CN101263658A - Method and apparatus for decoding linear packet code - Google Patents

Abstract

According to a method and apparatus taught herein, a decoding circuit and method decode linear block codes based on determining joint probabilities for one or more related subsets of bits in received data blocks. The use of joint probabilities enables faster and more 5 reliable determination of received bits, meaning that, for example, joint probability decoding requires fewer decoding iterations than a comparable decoding process based on single-bit probabilities. As a non-limiting example, the decoding circuit and method taught herein provide advantageous operation with Low Density Parity Check (LDPC) codes, and can be incorporated in a variety of communication systems and devices, such as those associated with wireless communication networks.

Description

The method and apparatus of decoding linear packet code

Background technology

The present invention relates generally to communication, and particularly relate to coming coded data to decode, all low-density checksums in this way of these linear block codess (LDPC) sign indicating number by linear block codes (linear blockcode).

In principle, the noisy communication channel communication channel is supported with the transfer of data until the arbitrarily low error rate of the capacity limit of this channel.Providing strong uses this channel capacity to propose great challenge with forward error correction coding practicality, while effectively.The simple relatively decode operation of various known coding techniques supports provides approaching as the ability by the defined theoretic channel capacity limit of signal to noise ratio (snr) simultaneously.Particularly, low-density checksum (LDPC) sign indicating number provides the excellent combination of implementing practicality and good channel capacity utilization.

In general, the LDPC representation realizes one type linear block codes of practical decoding embodiment, allows channel capacity to utilize near theoretic capacity limit simultaneously.Can find more information about LDPC sign indicating number and tradition decoding at R.G..Gallager in " the Low-densityparity-check codes " that deliver on the 8th volume 21-28 page or leaf at IRE Trans.Info.Theory in January, 1962, the document is examined by being incorporated by reference in this text at this.Other codes such as Turbo code provide confers similar advantages.

Because can come linear block codes is decoded with decode operation simple relatively, iteration, so linear block codes provides practical coding/decoding method.For example upgrade the probability of each bit in the data block that receives iteratively, traditional packet decoding method depends on probabilistic information by turn.

Summary of the invention

As in one or more embodiment that this instructs, recover to utilize linear block codes to come coded data from the data block that receives on the basis of the joint probability of decoding circuit and the method one or more related bits subclass in the data block that calculating receives.For example, in one embodiment, a kind of method that the data block that receives is decoded comprises the joint probability of the bit subset in this data block that receives of initialization, upgrade joint probability based on parity matrix, and utilize the joint probability after upgrading to recover coded data from this data block that receives, wherein the data block that this receives is encoded by the linear block codes of representing by parity matrix.Bit relationships in the data block that the parity-check matrix sign receives, and represent one group of parity check equations, this group parity check equations can be evaluated aspect joint probability.

In one or more embodiments, based on initialized value, joint probability calculation takes place once.In other embodiments, initialized value is used to the joint probability that the iteration acquisition is recomputated in the first time.Then, with the initial value of those results, perhaps as the basis of calculating final joint probability as one or more iteration subsequently.Can control the iterations of whether carrying out iteration and/or will carrying out according to iteration metric.In one embodiment, iteration metric depends on signal quality or the intensity that receives.

Though be not to be limited to this, be applicable to low-density checksum (LDPC) sign indicating number and other linear block codes at this Method and circuits of instructing.For LDPC and other linear block codess, (parity check) sign indicating number and outer block code were encoded in the data block that receives can be considered to utilize, and for example, given LDPC can be considered to represent two nested codes or cascaded code.In such embodiments, decoding circuit can be disposed utilize decreased number calculate initial joint probabilities by the defined bit relationships of given LDPC sign indicating number, and use remaining (or all) bit relationships to upgrade those initial joint probabilities then.Certainly, the decoder circuit configuration among such embodiment may be used on wherein given linear block codes " being looked " to comprise the situation of nested code, and may be used on wherein in fact using the situation of independent sign indicating number.

Certainly, the present invention is not limited to above feature and advantage.In fact, those skilled in the art are reading the following detailed description and will appreciate that additional feature and advantage when checking accompanying drawing.

Description of drawings

Fig. 1 is the block diagram that is comprised in an embodiment of the decoding circuit among the embodiment of wireless communication receiver.

Fig. 2 utilizes joint probability linear block-coded data to be carried out the logical flow chart of an embodiment of process of decoding logic.

Fig. 3 utilizes joint probability linear block-coded data to be carried out the logical flow chart of another embodiment of process of decoding logic.

Embodiment

Fig. 1 illustrates an embodiment who is configured to utilize the decoding circuit 10 of joint probability estimation in its packet decoding operation.Be compared to and in packet decoding, use single-bit probabilities, use joint probability especially to provide more reliable decoding result, thereby improved performance, reduced power loss etc. with less iteration.In this graphic extension, decoding circuit 10 is comprised in the wireless communication receiver 12, and this wireless communication receiver 12 can be understood that ordinary representation cordless communication network or intrasystem fixing or mobile receiver.Similarly, receiver 12 can comprise radio base station or mobile radio station, such as celluler radio phone, portable digital-assistant, beep-pager or other radio communication devices.

Receiver 12 comprises the antenna 14 that is used for receiving inputted signal, and described input signal comprises the signal of communication that has carried the data that block encoding is crossed (for example utilizing the LDPC sign indicating number to carry out the coded data piece).Receiver 12 further comprises front-end circuit 16, be used to filter the signal that receives with digitlization, also comprise demodulator circuit 18, this demodulator circuit 18 is configured to provide the soft value of estimating corresponding to each or associating bit for decoding circuit 10 in one or more embodiments.An embodiment of the decoding circuit 10 that can implement with hardware, software or its any combination comprises initializing circuit 20, counting circuit 22, evaluation circuits 24 and optional control circuit 26, to support its processing to the data block that receives.Decoding circuit 10 is handled from the data block data recovered that receives by its joint probability and is passed to one or more additional treatment circuits 30.

Fig. 2 illustrates an embodiment of the processing logic of implementing by one or more treatment circuits, and described treatment circuit comprises the decoding circuit 10 that is used for realizing joint probability.Shown processing starts from the joint probability (step 100) of the bit subset in the data block that initialization receives, and proceeds to based on the parity matrix that is associated with data that receive that block encoding is crossed and upgrade joint probability (step 102).As will this describe in detail, upgrading processing and can comprise single step (one-shot) operation after a while, wherein can calculate final joint probability by initial joint probability estimation, perhaps this renewal is handled and can be comprised iterative process.No matter any situation takes place, and all is used to recover coded data (step 104) from this data block that receives from the joint probability of such renewal that processing obtained.Should be appreciated that and to carry out the processing action shown in these on the afoot basis, as the part of the signal processing that receives.

In one or more embodiments, the joint probability of the bit subset in the data block that initialization receives comprises utilizes parity matrix bit subset to be designated the coupling bit in the data block that receives and the probable value of the possible bit value combinations in each bit subset of initialization.In other words, parity matrix identifies the related bits in this data block that receives.For fixed code, can be scheduled to known relationship, and use by decoding circuit 10 certainly.Therefore yet decoding circuit 10 can be configured to assess any given parity matrix, with the sign related bits and obtain to distinguish and adapt to the flexibility of different code.In at least one embodiment, the linear block codes that is associated with the data block that receives comprises the LDPC sign indicating number, and parity matrix comprises one group of parity check equations, and wherein each parity check equations identifies the coupling bit in this data block that receives.

Yet, the sign related bits is in order that determine bit subset, estimate the joint probability of this bit subset, in one or more embodiments, the process of upgrading joint probability comprises basis as upgrades joint probability by all correlation combiner of the represented joint probability of parity matrix.

The basis of the wide in range coding/decoding method that is used to understand above explanation is provided at the more detailed mathematic(al) treatment of example embodiment.Therefore, by limiting examples, following matrix H is represented example LDPC sign indicating number parity matrix, wherein every line display parity check equations (PCE), and the bit position in the data block that receives is shown in every tabulation:

$H=\left[\begin{array}{c}100010001000\\ 010001000100\\ 001000100010\\ 000100010001\\ 100001000010\\ 010010000001\\ 001000011000\\ 000100100100\end{array}\right]$ Equation (1)

Equation (2) expression is corresponding to the generator matrix G of parity check matrix H.

$\left[\begin{array}{c}100100111010\\ 010100000101\\ 001100110000\\ 000010011001\\ 000001100110\end{array}\right]$ Equation formula (2)

From above matrix as seen, this example block code length 12 bits, and indicated related bits by 8 PCE that rectangular array is represented.For example, PCE 1 has indicated bit position 1,5 relevant with 9.More particularly, according to even parity check, bit 1,5 in any data block that receives of PCE 1 expression and 9 value must amount to zero.Therefore, correct bit that receives 1,5 and 9 value comprise 0,0,0}, 0,1,1}, 1,1,0} and 1,0,1}.Similarly, PCE 2 has indicated bit 2,6 and 10 to amount to zero, and PCE 3 has indicated bit 3,7 and 11 to amount to zero, or the like.

Notably, can also see from H that PCE 1 and PCE 6 comprise bit 5.The ability that bit between the PCE and within the PCE is associated provides the basis of the joint probability decoding of decoding circuit 10.Consider this point, consider the embodiment of joint probability decoding, it is right that this embodiment forms bit with relevant bit subset.(one of skill in the art will appreciate that and to use other subclass) such as using tlv triple etc.PCE 1 has indicated bit 1 and 5 to be associated, and first pair of related bits can be formed (1,5), and this is represented as d at this _(1,5)Can form other relevant bit groupings, i.e. (2,10), (3,9), (4,7), (6,11) and (8,12) similarly.

For each relevant bit (that is, possible bit value combinations) all had four kinds of probability, these four kinds of probability are corresponding to pattern (pattern) (0,0), (0,1), (1,0) and (1,1).Therefore, the probabilistic decoding according to the joint probability decode operation of decoding circuit 10 requires to keep a plurality of probable values for each bit subset of being correlated with.(shown in Fig. 1) memory 28 can be contained in the decoding circuit 10 or with decoding circuit 10 and is associated, is used to the different bit combinations that each subclass is associated and keeps a plurality of probable values.

Utilize d _(1,5)As an example, the eigenvalue of bit 1 and four correspondences of 5 is:

$P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(00\right)=(1-{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">P</figure-callout>}}_1^{\mathrm{int}})(1-P_5^{\mathrm{int}})$ Equation (3)

$P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right)=(1-{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">P</figure-callout>}}_1^{\mathrm{int}})P_5^{\mathrm{int}}$ Equation (4)

$P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(10\right)={\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">P</figure-callout>}}_1^{\mathrm{int}}(1-P_5^{\mathrm{int}})$ Equation (5)

$P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(11\right)={\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">P</figure-callout>}}_1^{\mathrm{int}}{P}_5^{\mathrm{int}}$ Equation (6)

Wherein, term " intrinsic " expression beginning or initial joint probability value.Single-bit probabilities is (such as P ₁ ^Int) represent that bit adopts the probability of the value " 1 " opposite with " 0 ".As mentioning in the discussion of Fig. 2, initializing circuit 20 can be configured to be provided with intrinsic joint probability value on the basis of the bit soft value of the correspondence that demodulator circuit 18 is exported.Expected demodulator embodiment in this discussion after a while, these demodulators embodiment provide joint demodulation information, rather than the soft value of single-bit.(example for demodulation process is served as in the 8PSK demodulation, and this demodulation process produces joint demodulation information.Utilize 8PSK, demodulating process is determined 3 bits for each symbol, and can utilize the relevant bit subset of these intrinsics to determine joint probability).

Expectation turns back to equation (3)-equation (6), can see, if then only need be calculated three kinds in the possibility pattern, because these four kinds of patterns must amount to 1 aspect probability.In any case, decoding circuit 10 at each relevant bit to (d _(1,5), d _(2,10), d _(3,9), d _(4,7), d _(6,11)And d _(8,12)) realize the processing of equation (3)-equation (6).Therefore, the joint probability of the interested related bits subclass of initialization, and also joint probability is handled the joint probability that continuation calculating was upgraded.

For the iteration first time, and utilize d _(1,5)As an example, form three groups of joint probabilities at bit 1 and 5, these probability are corresponding to getting rid of different PCE among three PCE, and these three PCE comprise one or two in bit 1 and 5.(one or two in the bit 1 and 5 appears in PCE 1,5 and 6.) for " 01 " pattern, upgrade joint probability when the i time iteration according to following equation:

$P_{1,\left(\mathrm{1,5}\right)}(01,i)=P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right)\mathrm{Pr}\left\{\mathrm{PCE}5\mathrm{met}\right|\left(\mathrm{1,5}\right)=01\left\}\mathrm{Pr}\right\{\mathrm{PCE}6\mathrm{met}|\left(\mathrm{1,5}\right)=01\}$ Equation (7)

$=P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right)(P_{5,\left(\mathrm{6,11}\right)}(00.i-1)+P_{5,\left(\mathrm{6,11}\right)}(11.i-1))\&CenterDot;\left(\frac{1-(1-2P_{\mathrm{6,2}}(i-1))(1-2P_{\mathrm{6,12}}(i-1))}{2}\right)$ Equation (8)

Wherein the subscript of P " 1, (1,5) " expression is from getting rid of PCE 1 joint probability calculation of bit 1 and 5.This eliminating has been avoided otherwise the probability that can occur departs from.Be appreciated that this departing from the sense, promptly extra relevant bit 5 and 9 information help to determine the bit 1 in the environment of PCE 1, but this information should be from PCE 1 itself.On the contrary, this additional knowledge should from comprise bit 5 or 9 other (remaining) PCE and come autocorrelative intrinsic joint probability (perhaps coming the joint probability of the renewal of comfortable preceding iteration).

Because to d _(6,11)In two bits all be comprised among the PCE 5, so can by and P _{5, (6,11)}(00, i-1)+P _{5 (6,11)}(11, i-1) calculate even number bit among the PCE 5 probability of (getting rid of bit 1).00,10 and 11 bit modes are similarly calculated renewal.Can also observedly be that single-bit probabilities occurs, such as P _6,2(i-1).Decoding circuit 10 can be configured to for example obtain single-bit probabilities by two joint probabilities are sued for peace as needs.For example, can calculate the single-bit probabilities of second bit of PCE 6 as followsly:

P _6,2=P _{6, (2,10)}(10, i-1)+P _{6, (2,10)}(11, i-1) equation (9)

Therefore, get rid of bit 2 outside the PCE 6 be 1 probability be get rid of outside the PCE 6 to d _(2,10)Be 10 probability and get rid of outside the PCE 6 to d _(2,10)It is the summation of 11 probability.

Joint probability outside the eliminating PCE 5 of 01 pattern is upgraded a little a bit difference, because this calculating comprises PCE 1, this PCE 1 itself comprises bit 1 and 5.Therefore, only remaining bit is a bit 9 in PCE 1.So, provide and 1 corresponding of PCE as followsly:

$P_{5,\left(\mathrm{1,5}\right)}(01.i)=P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right)\mathrm{Pr}\left\{\mathrm{PCE}1\mathrm{met}\right|\left(\mathrm{1,5}\right)=01\left\}\mathrm{Pr}\right\{\mathrm{PCE}6\mathrm{met}|\left(\mathrm{1,5}\right)=01\}$ Equation (10)

$=P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right){\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">P</figure-callout>}}_{1,9}(i-1)\&CenterDot;\left(\frac{1-(1-2P_{\mathrm{6,2}}(i-1))(1-2P_{\mathrm{6,12}}(i-1))}{2}\right)$ Equation (11)

When decoding circuit 10 was got rid of PCE 6, decoding circuit 10 was carried out similarly and is upgraded, and will be appreciated that for this example, similarly upgraded at each appearance of six bit centerings, and these six bits are to being d _(1,5), d _(2,10), d _(3,9), d _(4,7), d _(6,11)And d _(8,12)

In first time iteration, the probability on equation right side is initial value, is eigenvalue.In iteration (iteration in the middle of this is also referred to as) subsequently, determined value in the iteration before the probability tables of right-hand side is shown in.In last iteration, decoding circuit 10 to calculating independent one group of final joint probability, is assessed these relevant right joint probabilities of bit to each relevant bit.For example:

$P_{\left(\mathrm{1,5}\right)}(01,I)=$

$P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right)\mathrm{Pr}\left\{\mathrm{PCE}1\mathrm{met}\right|\left(\mathrm{1,5}\right)=01\left\}\mathrm{Pr}\right\{\mathrm{PCE}5\mathrm{met}|\left(\mathrm{1,5}\right)=01\}\mathrm{Pr}\left\{\mathrm{PCE}5\mathrm{met}\right|\left(\mathrm{1,5}\right)=01\}$

Equation (12)

$=P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right){\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{1,9}}(I-1)(P_{5,\left(\mathrm{6,11}\right)}(00,I-1)+P_{5,\left(\mathrm{6,11}\right)}(11.I-1))\&CenterDot;\left(\frac{1-(1-2P_{\mathrm{6,2}}(I-1))(1-2P_{\mathrm{6,12}}(I-1))}{2}\right)$

Equation (13)

In case obtained above joint probability, just can use these joint probabilities in every way.For example, decoding circuit 10 can be assessed the pattern that each joint probability also selects to have maximum probability.In the environment of above example, at each bit of assessing to (d _(1,5), d _(2,10), d _(3,9), d _(4,7), d _(6,11)And d _(8,12)), joint probability is handled and is launched the probable value of each possible bit to pattern (00,01,10,11).Determined jointly that to picking up pattern this is to bit at each bit with maximum probability value.Therefore can jointly determine hard decision for the data block bits that receives.Replacedly, can determine single-bit probabilities, and can carry out hard decision to each bit based on single-bit probabilities (for example referring to equation (9)) according to joint probability.

Notably, on the basis of picking up " the best " joint probability the data block that receives being decoded makes continuous error rate minimize usually, and pick up best single-bit probabilities each bit error is minimized.Therefore, the characteristic of just received data (and communication system that may be related and the type of application) can make a kind of method than another kind more preferably.

As further alternative, joint probability can be used as the basis that is used to other bits that the initial joint probabilities value is set, rather than the final joint probability of calculating of utilization drives the hard decision to the data bit that receives.For example, can be by determining joint probability for the ISN bit, and utilize those joint probabilities to come the joint probability deterministic process of the outer sign indicating number of initialization bit then, cascaded code or inside/outside block code are decoded.Therefore, for example, if utilize ISN that the data block that receives is encoded, and be to utilize the LDPC sign indicating number to encode then, then can to utilize ISN be that the bit subset of being correlated with is calculated joint probability to decoding circuit 10, and then those joint probabilities are passed to outer sign indicating number LDPC decoding.

If expectation, even then when not utilizing cascaded code that the data block that receives has been advanced coding, decoding circuit 10 also can be used following method, is about to be fed as the initialization value that the joint probability of outer sign indicating number bit is determined at the determined joint probability of ISN.For example, the parity matrix of equation (1) can be considered as comprising first block code and the Parity Check Bits of following.Utilize this method, regard the bit among the PCE 19 parity check of bit 1 and 5 as, and bit 10 plays the effect of the parity check of bit 2 and 6.Below underscore in equation (14) represented a kind of method, promptly the selected bit in the Parity Check Bits in the H matrix is treated as the Parity Check Bits of other bit subset in this matrix.

$H=\left[\begin{array}{c}10001000\underset{\&OverBar;}{1}000\\ 010001000\underset{\&OverBar;}{1}00\\ 0010001000\underset{\&OverBar;}{1}0\\ 00010001000\underset{\&OverBar;}{1}\\ 1000010000\underset{\&OverBar;}{1}0\\ 01001000000\underset{\&OverBar;}{1}\\ 00100001\underset{\&OverBar;}{1}000\\ 000100100\underset{\&OverBar;}{1}00\end{array}\right]$ Equation (14)

Continue above example in more detail, be divided into ISN and outer sign indicating number corresponding to the LDPC sign indicating number of parity check matrix H, wherein bit 9 serves as the parity check of bit 1 and 5, or the like.Therefore, can utilize 2/3 yard of simple rate (input bit 1 and 5, output bit 1,5 and 9) to create bit (1,5,9).Identical encoding relation extends to (bit 2,6,10), (3,7,11) and (4,8,12).Therefore, the interior decode procedure of this embodiment of decoding circuit 10 utilizes bit 9,10,11 and 12 to determine joint probability for bit (1,5), (2,6), (3,7) and (4,8).For example, (1,5) be that 01 probability can be that bit 1 is 0, bit 5 be 1 and bit 9 be 1 probability product " A " doubly.Determine the value of A, so that these joint probabilities amount to 1.After such processing, abandon ISN bit 9 to 12, and " remaining " outer sign indicating number can be the following LDPC sign indicating number that provides:

$\left(\begin{array}{c}10100110\\ 01011001\\ 10101001\\ 01010110\end{array}\right)$ Equation (15)

Utilize determined these joint probabilities, the initialization value that decoding circuit 10 recomputates these joint probabilities as iteration, or as the basis at (1,5), (2,6), final joint probability calculation that (3,7) and (4,8) related bits is right.

Though the inside/outside decoding strengthens the particular procedure embodiment that has represented decoding circuit 10, but Fig. 3 shows the wide in range embodiment that joint probability is determined method, and this joint probability determines that method can be implemented in one or more treatment circuits that comprise decoding circuit 10.For example should be appreciated that decoding circuit 10 can comprise all baseband processing circuitries or part baseband processing circuitry, such as digital signal processor or application-specific integrated circuit (ASIC) (ASIC), and decoding circuit 10 can be based on hardware or software or its combination in any.In at least one embodiment, decoding circuit 10 comprises computer product, such as computer program instruction code or (synthesizable) logical file that can be comprehensive.In other embodiments, decoding circuit 10 is fixed in the integrated circuit (IC) apparatus as hardware or software.

Under any circumstance, below handling all be that joint probability definite (this method directly extends to other big or small related bits subclass) in bit-pair context provides general formula.Though graphic extension provides the equation of determining at 01 bit mode, should be appreciated that decoding circuit 10 also carries out this equation at other to pattern (00,10,11).Therefore, can be with γ (n ₁, n ₂) be defined as the set of PCE, consequently Or

All in this equation.

It is that each is to (n that processing starts from as follows ₁, n ₂) calculating intrinsic joint probability value (step 110):

$P_{({\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}}_1,n_2)}^{\mathrm{int}}\left(01\right)=(1-{\mathrm{<figure-callout\; id="1"\; label="P\; n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">P</figure-callout>}}_{n_{}}^{{}_1})P_{n_2}^{\mathrm{int}}$ Equation (16)

On the one or many iteration, upgrade these joint probabilities that are initialised (step 112 and 114) then.For for the first time such iteration, the soft value of joint probability is set to determined initial value in step 110.That is, for j ∈ γ (n ₁, n ₂),

$P_{j({\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}}_1,n_2)}\left(\mathrm{01,0}\right)=P_{({\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}}_1,n_2)}^{\mathrm{int}}\left(01\right)$ Equation (17)

Wherein, symbol

Expression is determined at the probability of 01 bit mode of the 0th (for the first time) iteration.For each the i time iteration subsequently, i=1, ... I-1, at all possible bit combination (pattern) in the just evaluated associated subset, decoding circuit 10 upgrades these and unites soft probability by recomputate the soft probability of associating according to equation (8) and equation (11).

Some embodiment of decoding circuit 10 always recomputates the renewal of carrying out joint probability by iteration.Even so, still such embodiment can be disposed to come according to the iteration metric restriction or otherwise determine the number that iteration recomputates.In one embodiment, iteration metric is defined number, in the nonvolatile memory that can be maintained in the receiver 12.In another embodiment, decoding circuit 10 is derived iteration metric based on iteration result (for example, the variation of the joint probability soft on the one or many iteration) one by one.Another iteration metric is to determine hard decision and check outer error detecting code (such as CRC), to check whether remain any mistake.In another embodiment, decoding circuit 10 is according to regulating iteration metric by channel quality estimation circuit 32 estimated channel quality or the intensity of Fig. 1.

In such embodiments, if relatively low channel quality (or signal strength signal intensity) situation is preponderated, then decoding circuit 10 can be realized the iteration more than once.Yet, if estimated channel quality (or intensity) has exceeded predetermined threshold value, this can be provided with according to valid data or type of service, then decoding circuit 10 can " convert " into single-shot joint probability and determines, wherein utilize the value of initialization to come these joint probabilities of initialization, and directly carry out final joint probability calculation.

In any case for final iteration ("Yes" that comes from step 114), decoding circuit 10 upgrades joint probability (step 116) by carry out final calculating according to equation (13).Decoding circuit 10 and may be that other circuit in the receiver 12 are carried out desired reprocessing (step 118).For example, decoding circuit 10 can be carried out the hard decision estimation to the bit in the data block that receives based on joint probability, as before this explains.Additional treatment circuit 30 can further be handled these hard bits, and these hard bits can comprise control/signaling information, user/application data, voice etc.

In one or more embodiments, decoding circuit 10 is by approximate incorporate its joint probability into and determine to improve in the method its performance and/or simplify its operation in above processing environment one or more.For example, decoding circuit 10 can be simplified its calculating by utilizing corresponding single-bit probabilities to be similar to joint probability, rather than utilizes the joint probability of all bits in the associated subset, at this associated subset assessment joint probability.For example, in the environment as the dibit subclass on joint probability basis, the single-bit that provides equation (8) approximate as followsly:

$P_{\left(\mathrm{1,}(\mathrm{1,5}\right)}\left(01\right)=P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right)\mathrm{Pr}\left\{\mathrm{PCE}5\mathrm{met}\right|\left(\mathrm{1,5}\right)=01\left\}\mathrm{Pr}\right\{\mathrm{PCE}6\mathrm{met}|\left(\mathrm{1,5}\right)=01\}$ Equation (18)

$=P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right)\left(\frac{1+(1-2P_{5,6})(1-2P_{5,11})}{2}\right)\left(\frac{1-(1-2P_{\mathrm{6,2}})(1-2P_{\mathrm{6,12}})}{2}\right)$ Equation (19)

Can also utilize logarithm probability and log-likelihood ratio (LLR) to come work decoding circuit 10 configurations, rather than directly utilize probable value to come work.Utilize for simplicity and the abridged iteration index, following equation provides the basis that is used for realizing at decoding circuit 10 one or more embodiment of joint probability decoding, but be to use calculating based on logarithm, should be appreciated that these equatioies have used bit 1 and 5 and from the parity check matrix H environment as an example before the equation (1):

$\log P_{\left(\mathrm{1,5}\right)}\left(01\right)=\log P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right)+\mathrm{<figure-callout\; id="1"\; label="log\; P"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">log</figure-callout>}{P}_{}{}_{\mathrm{1,9}}+\log \left(\frac{1+(1-2P_{\mathrm{5,6}})(1-2P_{\mathrm{5,11}})}{2}\right)$

$+\log \left(\frac{1-(1-2P_{\mathrm{6,2}})(1-2P_{\mathrm{6,12}})}{2}\right)$

$=\log P_{\left(\mathrm{1,5}\right)}^{\mathrm{int}}\left(01\right)+\mathrm{<figure-callout\; id="1"\; label="log\; P"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">log</figure-callout>}{P}_{}{}_{\mathrm{1,9}}+\log \left(\frac{1+\tanh (\mathrm{LL}{R}_{\mathrm{5,6}}/2)\tanh (\mathrm{LL}{R}_{\mathrm{5,11}}/2)}{2}\right)$

$+\log \left(\frac{1-\tanh (\mathrm{LL}{R}_{\mathrm{6,2}}/2)\tanh (\mathrm{<figure-callout\; id="12"\; label="LL\; R"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">LL</figure-callout>}{R}_{}{}_{\mathrm{6,12}}/2)}{2}\right)$ Equation (20)

If n is (n ₁, n ₂) in n ₁, then

$\mathrm{LL}{R}_{j,n}=\log \left(\frac{e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(00\right)}+e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(01\right)}}{e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(10\right)}+e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(11\right)}}\right)$ Equation (22)

$=\log \left(\frac{e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(00\right)}(1+e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(01\right)-\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(00\right)})}{e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(10\right)}(1+e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(11\right)-\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(10\right)})}\right)$ Equation (23)

$=\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(00\right)+\log (1+\frac{P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(01\right)}{P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(00\right)})$

$-\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(10\right)-\log (1+\frac{P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(11\right)}{P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(10\right)})$ Equation (24)

And, if n is (n ₁, n ₂) in n ₂, then

$\mathrm{LL}{R}_{j,n}=\log \left(\frac{e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(00\right)}+e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(10\right)}}{e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(01\right)}+e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(11\right)}}\right)$ Equation (25)

$=\log \left(\frac{e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(00\right)}(1+e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(10\right)-\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(00\right)})}{e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(01\right)}(1+e^{\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(11\right)-\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(01\right)})}\right)$ Equation (26)

$=\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(00\right)+\log (1+\frac{P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(10\right)}{P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(00\right)})$

$-\log P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(01\right)-\log (1+\frac{P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2}\left(11\right)}{P_{(\mathrm{<figure-callout\; id="1"\; label="n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">n</figure-callout>}1,n2)}\left(01\right)})$ Equation (27)

In addition, decoding circuit 10 can be configured to by be similar to item according to following equation

Product and reduce above complexity of calculation based on logarithm:

$\mathrm{\&Pi;}\tanh (\mathrm{<figure-callout\; id="1"\; label="LL\; R\; n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">LL</figure-callout>}{R}_n{1,n2}_{}/2)\&ap;\mathrm{sgn}\left(\mathrm{\&Pi;}\tanh (\mathrm{<figure-callout\; id="1"\; label="LL\; R\; n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">LL</figure-callout>}{R}_n{1,n2}_{}/2)\right)\underset{n}{\mathrm{<figure-callout\; id="1"\; label="min\; n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">min</figure-callout>}}\left|\tanh (\mathrm{<figure-callout\; id="1"\; label="LL\; R\; n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">LL</figure-callout>}{R}_n{1,n2}_{}/2)\right|$ Equation (28)

This equation further is reduced to:

$\mathrm{sgn}(\mathrm{\&Pi;<figure-callout\; id="1"\; label="LL\; R\; n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">LL</figure-callout>}{R}_n{1,n2}_{})\underset{n}{\mathrm{<figure-callout\; id="1"\; label="min\; n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">min</figure-callout>}}\left|\tanh (\mathrm{<figure-callout\; id="1"\; label="LL\; R\; n"\; filenames="A20068002157000181.png"\; state="\{\{state\}\}">LL</figure-callout>}{R}_n{1,n2}_{}/2)\right|$ Equation (29)

So that decoding circuit 10 can be similar to its probability assessment based on the symbol of function given in the equation (29).

No matter whether decoding circuit 10 is disposed the simplification mathematical procedure that uses equation (24), equation (27) and/or equation (29), can incorporate these decoding circuit 10 configurations into other treatment variable, these treatment variables are providing the performance that is enhanced or be simplified under some situation at least.For example, decoding circuit 10 can be configured to " mixing " decoder.Under the situation of the decoding situation of mixing, decoding circuit 10 is determined that joint probability incorporate in its some decoding calculating, rather than other decoding is calculated.For example, the iteration first time that probability is determined has been used joint probability, and decoding circuit 10 extracts single-bit probabilities from those results then.In iteration subsequently, decoding circuit 10 uses these single-bit probabilities that derive, rather than advances the joint probability calculation that is more related to by this iterative process.

Certainly, the present invention is not limited to above-mentioned discussion, also is not limited to accompanying drawing.In fact, the present invention is only limited by following claims and content with equal legal effect thereof.

Claims (32)

1, a kind of method that the data block that receives is decoded, the described data block that receives is encoded by the linear block codes of being represented by parity matrix, and this method comprises:

The joint probability of the bit subset in the described data block that receives of initialization;

Upgrade described joint probability based on parity matrix; And

Utilize the joint probability of upgrading to recover coded data from the described data block that receives.

2, method according to claim 1, wherein, the joint probability of the bit subset in the described data block that receives of initialization comprises: utilize parity matrix that bit subset is designated the coupling bit in the described data block that receives and the probable value of the possible bit value combinations in each bit subset of initialization.

3, method according to claim 1, wherein, linear block codes comprises low-density checksum (LDPC) sign indicating number, and wherein parity matrix comprises one group of parity check equations, and each parity check equations has identified the coupling bit in the described data block that receives.

4, method according to claim 1 wherein, is upgraded described joint probability based on parity matrix and is comprised making up according to the relevant joint probabilities of represented all of parity matrix and upgrade described joint probability.

5, method according to claim 1 wherein, is upgraded described joint probability based on parity matrix and is comprised: for the one or many iteration, recomputate joint probability; And the joint probability that will recomputate is as the initial value of next iteration; And in the end calculate final joint probability in the iteration.

6, method according to claim 5 wherein, in the iteration, recomputates joint probability and comprises in the middle of each: for each joint probability, gets rid of the corresponding equation in the parity check equations from this recomputates, to avoid recomputating deviation.

7, method according to claim 6, wherein, in the end calculating final joint probability in the iteration comprises: for each joint probability, calculate final joint probability by the corresponding equation that comprises in the parity check equations in final joint probability calculation.

8, method according to claim 1, wherein, utilize the joint probability of upgrading to recover coded data and comprise one of following operation: the hard bit decision of determining the described data block that receives on the basis of selecting maximum joint probability from the described data block that receives; On the basis of selecting maximum single-bit probabilities, determine the hard bit decision of the described data block that receives; Perhaps provide the joint probability of upgrading as the soft information that is used for the described data block that receives is carried out outer-decoder.

9, method according to claim 1 further comprises joint probability is expressed as log-likelihood.

10, method according to claim 9, further comprise single-bit probabilities is expressed as according to the determined log-likelihood ratio of log-likelihood, and wherein upgrading described joint probability based on parity matrix comprises: upgrade each in the iteration for one or many, identify maximum probability by the value of symbol according to the hyperbolic tangent function that is used for representing log-likelihood ratio and revise joint probability estimation.

11, method according to claim 1, wherein, upgrade the RST that described joint probability comprises that assessment receives based on parity matrix, and, carry out the iteration or non-iterative the recomputating of joint probability selectively according to parity matrix based on described assessment.

12, method according to claim 1, wherein, upgrading described joint probability based on parity matrix comprises: the related bits information of utilizing parity matrix to provide recomputates joint probability iteratively, and stops recomputating of described iteration on the basis of assessment iteration metric.

13, method according to claim 1, wherein, the joint probability of the bit subset in the described data block that receives of initialization comprises one of following operation: generate initial joint probability according to the soft value of the single-bit that demodulating process provided; Perhaps generate initial joint probability according to the associating bit soft value that demodulating process provided.

14, method according to claim 1, wherein, upgrading described joint probability based on parity matrix comprises: utilize the related bits information in the parity matrix to upgrade joint probability according to the iteration of desired number, the joint probability that utilization is upgraded the renewal that obtains by iteration is determined single-bit probabilities, and recovers coded data based on single-bit probabilities.

15, a kind of decoding circuit, it is configured to the data block that receives is decoded, the described data block that receives is encoded by the linear block codes of being represented by parity matrix, and this decoding circuit comprises one or more treatment circuits, and described treatment circuit is configured to:

The joint probability of the bit subset in the described data block that receives of initialization;

Upgrade described joint probability based on parity matrix; And

Utilize the joint probability of upgrading to recover coded data from the described data block that receives.

16, decoding circuit according to claim 15, wherein, decoding circuit is configured to: by utilizing parity matrix bit subset is designated coupling bit in the described data block that receives, and the probable value by the possible bit value combinations in each bit subset of initialization, the joint probability of the bit subset in the described data block that receives of initialization.

17, decoding circuit according to claim 15, wherein, linear block codes comprises low-density checksum (LDPC) sign indicating number, and wherein, parity matrix comprises one group of parity check equations, the coupling bit in the described data block that receives of each parity check equations sign.

18, decoding circuit according to claim 15, wherein, decoding circuit is configured to: upgrade described joint probability by making up according to all represented relevant joint probabilities of parity matrix, upgrade described joint probability based on parity matrix.

19, decoding circuit according to claim 15, wherein, decoding circuit is configured to: by recomputating joint probability at the one or many iteration, and by the joint probability that will recomputate initial value as next iteration, and, upgrade described joint probability based on parity matrix by in the end calculating final joint probability in the iteration.

20, decoding circuit according to claim 19, wherein, decoding circuit is configured to: by getting rid of corresponding equation in the parity check equations at each joint probability avoiding recomputating deviation from recomputate, recomputate joint probability in the iteration in the middle of each in the middle of one or many in the iteration.

21, decoding circuit according to claim 20, wherein, decoding circuit is configured to: by calculating final joint probability at each joint probability by the corresponding equation that comprises in the parity check equations in final joint probability calculation, in the end calculate final joint probability in the iteration.

22, decoding circuit according to claim 15, wherein, decoding circuit is configured to recover coded data by carrying out the joint probability that one of following operation utilization upgraded from the described data block that receives: the hard bit decision of (a) determining the described data block that receives on the basis of selecting maximum joint probability; (b) the hard bit decision of definite described data block that receives on the basis of selecting maximum single-bit probabilities; Perhaps (c) provides the joint probability of upgrading as the soft information that is used for the described data block that receives is carried out outer-decoder.

23, decoding circuit according to claim 15, wherein, decoding circuit is configured to joint probability is expressed as log-likelihood.

24, decoding circuit according to claim 23, wherein, decoding circuit is configured to: single-bit probabilities is expressed as according to the determined log-likelihood ratio of log-likelihood; And at each iteration in one or more renewal iteration, come to upgrade described joint probability based on parity matrix by revising joint probability estimation based on the sign maximum probability, wherein the value of symbol according to the hyperbolic tangent function that is used for representing log-likelihood ratio by decoding circuit identifies maximum probability.

25, decoding circuit according to claim 15, wherein, decoding circuit is configured to: the RST that receives by assessment, and based on described assessment by carrying out iteration or non-iterative recomputating according to parity matrix selectively to joint probability, upgrade described joint probability based on parity matrix.

26, decoding circuit according to claim 15, wherein, decoding circuit is configured to: recomputate joint probability iteratively by the related bits information of utilizing parity matrix to provide, and, upgrade described joint probability based on parity matrix by on the basis of assessment iteration metric, stopping recomputating of this iteration.

27, decoding circuit according to claim 15, wherein, decoding circuit is configured to the joint probability of the bit subset in the described data block that receives of initialization by carrying out one of following operation: (a) generate initial joint probability by the soft value of the single-bit that demodulating process provided; Perhaps (b) generates initial joint probability by the associating bit soft value that demodulating process provided.

28, decoding circuit according to claim 15, wherein, decoding circuit comprises initializing circuit, and this initializing circuit is configured to initialization one class value, described class value is represented each joint probability, and described class value is corresponding to the bit combination of the supposition that is comprised in each joint probability.

29, decoding circuit according to claim 28, wherein, decoding circuit further comprises counting circuit that is configured to upgrade described joint probability and the evaluation circuits that is configured to provide soft or hard decision, and described soft or hard decision is used for recovering coded data on assessment comes from the basis of joint probability of renewal of counting circuit.

30, decoding circuit according to claim 29, wherein, counting circuit comprises the iteration control circuit or is associated with the iteration control circuit that this iteration control circuit is configured to control by the performed a plurality of joint probabilities of counting circuit and recomputates iteration.

31, decoding circuit according to claim 15, wherein, decoding circuit is configured to: come iteration to upgrade joint probability by utilize the related bits information in the parity matrix according to the iteration of desired number, upgrade described joint probability based on parity matrix; And utilize the joint probability of the renewal that obtains by the iteration renewal to determine single-bit probabilities; And recover coded data based on single-bit probabilities.

32, a kind of method that linear block codes is decoded, it comprises:

Based on the bit relationships that first identified by parity matrix, first joint probability of first subclass of the related bits in the data block of determining to receive; And

First joint probability as initial value, is used for determining second joint probability of second subclass of the related bits of the data block that receives.

Images