Method and device for decoding low-density parity check code and optical information reproducing apparatus using the same

The present invention relates to a decoding method and a decoding device of a low density parity check (hereinafter referred to as LDPC), and an optical information reproduction device using the decoding method.

Optical information processing devices that process optical information include CDs (compact discs), DVDs (digital versatile discs), HD-DVDs, Blu-ray discs, and near-field optical processing devices. Recently, in response to the increasing demand for next-generation storage systems having a large capacity storage capacity, optical information processing devices using holography have been attracting attention.

An optical information processing device using holography is a page-oriented memory in principle of recording and reproducing image information, and uses a parallel signal processing input / output system, so that it is a bit unit system. The data transmission rate can be fundamentally increased compared to CDs and DVDs. In addition, a multiplexing technique for superimposing and recording image information on the same location on a recording medium can be used, and the storage density can be dramatically improved.

An optical information processing device using holography superimposes information light (information beam) including image information of original data and reference light (reference beam) on an optical information recording medium (recording medium) and irradiates the light. The interference pattern is recorded on an optical information recording medium. In order to reproduce the recorded light information, when the optical information recording medium is irradiated with the reference light, the reference light is diffracted by the interference pattern to generate the reproduced light.

The image of the data page reproduced via the regenerated light is detected using a light receiving array element such as a CMOS sensor (complementary metal-oxide semiconductor sensor) or a CCD (charge coupled device). The detected data page is restored to the original data through a series of signal processing and decoding processes.

However, when detecting an image of a data page, an error may occur due to the characteristics of the channel due to shrinkage or rotation of the optical information recording medium. For example, due to mis-alignment, the image pixels of the data page (hereinafter referred to as'data pixels') and the pixels of the light receiving array element (hereinafter referred to as'detection pixels') may not be matched with each other. The above error results in a high bit error rate (BER).

Various error correction codes such as the Reed-Solomon code have been proposed to lower the BER. Recently, the use of LDPC codes with error correction performance that is close to the Shannon limit has been proposed.

The LDPC code is a linear block code in which the elements of the parity check matrix are mostly '0'. A general parity check code is composed of a block composed of information symbols and a parity check symbol which is a modulo sum of a specific information symbol, and forms one code word. The relationship between this check symbol and the information symbol can be shown by the parity check matrix H. The parity check matrix H is represented by a set of linear homogeneous equations. That is, the LDPC code is a kind of parity check code, and is a coding method using a parity check matrix H in which most of the elements are 0 and simply have a small number and randomly scattered weighted values.

The coding process of the LDPC code having the parity check matrix H is as follows. When the above H matrix is obtained, the generator matrix G corresponding to the H matrix is obtained by using the relation of GHT = 0. The codeword C corresponding to the information symbol block X is calculated by C = XG. Regular LDPC code if the number of "1" s per column is W and the number of "1" s per row is constant at W x (N / M) for the H matrix (M x N). That is. If the number of "1" s per column is not constant and the number of "1" s per row is not exactly W × (N / M), it is called an irregular LDPC code. Non-uniform LDPC codes generally have better error correction capability, but are more difficult to implement in hardware.

Decoding the LDPC code is to search the received signal vector for the most probabilistically approximate codeword whose product with the parity check matrix H satisfies 0. Of the LDPC code decoding methods, the sum-product algorithm performs soft decision iterative decoding using probability values. The sum-product algorithm iteratively decodes probabilistic messages on a coded graph so that they converge to codewords that meet the maximum likelihood criterion while exchanging messages with each other between nodes.

As another decoding method of LDPC code, there is an LLR-BP (belief propagation) algorithm that uses a log-likelihood ratio (LLR). The LLR-BP algorithm will be described below.

If the code word is c, the transmitted signal is x, the received signal is y, and the channel noise is n, then y = [yn] = x + n. Here, the code word c = (c1, c2, ..., cN) is mapped by the transmission signal x = (x1, x2, ..., xN). Decoding is the process of finding the signal that maximizes the probability of a codeword for the received signal. That is,<maths num="1"><img file="JP5118317B2_D0001.tif" /></maths>Codeword that maximizes the value of<maths num="2"><img file="JP5118317B2_D0002.tif" /></maths>Is to ask.

The magnitude of the parity check matrix H is M × N and is indicated by H = [h (m, n)]. The set of bit nodes corresponding to the mth check node is displayed as N (m) = {n | h (m, n) = 1}. Similarly, the set of check nodes corresponding to the nth bit node is displayed as M (n) = {m | h (m, n) = 1}. The magnitudes of the sets N (m) and M (n) are indicated by | N (m) | and | M (n) |, respectively.<maths num="3"><img file="JP5118317B2_D0003.tif" /></maths>Means N (m) excluding the nth bit.<maths num="4"><img file="JP5118317B2_D0004.tif" /></maths>Means M (n) excluding the mth check.

The notations used in the following iterative decoding algorithms are: Fn is the LLR of the nth bit obtained from the received signal yn. zmn the n-th bit to go to the m-th check node in the N-th bit node is the LLR of the bet. zn is the posteriori LLR of the nth bit calculated by each iteration. Lmn is the nth bit LLR that goes to the nth bit node at the mth check node. (1) Initialization step For each m and n<maths num="5"><img file="JP5118317B2_D0005.tif" /></maths>Initialize to. (2) Row-direction iterative decoding It is defined as follows for each m and n.<maths num="6"><img file="JP5118317B2_D0006.tif" /></maths><maths num="7"><img file="JP5118317B2_D0007.tif" /></maths>(3) Column-direction iterative decoding Update for each m and n as follows.<maths num="8"><img file="JP5118317B2_D0008.tif" /></maths><maths num="9"><img file="JP5118317B2_D0009.tif" /></maths>(4) Provisional decoding As follows<maths num="10"><img file="JP5118317B2_D0010.tif" /></maths>To decide.<maths num="11"><img file="JP5118317B2_D0011.tif" /></maths><maths num="12"><img file="JP5118317B2_D0012.tif" /></maths>if,<maths num="13"><img file="JP5118317B2_D0013.tif" /></maths>If so, the decryption process is stopped<maths num="14"><img file="JP5118317B2_D0014.tif" /></maths>Is determined to be the correct decoding result. if,<maths num="15"><img file="JP5118317B2_D0015.tif" /></maths>If the decoding has not reached the maximum number of iterations, the process is repeated from step (2) above. if,<maths num="16"><img file="JP5118317B2_D0016.tif" /></maths>When the decryption is executed the maximum number of iterations, the decryption is stopped and the decryption failure is declared.

In an optical information processing device using holography, two-dimensional inter-symbol interference occurs due to misalignment between the light receiving array element and the image of the data page. Therefore, more efficient and improved BER decoding of LDPC codes is required.<patcit num="1"><text>JP 2004-186940</text></patcit>

<p>The technical problem to be solved by the present invention is a method and apparatus for decoding a low density parity check code that corrects an error due to misalignment, and optical information using the same, in order to solve the above problems. The purpose is to provide a reproduction device. Another technical problem to be solved by the present invention is to provide an efficient low-density parity check code decoding method and a decoding device, and an optical information reproduction device using the same.</p>

<p>According to one embodiment of the present invention, there is provided a method for decoding an LDPC code that decodes a received signal encoded by an LDPC code. First,<u style="single">Of the LDPC code</u>bit<u style="single">node</u>The initial value of the above received signal<u style="single">Using</u>initialize.<u style="single">Based on the initial value of the received signal, the LDPC code</u>In the row and column directions<u style="single">, Said bit node</u>Bits above after iterative decoding<u style="single">node</u>After the fact, it is determined whether or not to proceed with the iterative decoding again according to the above-mentioned posterior value. When proceeding with iterative decoding, the above<u style="single">Bit node</u>Compare the ex post facto value with the set value<u style="single">, Based on the adjustment level due to the comparison and misalignment</u>The above bit<u style="single">node</u>Update the initial value of.</p><p>According to another embodiment of the present invention, there is provided an LDPC code decoding device that decodes a received signal encoded by the LDPC code. The above LDPC code decoding device<u style="single">, The LDPC code</u>bit<u style="single">node</u>The initial value of the above received signal<u style="single">Using</u>The initialization part to initialize and<u style="single">Based on the initial value of the received signal, the LDPC code</u>In the row and column directions<u style="single">, Said bit node</u>Bits above after iterative decoding<u style="single">node</u>An iterative decoding unit that obtains the posterior value of<u style="single">Bit node</u>After the fact<u style="single">Setting</u>Compare with the fixed value<u style="single">, Based on the adjustment level due to the comparison and misalignment</u>the above<u style="single">LDPC code</u>bit<u style="single">node</u>Includes a partial compensation unit that updates the initial value of.</p><p>According to another embodiment of the present invention, there is provided an optical information reproducing device that reproduces optical information via reproduced light generated by incidenting reference light into an optical information recording medium. The optical information reproduction device receives a signal of an optical information detector that detects the reproduced light and detects an image of a data page, an equalizer that equalizes the image of the data page, and an output of the equalizer. It includes a data decoding unit that decodes the LDPC code.<u style="single">The data decoding unit initializes the bit node of the LDPC code using the initial value of the received signal, and the row direction and the column direction of the LDPC code based on the initial value of the received signal. Then, an iterative decoding unit that repeatedly decodes the bit node to obtain the posterior value of the bit node, an iterative determination unit that determines whether to proceed with the iterative decoding again according to the posterior value, and an iterative decoding unit. When proceeding, it includes a partial compensation unit that compares the ex post value of the bit node with the set value and updates the initial value of the bit node of the LDPC code based on the adjustment level due to the comparison and misalignment.</u></p>

<p>According to the present invention, it is possible to increase the processing speed of decoding the LDPC code via the modified LLR-BP algorithm, and the BER can be improved. Therefore, it is possible to improve the reliability of the optical information processing apparatus using holography with severe misalignment.</p>

Hereinafter, desirable embodiments of the present invention will be described in detail with reference to the accompanying drawings. The same reference number indicates the same component throughout the specification.

FIG. 1 is a block diagram showing an optical information processing apparatus according to an embodiment of the present invention. The optical information processing apparatus puts the data on the information light, causes the information light to enter the optical information recording medium together with the reference light, and records the optical information. In addition, the data is reproduced via the reproduced light that is reproduced by incidenting only the reference light on the optical information recording medium. As described above, the optical information processing device is an optical information recording / reproducing device capable of recording and reproducing optical information. As another embodiment, if only the function of reproducing optical information via reference light is provided by closing (making it non-transmissive) the spatial light modulator, this optical information processing apparatus can be used as an optical information reproducing apparatus. Can be done. As another embodiment, if only the function of closing the optical information detector (stopping the detection function) and recording optical information is provided, this optical information processing apparatus becomes an optical information recording apparatus.

Referring to FIG. 1, the optical information processor (100) includes a light source (110), a beam splitter (120), a multiplexing device (133), a spatial light modulator (140), and light. It includes an information detector (160), an equalizer (170), a data encoding unit (180) and a data decoding unit (200).

The light emitted from the light source (110) is separated into reference light (R) and information light (I) via an optical separator (120). The reference light (R) is reflected by the multiplexing device (133) through the first shutter (131) and is incident on the optical information recording medium (150) at a predetermined angle.

The information light (I) passes through the second shutter (134), changes its path by the reflector (134), and enters the spatial light modulator (140). At this time, the binary data of each encoded page provided by the data encoding unit (180), that is, the data page information is input to the spatial light modulator (140). The data encoding unit (180) encodes the input data with an LDPC code and provides it to the spatial light modulator (140) on a page-by-page basis.

The spatial optical modulator (140) optically modulates the data page information input from the data encoding unit (180) to generate a two-dimensional imaged data page, which is then used as the incident information light (I). ) And incident on the optical information recording medium (150).

When the reference light (R) and the information light (I) are incident on the optical information recording medium (150), the incident reference light (R) and the information light (I) are inside the optical information recording medium (150). A light induced generation of mobile charge occurs due to the intensity of the interference pattern generated by the interference with the light induced generation of mobile charge, and the interference pattern is recorded.

The multiplexing device (133) adjusts the angle at which the reference light (R) is incident on the optical information recording medium (150) to realize angle multiplexing. The multiplexing device (133) is preferably a rotating mirror such as a galvano mirror.

On the other hand, in order to reproduce the recorded data, it is advisable to irradiate the optical information recording medium (150) with only the reference light (R). During reproduction, the first shutter (131) passes the reference light (R) separated by the optical separator (120), while the second shutter (132) blocks the information light (I).

At this time, the reference light (R) that passes through the first shutter (131) and is emitted from the multiplexing device (133) is diffracted by the interference pattern recorded on the optical information recording medium (150) and is diffracted by the data page. Regenerated light with the image of is generated. The reproduced light is detected as an image of a data page by an optical information detector (160). The image of the detected data page is equalized via the equalizer (170) and decoded by the data decoding unit (200).

The optical information detector (160) is composed of a light receiving array element such as a CMOS sensor or a CCD. The equalizer (170) can use a well-known configuration such as the MMSE (minimum mean square error) equalizer. The data decoding unit (200) is a device that decodes the LDPC code. The data decoding unit (200) decodes the LDPC code output from the equalizer (170) and outputs the final output data.

On the other hand, the image pixels of the data page (hereinafter referred to as "data pixels") and the pixels of the light receiving array element (hereinafter referred to as "detection pixels") may not be matched with each other due to misalignment. Generally, when misalignment occurs within one pixel, it is affected by the surrounding eight pixels.

Figure 2 is an example of a diagram showing the case where misalignment occurs. Referring to FIG. 2, when misalignment occurs, the detected pixel (P) and the corresponding original data pixel (s0) do not match. In the absence of misalignment, the detected pixels (P) exactly match the original data pixels (s0).

When misalignment occurs, the bit data detected by the detection pixel (P) is affected by the three peripheral data pixels (s1, s2, s3) outside the original data pixel (s0). Here, the position, direction, and number of peripheral data pixels may vary, and are not necessarily limited to the illustrated form. For example, if misalignment occurs only in the vertical direction, the detected pixel (P) will be affected by one peripheral data pixel (s1).

Depending on the direction of misalignment, the bit data of the detected pixel (P) increases or decreases depending on the bit value of the peripheral data pixels (s1, s2, s3). For example, when the value of the data pixel (s0) is "0", the value of the detection pixel (X) is added to the value of the peripheral data pixels (s1, s2, s3) to increase the initial LLR value. If the values of the peripheral data pixels (s1, s2, s3) are all "1", the value of the detected pixel (X) may be "1", which is an erroneous result.

When the value of the data pixel (s0) is "1", the value of the detection pixel (P) may be added to the value of the peripheral data pixels (s1, s2, s3) to reduce the initial LLR value. If the values of the peripheral data pixels (s1, s2, s3) are all "0", the value of the detected pixel (P) can be "0", which is an erroneous result.

On the other hand, the opposite situation may occur as a result of misalignment. For example, if the value of the data pixel (s0) is "0" and the value of the peripheral data pixels (s1, s2, s3) is "0", the value of the detection pixel (P) is "0". May have a positive effect on decryption. Also, if the value of the data pixel (s0) is "1" and the value of the peripheral data pixels (s1, s2, s3) is "1", the value of the detection pixel (P) is "1", and so on. May have a positive effect on decoding.

Therefore, the present invention partially compensates for adverse effects on decoding performance as a result of misalignment in the iterative decoding step through a probabilistic iterative decoding process. Through this, the convergence speed can be increased and the BER can be decreased. FIG. 3 is a block diagram showing the data decoding section of FIG. Referring to FIG. 3, the data decoding unit (200) includes an initialization unit (210), an iterative decoding unit (220), an iterative determination unit (230) and a partial compensation unit (240).

The initialization unit (210) initializes the received signal. That is, the initialization unit (210) initializes each bit (zmn) to the LLR of the received signal. The iterative decoding unit (220) iteratively decodes each in the row direction and the column direction to obtain the post-liter LLR (zn) of the bit. The iterative decoding unit (220) first executes the row-direction iterative decoding, and then performs the column-direction iterative decoding using the result of the row-direction iterative decoding. The iterative decoding unit (220) can be divided into a row direction iterative decoding unit and a column direction iterative decoding unit.

Iterative Judgment (230) is a tentative codeword via the sign of the posterior LLR (zn)<maths num="17"><img file="JP5118317B2_D0017.tif" /></maths>Is determined, and whether or not the iterative decoding proceeds again is determined through this. The method of deciding whether or not to repeat will be described later with reference to FIG. If the iterative decoding succeeds and ends, the iterative decision unit (230) uses the codeword<maths num="18"><img file="JP5118317B2_D0018.tif" /></maths>Is output as output data.

When the iterative decoding proceeds again, the partial compensation unit (240) updates the initial values of the above bits. The partial compensation unit (240) compares the post-LLR (zn) with the preset LLR reference (LD) of the bit (P) to be compensated and updates the initial value. Here, the degree of updating of the initial value changes depending on the direction and size of the misalignment.

The decoding method using the data decoding unit (200) will be described below with reference to FIG. FIG. 4 is a flowchart of an LDPC code decoding method.

If the code word is c, the transmitted signal is x, the received signal is y, and the channel noise is n, then y = [yn] = x + n. Here, the codeword c = (c1, c2, ..., cN) is mapped by the transmission signal x = (x1, x2, ..., xN). Decoding is the process of finding the signal that maximizes the probability of a codeword for the received signal. That is,<maths num="19"><img file="JP5118317B2_D0019.tif" /></maths>Codeword that maximizes the value of<maths num="20"><img file="JP5118317B2_D0020.tif" /></maths>Is to ask.

The magnitude of the parity check matrix H is m × n and is indicated by H = [h (m, n)]. The set of bits corresponding to the mth check is displayed as N (m) = {n | h (m, n) = 1}. The set of checks corresponding to the nth bit is displayed as M (n) = {m | h (m, n) = 1}. The magnitudes of the sets N (m) and M (n) are indicated by | N (m) | and | M (n) |, respectively.<maths num="21"><img file="JP5118317B2_D0021.tif" /></maths>Means N (m) excluding the nth bit.<maths num="22"><img file="JP5118317B2_D0022.tif" /></maths>Means M (n) excluding the mth check.

The symbols used in the iterative decoding algorithm are: Fn is the LLR of the nth bit obtained from the received signal yn. zmn is the nth bit LLR that goes to the Mth check node at the nth bit node. zn is the posteriori LLR of the nth bit calculated in each iteration. Lmn is the nth bit LLR that goes to the nth bit node at the mth check node. p is the bit to be compensated for. s is a peripheral bit vector that affects bit p depending on the misalignment direction. For example, if the surrounding 3 bits affect bit p, then s = (s1, s2, s3). V is a vector indicating the direction and magnitude of misalignment. D (s, V) is the addition / subtraction level of bit p by s and V. Ld is the LLR criterion of bit p that determines whether Fp is updated or not. d0 and e0 are constants that determine the LLR that is reduced during the update. d1 and e1 are constants that determine the LLR that is increased during the update.

For each m and n in the initialization step<maths num="23"><img file="JP5118317B2_D0023.tif" /></maths>Initialize as follows (step S110). Once the initialization is complete, the following steps are for iterative decoding.

Row-wise iterative decoding steps for each m and n<maths num="24"><img file="JP5118317B2_D0024.tif" /></maths><maths num="25"><img file="JP5118317B2_D0025.tif" /></maths>Is defined as (step S120).

Following the row-direction iterative decoding step, the column-direction iterative decoding step is for m and n, respectively.<maths num="26"><img file="JP5118317B2_D0026.tif" /></maths><maths num="27"><img file="JP5118317B2_D0027.tif" /></maths>Update as per (step S130).

The provisional decryption step is<maths num="28"><img file="JP5118317B2_D0028.tif" /></maths><maths num="29"><img file="JP5118317B2_D0029.tif" /></maths>Using<maths num="30"><img file="JP5118317B2_D0030.tif" /></maths>(Step S140). next,<maths num="31"><img file="JP5118317B2_D0031.tif" /></maths>The iteration is determined according to the value of (step S150). if,<maths num="32"><img file="JP5118317B2_D0032.tif" /></maths>If, the decryption process is stopped and<maths num="33"><img file="JP5118317B2_D0033.tif" /></maths>Is determined to be the correct decoding result (step S155). if,<maths num="34"><img file="JP5118317B2_D0034.tif" /></maths>If, it is determined whether or not to repeat the decoding for the maximum number of iterations (step S160). After decrypting for the maximum number of iterations, abort decryption and declare decryption failure (step S165). If the decoding is not performed for the maximum number of iterations, probabilistic partial compensation is performed (step S170).

if,<maths num="35"><img file="JP5118317B2_D0035.tif" /></maths>When it is<maths num="36"><img file="JP5118317B2_D0036.tif" /></maths>Calculate d0 as follows. At this time, update to Fn = Fn-d0e0.

if,<maths num="37"><img file="JP5118317B2_D0037.tif" /></maths>Then, mathematical formula 6<maths num="38"><img file="JP5118317B2_D0038.tif" /></maths>Calculate d1 as follows. At this time, update to Fn = Fn-d1e1. The initial value is<maths num="39"><img file="JP5118317B2_D0039.tif" /></maths>Is updated. The degree to which the initial value is updated depends on the direction and size of the misalignment.

After updating the initial value as described above, the iterative decoding is executed again from the row direction iterative decoding step (step S120). Ld, e0, and e1 are values that do not change during the decoding process, and are appropriately selected and adjusted according to the channel status. It is desirable to make e 0 larger than e 1. In the case of D (s, V) used when calculating d0 and d1, it is related to the intensity of the part added by misalignment. However, since it is difficult to calculate D (s, V) accurately, it may be calculated so as to be proportional to the added width and the distance from the center of the pixel. As the probability value P (s), the value calculated at the time of provisional decoding may be used.

According to the present invention, among the bits that are larger or smaller than the original value due to misalignment, the initial LLR of the bits whose probability becomes a certain level or more in the decoding process is changed. Along with this, the convergence speed of the LLR-BP algorithm can be increased and the BER can be decreased. That is, it compensates according to the ratio of the influence of the misalignment to the initial LLR of the bit presumed to have a normal LLR despite being affected by the misalignment. As a result, by increasing the probability of the bit having the LLR in the correct direction, the error correction effect of the bit having the LLR in the wrong direction is also increased.

FIG. 5 is a graph showing the performance of the decoding method in terms of noise deviation vs. BER. Uncoded is the case where the equalizer and LDPC code are not used, and MMSE is the case where only the MMSE equalizer is used. The MMSE equalizer used a well-known pattern that was large enough and was convolutionized by 3x3. "MMSE + LDPC" is a case where the decoding method by the LLR-BP algorithm by the conventional technology is used for the MMSE equalizer. Proposed is the case where the decoding method by the modified LLR-BP algorithm according to the present invention is used for the MMSE equalizer.

The Nyquist size was set to 1, and the misalignment was set to 1/8 in the horizontal direction and 3/8 in the vertical direction. The LDPC code was set to 2500 in length, and simulation was performed for each of the cases of coding rates of 0.7, 0.8, and 0.9. Also, in the modified LLR-BP algorithm, Ld = 0.2, e0 = 0.633, e1 = 0.1267 were set, and D (S, V) was calculated as the ratio of the area to be added for simple calculation. .. P (s) was determined during the provisional decoding<maths num="40"><img file="JP5118317B2_D0040.tif" /></maths>Was used.

As a result of lowering the BER with the MMSE equalizer and performing decoding with an LDPC code that has strong error correction capability, the BER can be dramatically improved compared to using only "uncoded" or the equalizer. Understood. Further, as the coding rate increases, the decoding method according to the present invention has a lower BER than the decoding method according to the prior art. In particular, when the coding rate is 0.9, most of the errors that cannot be corrected by the LLR-BP algorithm according to the prior art are corrected by using the LLR-BP algorithm corrected by the present invention.

<figref num="1">It is a block diagram which showed the optical information processing apparatus by one Example of this invention.</figref><figref num="2">It is an example of the figure which showed the case where the misalignment occurred.</figref><figref num="3">It is a block diagram which showed the data decoding part of FIG.</figref><figref num="4">It is a flowchart about the decoding method of the LDPC code.</figref><figref num="5">It is a graph which showed the performance of a decoding method by noise deviation vs. BER.</figref>

Code description

100 Optical information processing device 110 light source 120 optical separator 131 First shutter 132 Second shutter 133 Multiplexer 134 reflector 140 Spatial Light Modulator 150 Optical information recording medium 160 Optical information detector 170 equalizer 180 Data encoding section 200 Data Decoding Department