}}_0,z_0)=(1,\underset{u\&Element;N\left(m\right)-\left\{n\right\}}{\mathrm{\&Pi;}}\tanh \frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2},0)$

Thereby the output valve of CORDIC-3 does $\mathrm{Arctan}h\left(\underset{u\&Element;N\left(m\right)-\left\{n\right\}}{\mathrm{\&Pi;}}\mathrm{Tanh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2}\right),$ Need again shift unit move to left (being equivalent to multiply by 2) be L _{N ← m} ^(k)(cn), calculate L _{N ← m} ^(k)(cn) circuit implementation structure figure is shown in Fig. 4 (the latter half).

Final step is that decoding judgement and decoding finish, and the circuit of these functions is realized very simple, and is prior art, so in Fig. 4, omit.

Though in this manual, a kind of most preferred embodiment according to the present invention has carried out detailed elaboration to the present invention, and the professional in technical field still can make various improvement and modification to the present invention.Therefore, additional claim attempt comprises all improvement modification that belong to flesh and blood of the present invention.

Claims (5)

1. the interpretation method of MIMO OFDM system low density correcting code; Based on log-likelihood ratio belief propagation principle; The log-likelihood ratio belief propagation comprise initialization, variable node to check-node transmit information, check-node transmits information, decoding judgement to variable node; It is characterized in that: the hyperbolic tangent function from check-node to variable node and the inverse function thereof that transmit information for initialized logarithmic function and; Utilize CORDIC; The angular discretization of the rotation that above-mentioned function is related to is a fixed value, utilizes these fixed values progressively to deflect into the angle of required rotation then, obtains the difference equation of CORDIC;

Wherein, In the calculating of the logarithmic function during initialization; Channel information is exported behind divider and is passed through multiplier successively, and adder and subtracter obtain a vector, is the CORDIC under the vector pattern of algorithm of initial vector with carrying out with this vector; Be defined as first kind of CORDIC, the output result of this first kind of CORDIC is the check-node initialization information;

When check-node transmits information to variable node; Earlier variable node information is moved to right through shift unit; Form a vector; With carrying out with this vector is the CORDIC under the rotary mode of initial vector algorithm; Being defined as second kind of CORDIC, the component in this second kind of CORDIC output vector is formed a vector through divider output through multiplier again, is the CORDIC under the vector pattern of algorithm of initial vector with carrying out with this vector; Be defined as the third CORDIC, the component in this third CORDIC output vector is moved to left through shift unit promptly obtains the information that check-node passes to variable node.

2. according to the interpretation method of the said MIMO OFDM system low density correcting code of claim 1; It is characterized in that: through selector is set; The rotary mode of CORDIC and vector pattern are carried out multiplexing so that in a CORDIC, realize above-mentioned first, second, third kind of CORDIC.

3. according to the interpretation method of claim 1 or 2 said MIMO OFDM system low density correcting codes, concrete steps are following:

At first realize the initialization of the check-node information of log-likelihood ratio belief propagation algorithm through first kind of CORDIC, adder, subtracter, multiplier and divider:

${\mathrm{\&Lambda;}}_c\left(c_n\right)=\ln \frac{P(x_n=1|y_n)}{P(x_n=0|y_n)};$

Realize that through adder variable node transmits information to check-node in the loe-density parity-check code then:

${\mathrm{\&Lambda;}}_{n\&RightArrow;m}^{\left(k\right)}\left(c_n\right)={\mathrm{\&Lambda;}}_c\left(c_n\right)+\underset{u\&Element;M\left(n\right)-\left\{m\right\}}{\mathrm{\&Sigma;}}{L}_{n\&LeftArrow;u}^{(k-1)}\left(c_n\right);$

Then utilize second kind to realize that with the third CORDIC, shift unit, multiplier and divider check-node transmits information to variable node:

$L_{n\&LeftArrow;m}^{\left(k\right)}\left(c_n\right)=2\arctan h\left(\underset{u\&Element;N\left(m\right)-\left\{n\right\}}{\mathrm{\&Pi;}}\tanh \frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2}\right);$

Utilize adder and selector to realize the decoding judgement at last:

${\mathrm{\&Lambda;}}^{\left(k\right)}\left(c_n\right)={\mathrm{\&Lambda;}}_c\left(c_n\right)+\underset{u\&Element;M\left(n\right)}{\mathrm{\&Sigma;}}{L}_{n\&LeftArrow;u}^{\left(k\right)}\left(c_n\right)$

${\hat{c}}_n=\left\{\begin{array}{cc}0& {\mathrm{\&Lambda;}}^{\left(k\right)}\left(c_n\right)>0\\ 1& {\mathrm{\&Lambda;}}^{\left(k\right)}\left(c_n\right)\&le;0\end{array}\right.;$

And judge whether the code word after the judgement is correct; Promptly whether

sets up; If correctly then iteration finishes; Otherwise the continuation iteration is up to maximum iteration time.

4. according to the interpretation method of the said MIMO OFDM system low density correcting code of claim 3, its initial method is following:

(1) with channel information, i.e. channel transition probability P (x _n=1|y _n) and P (x _n=0|y _n) be output as through after the divider $\frac{P(x_n=1|y_n)}{P(x_m=0|y_n)};$

(2) with channel transition probability ratio

Vector through obtaining after multiplier, adder and the subtracter does $({\left[\frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}\right]}^2+1,{\left[\frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}\right]}^2-\mathrm{1,0});$

(3) with vector $({\left[\frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}\right]}^2+1,{\left[\frac{P(x_n=1|y_n)}{P(x_n=0|y_n)}\right]}^2-\mathrm{1,0})$ As the initial vector of coordinate rotation digital computer algorithm vector pattern, form first kind of CORDIC, component z in the output vector of this first kind of CORDIC _nBe

5. according to the interpretation method of the said MIMO OFDM system low density correcting code of claim 3, its check-node is realized through following steps to variable node transmission information:

(1) with variable node information

Move to right through shift unit and to realize multiply by Promptly realize

Thereby form vector $(\mathrm{0,1},\frac{{\mathrm{\&Delta;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2});$

(2) with

Be the initial vector of coordinate rotation digital computer algorithm rotary mode, obtain second kind of CORDIC, the component in the output vector in this second kind of CORDIC $x_n=A_n\mathrm{Cosh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2},$ $y_n=A_n\mathrm{Sinh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2};$

(3) x above _n, y _nObtain through after the divider $\frac{y_n}{x_n}=\frac{A_n\mathrm{Sinh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2}}{A_n\mathrm{Cosh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2}}$ $=\mathrm{Tanh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2};$

(4) $\mathrm{Tanh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2}$ Obtain through behind the multiplier again $\underset{u\&Element;N\left(m\right)-\left\{n\right\}}{\mathrm{\&Pi;}}\mathrm{Tanh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2},$ Form vector $(1,\underset{u\&Element;N\left(m\right)-\left\{n\right\}}{\mathrm{\&Pi;}}\mathrm{Tanh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2},0);$

(5) with

Be the initial vector of coordinate rotation digital computer algorithm vector pattern, obtain the third CORDIC, component in the output vector of this third CORDIC $z_n=\mathrm{Arctanh}\left(\underset{u\&Element;N\left(m\right)-\left\{n\right\}}{\mathrm{\&Pi;}}\mathrm{Tanh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2}\right);$

(6) will

The process shift unit moves to left and obtains the information that check-node passes to variable node $L_{n\&LeftArrow;m}^{\left(k\right)}\left(c_n\right)=2\mathrm{Arctanh}\left(\underset{u\&Element;N\left(m\right)-\left\{n\right\}}{\mathrm{\&Pi;}}\mathrm{Tanh}\frac{{\mathrm{\&Lambda;}}_{u\&RightArrow;m}^{\left(k\right)}\left(c_u\right)}{2}\right).$

Images