CN101931417B - Anti-multipath interference LDPC decoder - Google Patents

Abstract

The invention discloses an anti-multipath interference LDPC decoder which comprises a decoding module and a time-varying FIR filter module for inhibiting the multipath interference. The time-varying FIR filter module in the decoder of the invention is independent from the LDPC decoding module and is taken as a preprocessor of the LDPC decoder to eliminate the most of the multipath interference before the LDPC decodes, thus the LDPC decoder can acquire good BER decoding performance and perform right coding on the signals under the environment of the multipath interference.

Description

A kind of anti-multipath interference LDPC decoder

Technical field

The present invention relates to the decryption technology in the message transmitting procedure, relate in particular to a kind of anti-multipath interference LDPC decoder.

Background technology

The decoder of existing loe-density parity-check code (low-density parity-check codes, LDPC) can't be worked under multipath channel, presents the very poor error rate (bit error rate, BER) performance, and its BER may be greater than 25%.On the other hand, the performance of LDPC long code in some cases even be better than Turbo code, its BER curve only departs from Shannon limit less than 0.1dB.Under the condition of bandwidth or transmission rate permission, introduce the BER performance that the LDPC encoding and decoding can improve system greatly.But, the at present design of LDPC decoder is not considered to eliminate multipath and is disturbed (multipath interference, MPI) applied environment, the excellent properties of LDPC code only limit under additive white Gaussian noise (additive white Gaussian noise, the AWGN) channel circumstance.

For example, in experiment once, be the BER performance of checking LDPC code under the MPI applied environment, consider that code length is 378, code rate is respectively 1/2 and 1/6 random LDPC code and passes through MPI.The MPI index is: main footpath number is that 3, MPC (channel impulse response, ChannelImpulse Response) impulse response length is 8, channel impulse response h=[0.4,0,0.3,0,0.2,0,0,0.1].Obtain BER curve as shown in Figure 1, can find out: even adopt 1/6 low code rate, the falling characteristic as the waterfall in the awgn channel does not appear in the BER curve of traditional LDPC decoder yet.

Therefore, traditional LDPC decoder is under the MPI environment, and BER is seriously deteriorated, and the LDPC decoder can't be to the signal correct decoding under the MPI channel.

Summary of the invention

The object of the present invention is to provide a kind of anti-multipath interference LDPC decoder.Based on the present invention, can overcome traditional LDPC decoder under the MPI environment, BER is seriously deteriorated, and the LDPC decoder can't be to the problem of the signal correct decoding under the MPI channel.

A kind of anti-multipath interference LDPC decoder of the present invention comprises decoder module and filtration module, described decoder module cascade filtering module, and described filtration module suppresses for multipath is disturbed.

Above-mentioned anti-multipath interference LDPC decoder becomes the FIR filtration module when preferred described filtration module is, the channel signal that this module receives be x ( _k) be divided into:

Wherein, x _d(k) be that channel is at the bit sequence of data slot output; x _t(k) be that channel is at the bit sequence of training time slot output; Described decoder also includes the coefficient adjustment module, become FIR (Finite Impulse Response finite impulse response) filtration module when described and be connected, be used for adjusting the filter factor that becomes the FIR filtration module when described at the training time slot and disturb to suppress multipath.

Above-mentioned anti-multipath interference LDPC decoder, preferred described coefficient adjustment module comprises that local reference signal acquiring unit and least mean square algorithm adjustment unit, local reference signal acquiring unit are used for the parity check matrix H according to the LDPC coding, the LDPC coded samples d (k) of pre-stored described training time slot bit sequence; The least mean square algorithm adjustment unit is used for according to following rule, and the coefficient that becomes the FIR filtration module when described is adjusted: the time become the coefficient vector w (k) of FIR filtration module=[w ₀(k), w ₁(k), w ₂(k) ..., w _N(k)] T, error signal e (k), training time slot bit sequence x _t(k) and the satisfied pass of convergence factor μ (k) be w (k+1)=w (k)+2 μ (k) e (k) x _t(k); Wherein, convergence factor is $\mathrm{\μ}\left(k\right)=\frac{N+1}{2N}{x}_t^T\left(k\right)x_t\left(k\right);$ Error signal is $e\left(k\right)=d\left(k\right)-y\left(k\right)=d\left(k\right)-x_t^T\left(k\right)w\left(k\right).$

Above-mentioned anti-multipath interference LDPC decoder, preferred described local reference signal d (k) obtains in the following way: described parity check matrix H is decomposed into H=[A B], submatrix A nonsingular matrix wherein; According to G=[A ^-1B I], obtain generator matrix G by A and B, wherein, I is M * M unit matrix; Original information bits vector u is set at the tail end of coded sequence, check bit vector c _pOccupy the front end of coded sequence, according to c=[c _pU]=uG obtains the ldpc coded signal c that will transmit; According to d (k)=c _t(k), k=1 ..., N obtains local reference signal d (k), wherein, and c _t=[c _t(1) ... c _t(N)]=u _tG.

Above-mentioned anti-multipath interference LDPC decoder, the parity check matrix H of preferred described LDPC coding is obtained in the following way: the row of establishing check matrix heavily is k _Row, column weight is k _ColWith k _ColIndividual 1 is placed on the check matrix first row randomly, obtains h (1)=[h (1,1) ... h (N, 1)] ^TWith k _ColIndividual 1 is placed on the check matrix secondary series randomly, obtains h (2)=[h (1,2) ... h (N, 2)] ^TWhether check satisfies h ^T(1) h (2)＜2; If do not satisfy, again with k _ColIndividual 1 is placed on the check matrix secondary series randomly, until satisfy h ^T(1) h (2)＜2; For i=3 ..., N is with k _ColIndividual 1 is placed on check matrix i randomly lists, and obtains h (i)=[h (1, i) ... h (N, i)] ^T, whether check satisfies h ^T(i) h (j)＜2, j=1 ..., i-1 and H _iRow heavily equal k _Row, wherein

$H_i=\left[\begin{array}{cccc}h\left(\mathrm{1,1}\right)& h\left(\mathrm{1,2}\right)& \·\·\·& h(1,i)\\ h\left(\mathrm{2,1}\right)& h\left(\mathrm{2,2}\right)& \·\·\·& h(2,i)\\ \·& \·& & \·\\ \·& \·& & \·\\ \·& \·& & \·\\ h(M,1)& h(M,2)& \·\·\·& h(M,i)\end{array}\right]$

$=\left[\begin{array}{ccc}h\left(1\right)& \·\·\·& h\left(i\right)\end{array}\right],$

If do not satisfy, again with k _ColIndividual 1 is placed on check matrix i randomly lists, until satisfy h ^T(i) h (j)＜2, j=1 ..., i-1 and H _iThe heavy condition of row.When i=N, obtain

$H_N=\left[\begin{array}{cccc}h\left(\mathrm{1,1}\right)& h\left(\mathrm{1,2}\right)& \·\·\·& h(1,N)\\ h\left(\mathrm{2,1}\right)& h\left(\mathrm{2,2}\right)& \·\·\·& h(2,N)\\ \·& \·& & \·\\ \·& \·& & \·\\ \·& \·& & \·\\ h(M,1)& h(M,2)& \·\·\·& h(M,N)\end{array}\right]$

$=\left[\begin{array}{ccc}h\left(1\right)& \·\·\·& h\left(N\right)\end{array}\right]$

The output signal that above-mentioned anti-multipath interference LDPC decoder, preferred described decoder module utilize the confidence spread algorithm to become the FIR filtration module when described is decoded.

Above-mentioned anti-multipath interference LDPC decoder, preferred described decoder module comprise initialization unit, horizontal iteration unit, vertical iteration unit and decoding unit, and wherein: initialization unit is used for foundation $f_n^{}$ ${}_1^{}=1/(1+\exp (-2a{y}_n/{\mathrm{\&sigma;}}^2),$ ${\mathrm{<figure-callout\; id="0"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{}=1-{\mathrm{<figure-callout\; id="1"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{}$ Calculating parameter f _n ¹, f _n ⁰, wherein, σ ²=N ₀/ 2 is the variance of interchannel noise; A is the channel fading factor, y _nBecome the soft output of FIR filtration module when described; And, order ${\mathrm{<figure-callout\; id="1"\; label="q\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">q</figure-callout>}}_{\mathrm{mn}}^{}={\mathrm{<figure-callout\; id="1"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{},$ ${\mathrm{<figure-callout\; id="0"\; label="q\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">q</figure-callout>}}_{\mathrm{mn}}^{}={\mathrm{<figure-callout\; id="0"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{};$ Horizontal iteration unit is used for m=1 ..., M, n ∈ N (m), foundation $d{q}_{\mathrm{mn}}=q_{\mathrm{mn}}^0-{\mathrm{<figure-callout\; id="1"\; label="q\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">q</figure-callout>}}_{\mathrm{mn}}^{},$ $d{r}_{\mathrm{mn}}=r_{\mathrm{mn}}^0-{\mathrm{<figure-callout\; id="1"\; label="r\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">r</figure-callout>}}_{\mathrm{mn}}^{}=\underset{n^{\&prime;}\&Element;N\left(m\right)\backslash n}{\mathrm{\&Pi;}}{\mathrm{dq}}_{{\mathrm{mn}}^{\&prime;}},$ Calculate dq _Mn, dr _MnAnd then foundation $r_{\mathrm{mn}}^{}$ ${}_0^{}=(1+d{r}_{\mathrm{mn}})/2,$ ${\mathrm{<figure-callout\; id="1"\; label="r\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">r</figure-callout>}}_{\mathrm{mn}}^{}=(1-d{r}_{\mathrm{mn}})/2$ Obtain r _Mn ⁰, r _Mn ¹Vertical iteration unit is used for according to the r that obtains _Mn ⁰And r _Mn ¹Upgrade probable value q _Mn ⁰And q _Mn ¹Wherein, ${\mathrm{<figure-callout\; id="0"\; label="q\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">q</figure-callout>}}_{\mathrm{mn}}^{}={\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="0"\; label="mn\; f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">mn</figure-callout>}}{f}_n^{}{}_0^{}\underset{m^{\&prime;}\&Element;M\left(n\right)\backslash m}{\mathrm{\&Pi;}}{r}_{m^{\&prime;}\mathrm{<figure-callout\; id="0"\; label="n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">n</figure-callout>}}^0,$ ${\mathrm{<figure-callout\; id="1"\; label="q\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">q</figure-callout>}}_{\mathrm{mn}}^{}={\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="1"\; label="mn\; f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">mn</figure-callout>}}{f}_n^{}{}_1^{}\underset{m^{\&prime;}\&Element;M\left(n\right)\backslash m}{\mathrm{\&Pi;}}{r}_{m^{\&prime;}\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">n</figure-callout>}}^1;$ α mn is normalization coefficient, so that ${\mathrm{<figure-callout\; id="0"\; label="q\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">q</figure-callout>}}_{\mathrm{mn}}^{}+{\mathrm{<figure-callout\; id="1"\; label="q\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">q</figure-callout>}}_{\mathrm{mn}}^{}=1;$ And, foundation $q_n^o={\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="0"\; label="n\; f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">n</figure-callout>}}{f}_n^{}{}_0^{}\underset{m\&Element;M\left(n\right)}{\mathrm{\&Pi;}}{\mathrm{<figure-callout\; id="0"\; label="r\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">r</figure-callout>}}_{\mathrm{mn}}^{},$ ${\mathrm{<figure-callout\; id="1"\; label="q\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">q</figure-callout>}}_n^{}={\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="1"\; label="n\; f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">n</figure-callout>}}{f}_n^{}{}_1^{}\underset{m\&Element;M\left(n\right)}{\mathrm{\&Pi;}}{\mathrm{<figure-callout\; id="0"\; label="r\; mn"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">r</figure-callout>}}_{\mathrm{mn}}^{}$ Computation of Pseudo posterior probability q _n ⁰And q _n ¹Wherein, α _nBe normalization coefficient, so that ${\mathrm{<figure-callout\; id="0"\; label="q\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">q</figure-callout>}}_n^{}+{\mathrm{<figure-callout\; id="1"\; label="q\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">q</figure-callout>}}_n^{}=1;$ Decoding unit is used for deciphering according to presetting rule, and described presetting rule is, when ${\mathrm{<figure-callout\; id="1"\; label="q\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">q</figure-callout>}}_n^{}=1$ Season $r_n^{}$ ${}_1^{}=1,$ Otherwise ${\mathrm{<figure-callout\; id="1"\; label="r\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">r</figure-callout>}}_n^{}=0,$ M=0; If check equations rH ^T=0, then successfully decoded.

Above-mentioned anti-multipath interference LDPC decoder, the preferred FIR filtration module that becomes when becoming the FIR filtration module into variable step when described.

Than prior art, the present invention has following beneficial effect: become independently parts of two of FIR filtration module and LDPC decodings when the present invention is divided into decoder, the time become the FIR filtration module as the preprocessor of LDPC decoder, before the LDPC decoding, eliminate most of MPI; The LDPC decoder has good BER decoding performance, can be to the signal correct decoding under the MPI channel.

Description of drawings

Fig. 1 is that code length is the BER curve chart of 378 LDPC decoder under the multipath conditions;

Fig. 2 is the structural representation of baseband system model;

Fig. 3 is the burst structure schematic diagram of TDD-CDMA (Time Division-Code Division Multiple Access) system;

When being, Fig. 4 becomes the structural representation of FIR filtration module;

Fig. 5 A is the structural representation of the anti-multipath interference LDPC decoder preferred embodiment according to the present invention;

Fig. 5 B is the fundamental diagram of the anti-multipath interference LDPC decoder preferred embodiment according to the present invention;

When being, Fig. 6 becomes the output BER curve synoptic diagram of FIR filtration module;

Fig. 7 is anti-multipath interference LDPC decoder of the present invention and existing LDPC decoder BER curve performance comparison diagram.

Embodiment

For above-mentioned purpose of the present invention, feature and advantage can be become apparent more, below in conjunction with the drawings and specific embodiments the embodiment of the invention is described in further detail.

Core concept of the present invention is: become independently parts of two of FIR filtration module and LDPC decodings when decoder is divided into, the time become the FIR filtration module as the preprocessor of LDPC decoder, before the LDPC decoding, eliminate most of MPI.Below the present invention will be described in detail.

System and channel model

Consider a baseband coding modulation LDPC system, become the FIR filtration module when receiver end uses, with reference to Fig. 2, Fig. 2 is the structure chart of baseband system model.At coding side source bits sequence u (n) is carried out the LDPC coding and obtain c (n), ldpc coded signal c (n) will can describe in detail afterwards, and the code check that the codec of LDPC is set is 1/2.In decoding end, in time, become the FIR filtration module channel CIR (Channel Impulse Response-channel impulse response) estimated, and suppress MPI.

Among Fig. 2, MPC is by channel impulse response (CIR) expression, and ldpc coded signal c (n) transmits in MPC that to be subjected to channel noise jamming, interchannel noise be the white Gaussian noise of zero-mean, and noise power is N ₀, be designated as v (n).The baseband signal that obtains after the decoding end demodulation is:

x(n)＝c(n)*h(n)+v(n) (1)

Wherein, " * " represents convolution, and h (n) is the CIR of multipath channel, is real variable.

In awgn channel, the baseband signal that the reception signal obtains behind high frequency demodulation is:

x(n)＝c(n)+v(n) (2)

Although traditional LDPC decoder receives the BER performance that signal (2) has well for awgn channel, but the error code that causes for the convolution of ldpc coded signal c (n) in the formula (1) and h (n), the LDPC decoder is difficult to be correctly decoded, the BER performance of multipath fading and multipath transmisstion meeting grievous injury LDPC decoding.Therefore, the present invention proposes to become the FIR filtration module to suppress MPI when the LDPC decoder is preposition.

Consider the burst frame structure (as shown in Figure 3) of TDD-CDMA system, the TDD-CDMA base station sends training sequence in the wireless signal frame of down link.Burst sequences is made of data sequence, training sequence and protection intervening sequence GP, takies altogether the time span of 2560Tc, and Tc is the spread spectrum signal cycle.Training sequence in the burst sequences can have several different methods design, and it is the full 0 sequence, and complete 1 sequence also can design one 0, the sequence of 1 random distribution.The time become the FIR filtration module is estimated MPC with training sequence parameter.

Burst sequences frame structure divided data time slot among Fig. 3 part and training time slot part, after MPI and demodulation, the time become baseband signal that the FIR filtration module receives into:

X wherein _d(k) be that channel is at the bit sequence of data slot output; x _t(k) be that channel is at the bit sequence of training time slot output.Become the FIR filtration module during Fig. 2 preposition and at the training time slot channel parameter is estimated, at data slot MPC is filtered, for the decoder module of back provides the soft output signal y that has eliminated most of MPI (k).

For preventing the loss of the information that receives, among the present invention: in time, becomes the FIR filtration module and is soft output, and y (k) is the multi-system signal.The embodiment of the invention shows: but the time become the output establishment MPI of FIR filtration module, as SNR during greater than 3dB, the error rate can be reduced to below 20%, this is equivalent in the awgn channel situation LDPC decoder to the requirement of the input signal error rate.The time become the FIR filtration module for acquisition to the set output signal error rate, carry out the interative computation of necessary number of times.In application can according to the decoding delay of system and the error rate require to select different iterationses and the time become FIR filtration module processing signals length N.The conversion of multipath channel

Because the LDPC decoder does not possess the good decoding performance under the MPI, the present invention becomes the FIR filtration module when proposing to adopt multipath channel is transformed to awgn channel before the LDPC decoding. and in time, becomes the FIR filtration module and utilizes training sequence to obtain channel parameter, and the data sequence filtering of follow-up arrival processed, suppress MPI.Adaptive-filtering parameter adjustment algorithm is based on through the training sequence of LDPC coding, become identification and the setting of FIR filtration module parameter in the time of need to before data symbol sequence arrives, finishing, need design suitable the time become identification and the setup times that FIR filtration module length and iterations become FIR filtration module parameter when shortening, to avoid the bit drop-out of data symbol sequence.

The present invention becomes the FIR filtration module as shown in Figure 4 when using.The time become the FIR filtration module list entries into x (k)=[x (k), x (k-1) ..., x (k-N)] ^T, the time become the FIR filtration module coefficient be w _n(k), n=0 ..., N is expressed as w (k)=[w with vector form ₀(k), w ₁(k), w ₂(k) ..., w _N(k)] ^TIn Fig. 4, d (k) and y (k) be respectively training signal and the time become the output of FIR filtration module.

In time, become the output signal of FIR filtration module and input vector x (k) and coefficient following relation arranged:

y(k)＝x ^T(k)w(k) (4)

In formula (4), signal vector x (k) is bounded, and the time become the FIR filtration module output signal y (k) bounded whether, whether stable with timely changes FIR filtration module, by the convergence decision of filtration module parameter.

Decoding end local reference signal d (k) becomes the LDPC coded-bit sample of the pre-stored local training sequence of FIR filtration module when being.Among the present invention, local training sequence module has buffer, stores local training sequence, and receiver local reference signal d (k) obtains as follows.

Step 1: establishing LDPC code parity check matrix H dimension is M * N, and can be decomposed into A and B is two submatrixs

H＝[A B] (5)

Wherein submatrix A is nonsingular.

Step 2: obtain generator matrix by A and B

G＝[A ^-1B I] (6)

Wherein, I is M * M unit matrix.

Step 3: information bit vector u is placed on the tail end of coded sequence, and check bit vector c _pThen occupy the front end of coded sequence, namely through the vectorial c that obtains of coding, be the ldpc coded signal that will transmit and be

c＝[c _pu]＝uG， (7)

Step 4: be u with training sequence _tSubstitution becomes FIR filtration module local signal d (k)=c with formula (7) when obtaining _t(k), k=1 ..., N, c _t=[c _t(1) ... c _t(N)]=u _tG.D (k) is stored in the local training sequence module buffer.

At the training time slot, when adjusting according to error signal e (k), the LMS module becomes the coefficient of FIR filtration module,

$e\left(k\right)=d\left(k\right)-y\left(k\right)=d\left(k\right)-x_t^T\left(k\right)w\left(k\right)---\left(8\right)$

The output signal y (k) that becomes the FIR filtration module when making is tending towards local reference signal d (k).

LMS Least Mean Square (least mean square algorithm) becomes coefficient vector, error signal, the input signal vector x of FIR filtration module when providing _t(k) and the relation of convergence factor μ (k):

w(k+1)＝w(k)+2μ(k)e(k)x _t(k). (9)

Wherein convergence factor is:

${\mathrm{\&mu;}}_{\left(k\right)}=\frac{N+1}{2N}{x}_t^T\left(k\right)x_t\left(k\right)---\left(10\right)$

Error signal e (k) is defined by formula (8).

After becoming at that time the stable output signal of FIR filtration module, most of MPI will from the time become filtering the output y (k) of FIR filtration module.Like this, the error rate of LDPC decoder input signal is minimized, and has quite obtained an accurate awgn channel.Embodiment will verify this result.

Become the FIR filtration module during with fixed step size and compare, becoming the FIR filtration module during variable step has better performance at quick track side's mask.

Obtaining of LDPC code parity check matrix H

The below describes the acquisition methods of LDPC code parity check matrix H in detail.

The BER performance of LDPC decoder also depends on the structure of LDPC code and the algorithm of decoder.If the LDPC code that uses has 4 rings or code heavy little, the high performance decoder algorithm BER result that also can't obtain then.

The LDPC code can be divided into two large classes: random LDPC code and plan circulation (Quasi-Cyclic) LDPC code.Code without the 4 random LDPC codes that encircle is great, but Fast Convergent when high performance decoder algorithm is decoded, and decoding delay is little.The invention provides a kind of building method of LDPC code of nothing 4 rings:

Theorem 1 (LDPC code 4 ring verification theorems): the check matrix of establishing M * N is:

$H=\left[\begin{array}{cccc}h\left(\mathrm{1,1}\right)& h\left(\mathrm{1,2}\right)& \&CenterDot;\&CenterDot;\&CenterDot;& h(1,N)\\ h\left(\mathrm{2,1}\right)& h\left(\mathrm{2,2}\right)& \&CenterDot;\&CenterDot;\&CenterDot;& h(2,N)\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ h(M,1)& h(M,2)& \&CenterDot;\&CenterDot;\&CenterDot;& h(M,N)\end{array}\right]---\left(11\right)$

$=\left[\begin{array}{ccc}h\left(1\right)& \&CenterDot;\&CenterDot;\&CenterDot;& h\left(N\right)\end{array}\right].$

And if only if appoints and to get two column vectors

h(i)＝[h(1，i)...h(N，i)] ^T，i∈{1，...，M}

With

h(j)＝[h(1，j)...h(N，j)] ^T，j∈{1，...，M}，

And j ≠ i then has h ^T(i) h (j)＜2, check matrix (13) is without 4 rings.

Utilize theorem 1, can obtain the random LDPC code constructing method without 4 rings.The row of considering check matrix heavily is k _Row, column weight is k _Col, design procedure is as follows:

Step 1: the row of establishing check matrix heavily is k _Row, column weight is k _Col

Step 2: with k _ColIndividual 1 is placed on the check matrix first row randomly, obtains h (1)=[h (1,1) ... h (N, 1)] ^T

Step 3: with k _ColIndividual 1 is placed on the check matrix secondary series randomly, obtains h (2)=[h (1,2) ... h (N, 2)] ^T, whether check satisfies h ^T(1) h (2)＜2; If do not satisfy, again with k _ColIndividual 1 is placed on the check matrix secondary series randomly, until satisfy h ^T(1) h (2)＜2;

Step 4: for i=3 ..., N is with k _ColIndividual 1 is placed on check matrix i randomly lists, and obtains h (i)=[h (1, i) ... h (N, i)] ^T, whether check satisfies h ^T(i) h (j)＜2, j=1 ..., i-1 and H _iRow heavily equal k _Row, wherein

$H_i=\left[\begin{array}{cccc}h\left(\mathrm{1,1}\right)& h\left(\mathrm{1,2}\right)& \&CenterDot;\&CenterDot;\&CenterDot;& h(1,i)\\ h\left(\mathrm{2,1}\right)& h\left(\mathrm{2,2}\right)& \&CenterDot;\&CenterDot;\&CenterDot;& h(2,i)\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ h(M,1)& h(M,2)& \&CenterDot;\&CenterDot;\&CenterDot;& h(M,i)\end{array}\right],$

$=\left[\begin{array}{ccc}h\left(1\right)& \&CenterDot;\&CenterDot;\&CenterDot;& h\left(i\right)\end{array}\right]$

If do not satisfy, again with k _ColIndividual 1 is placed on check matrix i randomly lists, until satisfy h ^T(i) h (j)＜2, j=1 ..., the heavy condition of the row of i-1 and Hi.When i=N, obtain

$H_N=\left[\begin{array}{cccc}h\left(\mathrm{1,1}\right)& h\left(\mathrm{1,2}\right)& \&CenterDot;\&CenterDot;\&CenterDot;& h(1,N)\\ h\left(\mathrm{2,1}\right)& h\left(\mathrm{2,2}\right)& \&CenterDot;\&CenterDot;\&CenterDot;& h(2,N)\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ h(M,1)& h(M,2)& \&CenterDot;\&CenterDot;\&CenterDot;& h(M,N)\end{array}\right]$

$=\left[\begin{array}{ccc}h\left(1\right)& \&CenterDot;\&CenterDot;\&CenterDot;& h\left(N\right)\end{array}\right]---\left(13\right)$

H _NMatrix is the parity check matrix H of the anti-multipath interference LDPC decoder use of the present invention's proposition.

The decoding of confidence spread algorithm

Among the present invention, decoder module adopts BP (Belief Propagation-confidence spread) algorithm, also claims sum-product algorithm.Definition

N(m)≡{n：H ^T＝1) (14a)

Be the set of the information node adjacent with check-node m, namely check matrix m capable in the row number set of nonzero element, among the set N (m) k is arranged _RowIndividual element.

Similarly, definition

M(n)≡{m：H ^T＝1) (14b)

Be the check-node set adjacent with information node n, the namely line number set of nonzero element in the check matrix n row has k among the set M (n) _ColIndividual element.

The BP algorithm specific implementation process that the decoding of LDPC code is adopted is as follows:

Step 1 initialization

Make y _nBecome the output of FIR filter when being, then

${\mathrm{<figure-callout\; id="1"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{}=1/(1+\exp (-2a{y}_n/{\mathrm{\&sigma;}}^2));---\left(15a\right)$

${\mathrm{<figure-callout\; id="0"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{}=1-{\mathrm{<figure-callout\; id="1"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{},---\left(15b\right)$

Wherein, σ ²=N ₀/ 2 is the variance of interchannel noise; A is the channel fading factor, and

$q_{\mathrm{mn}}^1={\mathrm{<figure-callout\; id="1"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{};---\left(16a\right)$

$q_{\mathrm{mn}}^0={\mathrm{<figure-callout\; id="0"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{}.---\left(16b\right)$

The horizontal iteration of step 2: to m=1 ..., M, n ∈ N (m) calculates

${\mathrm{dq}}_{\mathrm{mn}}=q_{\mathrm{mn}}^0-q_{\mathrm{mn}}^1$ With ${\mathrm{dr}}_{\mathrm{mn}}=r_{\mathrm{mn}}^0-r_{\mathrm{mn}}^1=\underset{n^{\&prime;}\&Element;N\left(m\right)\backslash n}{\mathrm{\&Pi;}}{\mathrm{dq}}_{{\mathrm{mn}}^{\&prime;}},---\left(17\right)$

Obtain

$r_{\mathrm{mn}}^0=(1+{\mathrm{dr}}_{\mathrm{mn}})/2;---\left(18a\right)$

$r_{\mathrm{mn}}^1=(1-{\mathrm{dr}}_{\mathrm{mn}})/2.---\left(18b\right)$

The vertical iteration of step 3: utilize the r that obtains in the step 2 _Mn ⁰And r _Mn ¹Upgrade probable value q _Mn ⁰And q _Mn ¹

$q_{\mathrm{mn}}^0={\mathrm{\&alpha;}}_{\mathrm{mn}}{\mathrm{<figure-callout\; id="0"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{}\underset{m^{\&prime;}\&Element;M\left(n\right)\backslash m}{\mathrm{\&Pi;}}{r}_{m^{\&prime;}\mathrm{<figure-callout\; id="0"\; label="n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">n</figure-callout>}}^0;---\left(19a\right)$

$q_{\mathrm{mn}}^1={\mathrm{\&alpha;}}_{\mathrm{mn}}{\mathrm{<figure-callout\; id="1"\; label="f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">f</figure-callout>}}_n^{}\underset{m^{\&prime;}\&Element;M\left(n\right)\backslash m}{\mathrm{\&Pi;}}{r}_{m^{\&prime;}\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">n</figure-callout>}}^1,---\left(19b\right)$

Wherein, α _MnBe normalization coefficient, so that $q_{\mathrm{mn}}^0+q_{\mathrm{mn}}^1=1.$

While Computation of Pseudo posterior probability

${\mathrm{<figure-callout\; id="0"\; label="q\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">q</figure-callout>}}_n^{}={\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="0"\; label="n\; f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">n</figure-callout>}}{f}_n^{}{}_0^{}\underset{m\&Element;M\left(n\right)}{\mathrm{\&Pi;}}{r}_{\mathrm{mn}}^0;---\left(20a\right)$

${\mathrm{<figure-callout\; id="1"\; label="q\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">q</figure-callout>}}_n^{}={\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="1"\; label="n\; f\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">n</figure-callout>}}{f}_n^{}{}_1^{}\underset{m\&Element;M\left(n\right)}{\mathrm{\&Pi;}}{r}_{\mathrm{mn}}^0,---\left(20b\right)$

Wherein, α _nBe normalization coefficient, so that ${\mathrm{<figure-callout\; id="0"\; label="q\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00032.png"\; state="\{\{state\}\}">q</figure-callout>}}_n^{}+{\mathrm{<figure-callout\; id="1"\; label="q\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">q</figure-callout>}}_n^{}=1.$

Step 4 trial and error decoding

When ${\mathrm{<figure-callout\; id="1"\; label="q\; n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">q</figure-callout>}}_n^{}=1$ Season $r_n^{}$ ${}_1^{}=1,$ Otherwise $r_{\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="H2009102378610E00011.png,H2009102378610E00021.png"\; state="\{\{state\}\}">n</figure-callout>}}^1=0,$ If check equations rH m=0. ^T=0, then successfully decoded and finish, on the contrary then get back to step 2. in a single day successfully decoded (satisfy end condition or reach maximum cycle), obtain the information bit vector u after the error correction.The anti-multipath interference LDPC decoder preferred embodiment

With reference to Fig. 5 A, Fig. 5 B, Fig. 5 A is the structural representation of the anti-multipath interference LDPC decoder preferred embodiment according to the present invention, and Fig. 5 B is the fundamental diagram of the anti-multipath interference LDPC decoder preferred embodiment according to the present invention.Comprise: anti-multipath interference LDPC decoder coefficient adjustment module 52, the time become FIR filtration module 54 and decoder module 56.

Wherein, coefficient adjustment module 52 comprises local reference signal acquiring unit 521 and least mean square algorithm adjustment unit 522:

Local reference signal obtains single 521 parity check matrix H that are used for according to the LDPC coding, the LDPC coded samples d (k) of pre-stored described training time slot bit sequence.Least mean square algorithm adjustment unit 522 is used for according to following rule, and the coefficient that becomes the FIR filtration module when described is adjusted: the time become the coefficient vector w (k) of FIR filtration module=[w ₀(k), w ₁(k), w ₂(k) ..., w _N(k)] ^T, error signal e (k), training time slot bit sequence x _t(k) and the satisfied pass of convergence factor μ (k) be w (k+1)=w (k)+2 μ (k) e (k) x _t(k); Wherein, convergence factor is $\mathrm{\&mu;}\left(k\right)=\frac{N+1}{2N}{x}_t^T\left(k\right)x_t\left(k\right);$ Error signal is $e\left(k\right)=d\left(k\right)-y\left(k\right)=d\left(k\right)-x_t^T\left(k\right)w\left(k\right)$ The time become the FIR filtration module.

Decoder module 56 comprises initialization unit 561, horizontal iteration unit 562, vertical iteration unit 563 and decoding unit 564:

Initialization unit 561 is used for foundation $f_n^1=1/(1+\exp (-2{\mathrm{ay}}_n/{\mathrm{\&sigma;}}^2),$ $f_n^0=1-f_n^1$ Calculating parameter f _n ¹, f _n ⁰, wherein, σ ²=N ₀/ 2 is the variance of interchannel noise; A is the channel fading factor, y _nFor the time become the soft output of FIR filtration module; And, order $q_{\mathrm{mn}}^1=f_n^1,$ $q_{\mathrm{mn}}^0=f_n^0;$ Horizontal iteration unit 562 is used for m=1 ..., M, n ∈ N (m), foundation ${\mathrm{dq}}_{\mathrm{mn}}=q_{\mathrm{mn}}^0-q_{\mathrm{mn}}^1,$ ${\mathrm{dr}}_{\mathrm{mn}}=r_{\mathrm{mn}}^0-r_{\mathrm{mn}}^1=\underset{n^{\&prime;}\&Element;N\left(m\right)\backslash n}{\mathrm{\&Pi;}}{\mathrm{dq}}_{{\mathrm{mn}}^{\&prime;}},$ Calculate dq _Mn, dr _MnAnd then foundation $r_{\mathrm{mn}}^0=(1+{\mathrm{dr}}_{\mathrm{mn}})/2,$ $r_{\mathrm{mn}}^1=(1-{\mathrm{dr}}_{\mathrm{mn}})/2$ Obtain rmn0, r _Mn ¹Vertical iteration unit 563 is used for according to the r that obtains _Mn ⁰And r _Mn ¹Upgrade probable value q _Mn ⁰And q _Mn ¹Wherein, $q_{\mathrm{mn}}^0={\mathrm{\&alpha;}}_{\mathrm{mn}}{f}_n^0\underset{m^{\&prime;}\&Element;M\left(n\right)\backslash m}{\mathrm{\&Pi;}}{r}_{m^{\&prime;}n}^0,$ $q_{\mathrm{mn}}^1={\mathrm{\&alpha;}}_{\mathrm{mn}}{f}_n^1\underset{m^{\&prime;}\&Element;M\left(n\right)\backslash m}{\mathrm{\&Pi;}}{r}_{m^{\&prime;}n}^1;$ α _MnBe normalization coefficient, so that $q_{\mathrm{mn}}^0+q_{\mathrm{mn}}^1=1;$ And, foundation $q_n^0={\mathrm{\&alpha;}}_n{f}_n^0\underset{m\&Element;M\left(n\right)}{\mathrm{\&Pi;}}{r}_{\mathrm{mn}}^0,$ $q_n^1={\mathrm{\&alpha;}}_n{f}_n^1\underset{m\&Element;M\left(n\right)}{\mathrm{\&Pi;}}{r}_{\mathrm{mn}}^0$ Computation of Pseudo posterior probability q _n ⁰And q _n ¹Wherein, α _nBe normalization coefficient, so that $q_n^0+q_n^1=1;$ Decoding unit 564 is used for deciphering according to presetting rule, and described presetting rule is, when $q_n^1=1$ Season $r_n^1=1,$ Otherwise $r_n^1=0,$ M=0; If check equations rH ^T=0, then successfully decoded.

As can be seen from Figure 5B, decoder module not the time become in the closed loop of FIR filtration module, do not participate in interative computation at the training time slot, the coefficient adjustment module is independently carried out the accent coefficient adjustment of FIR time varing filter, and the LDPC decoder is only deciphered at data slot (time become the convergence of FIR filtration module after).Because the LDPC decoder is not participated in the interative computation of channel parameter identification, and can the interative computation amount directly have influence on the coefficient adjustment that becomes the FIR filtration module in the training sequence time slot in the time of finishing.LDPC decoder of the present invention is only deciphered at data slot (time become the convergence of FIR filtration module after), has that amount of calculation is little, hardware is realized simple advantage.

The feasibility of the anti-multipath interference LDPC decoder that proposes for checking the present invention has been carried out system emulation, and in this emulation experiment, the LDPC decoder of change FIR filter does not compare to anti-multipath interference LDPC decoder and when using.

Making multipath channel impact the length that swashs response is 8, and the multipath channel parameter is

h(n)＝e ^-0.5n，n＝0，1，...7。

The time become the length of FIR filter and the length of training sequence is 6N into the N=186.LDPC code length.

Adopt the random LDPC code constructing method of the present invention to obtain a random LDPC code.The iterations of LDPC decoder is 20 times.It is 1000 that the LDPC encoder sends the data frame number.

The time become the FIR filtration module output BER curve as shown in Figure 6, the time has become FIR filter establishment MPI, the BER of its output signal reduce and connect the LDPC decoder after making and decipher smoothly.For the LDPC decoder, the Signal Pretreatment equivalence of preposition adaptive equalizer is with the MPC conversion awgn channel that is as the criterion.

From Fig. 7, can find out the effect of filter.Wherein, curve a representative becomes the BER curve of the LDPC decoder of FIR filter when not having, and becomes the BER curve of the LDPC decoder of FIR filter during curve b; When SNR is 0～6dB, the BER that becomes the LDPC decoder of FIR filter when not adopting maintains about 0.25, without BER waterfall characteristic, and anti-multipath interference LDPC decoder is that BER waterfall characteristic appears in 2～4.5dB at SNR, when the BER of SNR during greater than 4.5dB is lower than 10-6, this and the LDPC decoder BER performance in awgn channel is close.

Above a kind of anti-multipath interference LDPC decoder provided by the present invention is described in detail, used specific case herein principle of the present invention and execution mode are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, all will change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.

Claims (5)

1. an anti-multipath interference LDPC decoder comprises decoder module and filtration module, it is characterized in that, and described decoder module and filtration module cascade, described filtration module suppresses for multipath is disturbed; Become the FIR filtration module when this filtration module is, the channel signal that this module receives is that x (k) is divided into:

Wherein, x _d(k) be that channel is at the bit sequence of data slot output; x _t(k) be that channel is at the bit sequence of training time slot output;

Described decoder also includes the coefficient adjustment module, becomes the FIR filtration module when described and is connected, and is used for adjusting the filter factor that becomes the FIR filtration module when described at the training time slot and disturbs to suppress multipath;

Described coefficient adjustment module comprises:

The local reference signal acquiring unit is used for the parity check matrix H according to the LDPC coding, the LDPC coded samples of pre-stored described training time slot bit sequence;

The least mean square algorithm adjustment unit is used for according to following rule, and the coefficient that becomes the FIR filtration module when described is adjusted: the time become the coefficient vector w (k) of FIR filtration module=[w ₀(k), w ₁(k), w ₂(k) ..., w _N(k)] ^T, error signal e (k), training time slot bit sequence x _t(k) and the satisfied pass of convergence factor μ (k) be w (k+1)=w (k)+2 μ (k) e (k) x _t(k); Wherein, convergence factor is

Error signal is

D (k) is local reference signal, and local reference signal d (k) is the LDPC coded samples of the training sequential bit sequence of preferential storage;

Wherein, y (k) for the time become the output signal of FIR filtration module, x _t ^T(k) be x _t(k) transposition;

Described decoder module comprises:

Initialization unit is used for foundation $f_n^1=1/(1+\exp (-2{\mathrm{ay}}_n/{\mathrm{\&sigma;}}^2),$ $f_n^0=1-f_n^1$ Calculating parameter

Wherein, σ ²=N ₀/ 2 is the variance of interchannel noise; A is the channel fading factor, y _nBecome the soft output of FIR filtration module when described; And, order

$q_{\mathrm{mm}}^0=f_n^0;$

Horizontal iteration unit is used for m=1 ...., M, n ∈ N (m), foundation ${\mathrm{dq}}_{\mathrm{mn}}=q_{\mathrm{mn}}^0-q_{\mathrm{mn}}^1,$ ${\mathrm{dr}}_{\mathrm{mn}}=r_{\mathrm{mn}}^0-r_{\mathrm{mn}}^1=\underset{n^{\&prime;}\&Element;N\left(m\right)\backslash n}{\mathrm{\&Pi;}}{\mathrm{dq}}_{{\mathrm{mn}}^{\&prime;}},$ Calculate dq _Mn, dr _MnAnd then foundation $r_{\mathrm{mn}}^0=(1+{\mathrm{dr}}_{\mathrm{mn}})/2,$ $r_{\mathrm{mn}}^1=(1-{\mathrm{dr}}_{\mathrm{mn}})/2$ Obtain

Vertical iteration unit is obtained for foundation

With

Upgrade probable value

With

Wherein, $q_{\mathrm{mn}}^0={\mathrm{\&alpha;}}_{\mathrm{mn}}{f}_n^0\underset{m^{\&prime;}\&Element;M\left(n\right)\backslash m}{\mathrm{\&Pi;}}{r}_{m^{\&prime;}n}^0,$ $q_{\mathrm{mn}}^1={\mathrm{\&alpha;}}_{\mathrm{mn}}{f}_n^1\underset{m^{\&prime;}\&Element;M\left(n\right)\backslash m}{\mathrm{\&Pi;}}{r}_{m^{\&prime;}n}^1;$ α _MnBe normalization coefficient, so that $q_{\mathrm{mn}}^0+q_{\mathrm{mn}}^1=1;$

And, foundation $q_n^0={\mathrm{\&alpha;}}_n{f}_n^0\underset{m\&Element;M\left(n\right)}{\mathrm{\&Pi;}}{r}_{\mathrm{mn}}^0,$ $q_n^1={\mathrm{\&alpha;}}_n{f}_n^1\underset{m\&Element;M\left(n\right)}{\mathrm{\&Pi;}}{r}_{\mathrm{mn}}^0$ The Computation of Pseudo posterior probability

With

Wherein, α _nBe normalization coefficient, so that

Wherein, N (m) is number set of the capable row of check matrix m, and the total number of this set is N, and M (n) is the line number set of check matrix n row, and the total number of this set is M;

Decoding unit is used for deciphering according to presetting rule, and described presetting rule is, when

Season

Otherwise

M=0; If check equations rH ^T=0, then successfully decoded; Wherein r is LDPC decoding output bit sequence, uses vector representation.

2. anti-multipath interference LDPC decoder according to claim 1 is characterized in that, described local reference signal d (k) obtains in the following way:

Described parity check matrix H is decomposed into H=[A B], submatrix A nonsingular matrix wherein; According to G=[A ^-1B I], obtain generator matrix G by A and B, wherein, I is M * M unit matrix;

Original information bits vector u is set at the tail end of coded sequence, check bit vector c _pOccupy the front end of coded sequence, according to c=[c _pU]=uG obtains the ldpc coded signal c that will transmit;

According to d (k)=c _t(k), k=1 ..., N obtains local reference signal d (k), wherein, and c _t=[c _t(1) ... c _t(N)]=u _tG.

3. anti-multipath interference LDPC decoder according to claim 2 is characterized in that, the parity check matrix H of described LDPC coding is obtained in the following way:

If the row of check matrix heavily is k _Row, column weight is k _ColWith k _ColIndividual 1 is placed on the check matrix first row randomly, obtains h (1)=[h (1,1) ... h (N, 1)] ^TWith k _ColIndividual 1 is placed on the check matrix secondary series randomly, obtains h (2)=[(1,2) ... h (N, 2)] ^T

Whether check satisfies h ^T(1) h (2)＜2; If do not satisfy, again with k _ColIndividual 1 is placed on the check matrix secondary series randomly, until satisfy h ^T(1) h (2)＜2;

For i=3 ..., N is with k _ColIndividual 1 is placed on check matrix i randomly lists, and obtains h (i)=[h (1, i) ... h (N, i)] ^T, whether check satisfies h ^T(i) h (j)＜2, j=1 .., i-1 and H _iRow heavily equal k _Row, wherein

$H_i=\left[\begin{array}{cccc}h\left(\mathrm{1,1}\right)& h\left(\mathrm{1,2}\right)& ...& h(1,i)\\ h\left(\mathrm{2,1}\right)& h\left(\mathrm{2,2}\right)& ...& h(2,i)\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ h(M,1)& h(M,2)& ...& h(M,i)\end{array}\right]$

$=\left[\begin{array}{ccc}h\left(1\right)& ...& h\left(i\right)\end{array}\right],$

If do not satisfy, again with k _ColIndividual 1 is placed on check matrix i randomly lists, until satisfy h ^T(i) h (j)＜2, j=1 ..., the heavy condition of the row of i-1 and H i; When i=N, obtain

$H_N=\left[\begin{array}{cccc}h\left(\mathrm{1,1}\right)& h\left(\mathrm{1,2}\right)& ...& h(1,N)\\ h\left(\mathrm{2,1}\right)& h\left(\mathrm{2,2}\right)& ...& h(2,N)\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ h(M,1)& h(M,2)& ...& h(M,N)\end{array}\right]$

$=\left[\begin{array}{ccc}h\left(1\right)& ...& h\left(N\right)\end{array}\right].$

4. anti-multipath interference LDPC decoder according to claim 3 is characterized in that, the output signal that described decoder module utilizes the confidence spread algorithm to become the FIR filtration module when described is decoded.

5. anti-multipath interference LDPC decoder according to claim 4 is characterized in that, becomes the FIR filtration module when becoming the FIR filtration module into variable step when described.

Images