CN101465974B - Method for implementing encode of nonlinear code - Google Patents

Abstract

The invention discloses a realizing method of an encoder for a nonlinear code. The method is that: the correlation additive value is determined by utilizing a tree-shaped adding structure; the exhaustive comparison is carried out on the correlation results by utilizing a tree-shaped comparing structure; then, a soft-decision output of each bit is determined. Therefore, the number of adders and comparators are greatly reduced. When calculating the correlation of an entire code word, the correlation is divided into the correlation of an info-bit part and the correlation of a check-bit part. Similar to the correlation of the info-bit part, the correlation of the check-bit part also uses the tree-shaped adding structure; only a simple reordering operation is added. When calculating the 256 correlation values of the entire code word, the negative symmetry property of a NR code table is utilized, thus leading the calculating amount of the addition to be reduced by one time. On the basis of a correlative MAP decoding algorithm, the realizing method can greatly reduce the addition and operation. The realizing method can be used for a digital television receiving chip according with the terrestrial broadcasting and transmitting of Chinese digital televisions, or other receiving chips of wireless transmitting systems which use NR codes.

Description

The implementation method of encode of nonlinear code

Technical field

The present invention relates to the Digital Television Terrestrial Broadcasting transmission field, the implementation method that particularly relates to a kind of NR (Nordstrom-Robinson nonlinear code) code decoder, can be used to meet the Digital Television receiving chip of China Digital TV terrestrial broadcasting transmission, perhaps other use the receiving chip of the wireless transmitting system of NR sign indicating number.

Background technology

The NR sign indicating number is a kind of nonlinear code of highly effective, it constitute multiple mode.

In the Digital Television Terrestrial Broadcasting transmission CNS a kind of 4QAM-NR mapping mode being arranged, is to increase the accurate orthogonal coding mapping of NR before the 4QAM sign map.

The accurate orthogonal coding of NR is that 8 information bit: X0X1X2X3X4X5X6X7 (X0 is preceding) are mapped as 16 bit: X0X1X2X3X4X5X6X7Y0Y1Y2Y3Y4Y5Y6Y7.

Wherein, X0X1X2X3X4X5X6X7 is an information bit, and Y0Y1Y2Y3Y4Y5Y6Y7 is the bit of deriving, and value is 0 or 1, and its restriction relation satisfies following formula:

Y0＝X7+X6+X0+X1+X3+(X0+X4)(X1+X2+X3+X5)+(X1+X2)(X3+X5)；

Y1＝X7+X0+X1+X2+X4+(X1+X5)(X2+X3+X4+X6)+(X2+X3)(X4+X6)；

Y2＝X7+X1+X2+X3+X5+(X2+X6)(X3+X4+X5+X0)+(X3+X4)(X5+X0)；

Y3＝X7+X2+X3+X4+X6+(X3+X0)(X4+X5+X6+X1)+(X4+X5)(X6+X1)；

Y4＝X7+X3+X4+X5+X0+(X4+X1)(X5+X6+X0+X2)+(X5+X6)(X0+X2)；

Y5＝X7+X4+X5+X6+X1+(X5+X2)(X6+X0+X1+X3)+(X6+X0)(X1+X3)；

Y6＝X7+X5+X6+X0+X2+(X6+X3)(X0+X1+X2+X4)+(X0+X1)(X2+X4)；

Y7＝X0+X1+X2+X3+X4+X5+X6+X7+Y0+Y1+Y2+Y3+Y4+Y5+Y6。

In the following formula, addition is that mould two adds computing, and multiplication is mould two multiplications.

NR sign indicating number (i.e. " the accurate orthogonal coding of NR ", unified in below describing with " NR sign indicating number " expression " the accurate orthogonal coding of NR ") have correlation properties preferably, each code word in the code table is carried out relevant with 256 code words, in 256 relevant outputs that produce 1 maximal correlation output 16 is arranged, 1 the relevant output-16 of minimum, have 30 relevant to be output as 0, other correlations are+4 or-4.

The decoding algorithm of NR sign indicating number has a lot, and the relevant MAP (relevant maximum a posteriori probability) that has commonly used deciphers and LOG-MAP (logarithm maximum a posteriori probability) decoding algorithm.For fear of the decoding of the complexity in finite field, can adopt relatively simply relevant MAP decoding algorithm.With the codeword information that receives and 256 local code words related operation that walks abreast, the corresponding Hard decision decoding result of maximum positive output.

Because TPC (Turbo product code), CTC (Turbo convolution code), LDPC very advanced codings such as (loe-density parity-check codes) are deciphered effect preferably in order to reach, all be to use the iterative decoding algorithm of soft inputting and soft output (SISO); Therefore for convenience of the SISO iterative decoding of rear end, 256 related operation results are divided into groups, seek the competition code word correlated results of every information bit correspondence one by one, thereby can obtain the Soft decision decoding output result of each information bit.

Adopt above-mentioned algorithm, even use relatively simply relevant MAP decoding algorithm, owing to the codeword information that receives and whole 256 code words in the code table need be carried out relevantly, and compare correlated results one by one, amount of calculation is also quite big.

Summary of the invention

The technical problem to be solved in the present invention provides a kind of implementation method of encode of nonlinear code, on the basis of relevant MAP decoding algorithm, can reduce the number of addition and operation greatly.

For solving the problems of the technologies described above, encode of nonlinear code implementation method of the present invention comprises the steps:

The first step receives the soft information of code word, utilizes the tree-shaped structure that adds to make it relevant with whole 256 code words in the NR code table, obtains correlated results;

Second step, utilize tree-shaped comparative structure that correlated results is carried out limit relatively, obtain the soft-decision output of each bit.

Owing to adopt said method, can significantly reduce the number of adder and comparator.The unreduced decoder that adopts algorithm of the prior art to realize, whenever translating a code word needs 15*256=3840 sub-addition and 2*127*8=2032 comparison; And adopt decoder after the simplification that method of the present invention realizes, whenever translating a code word only needs 254*2+128=636 sub-addition and 254+510=764 comparison.

Description of drawings

The present invention is further detailed explanation below in conjunction with accompanying drawing and embodiment:

Fig. 1 is the flow chart of the method for the invention;

Fig. 2 is the tree-shaped structural representation that adds that uses in the method for the present invention;

Fig. 3 is the tree-shaped comparative structure schematic diagram that uses in the method for the present invention.

Embodiment

As shown in Figure 1, NR code decoder implementation method of the present invention mainly was divided into for two steps, and the first step is for receiving the soft information of code word, and made it relevant with all code words in the code table, and second goes on foot to correlated results is compared, and obtains soft value.

Doing when relevant since each bit of code word all be+1 or-1, so in associative operation, do not need multiplication, only need addition.

Make (a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7) the soft information of code word for receiving, wherein (a0, a1, a2, a3, a4, a5, a6 a7) is the soft information of information bit, (b0, b1, b2, b3, b4, b5, b6 b7) is the soft information of check bit.The relevant of whole codeword can be divided into the relevant of the relevant of information bit part and check bit part, can be formulated as:

Correlation_codeword＝correlation_information_bits+correlation_check_bits

Wherein, Correlation_codeword is that all bits of whole codeword are relevant,

Correlation_information_bits is that information bit is relevant in the code word,

Correlation_check_bits is that check bit is relevant in the code word.

Can prove that 256 code words in the NR code table are negative symmetries, therefore only need to calculate the relevant of preceding 128 code words in the soft information of code word that receives and the NR code table, the value that back 128 code words are correlated with can be expressed as:

Correlation_second_half＝-Correlation_first_half(127:-1:0)

Wherein, Correlation_second_half is that back 128 code words are relevant,

Correlation_first_half is that preceding 128 code words are relevant.

Mark X (m:s:n) expression data variable X sequential element call number is from m to n, and step-length is s.

When the relevant of the information bit part of doing preceding 128 code words and check bit part relevant, the tree-shaped structure that adds among use Fig. 2 can the many add operations of merger, significantly reduce the number of times of add operation.

The relevant of information bit part of preceding 128 code words can directly utilize the tree-shaped structure that adds to obtain.

Being correlated with of the check bit part of preceding 128 code words, computational process is as follows:

(1), utilize the tree-shaped structure that adds to obtain add_first_b (0:1:127);

(2), make add_second_b=-add_first_b (127:-1:0);

(3), add_first_b and add_second_b are formed a column vector add_b;

(4), from add_b, take out 128 values according to the sequence number that reorders in the following table 1, as the check bit correlated results partly of preceding 128 code words.

Wherein, add_first_b is be correlated with a last half value before reordering of check bit,

Add_second_b is be correlated with a back half value before reordering of check bit,

Add_b is be correlated with a all values before reordering of check bit.

Table 1 sequence number that reorders

0，254，151，105，47，209，204，50，93，163，240，14，154，100，67，189，185，71，90，164，226，28，117，139，54，200，239，17，133，123，40，214，115，141，170，84，180，74，25，231，198，56，37，219，233，23，126，128，108，146，193，63，223，33，6，248，11，245，156，98，80，174，179，77，229，27，60，194，86，168，251，5，106，148，137，119，49，207，166，88，142，112，35，221，73，183，144，110，211，45，68，186，252，2，31，225，216，38，79，177，131，125，96，158，191，65，18，236，12，242，213，43，21，235，246，8，58，196，173，83，160，94，121，135，103，153，202，52

When the soft-decision output of calculating each bit, in the polarity of this bit correlated results is divided into two groups according to 256 code words, maximizing in two groups respectively, the difference between these two maximums just can be used as the soft-decision output of this bit.When each bit calculates soft information, all need 256 comparisons, the comparative result of considering the front can be used for the soft value of back and calculate, and therefore can utilize the tree-shaped comparative structure among Fig. 3, thereby significantly reduce the number of times of comparison.

The correlation of all 256 code words is placed on rightmost one row of tree-shaped comparative structure, just the 7th row.The a pair of consecutive value of the value of n-1 row by the n row relatively size obtained, and the number of the value of every like this row reduces by half successively, and the number of same comparator also can reduce by half, and is formulated as:

$a_0^{n-1}=\max (a_0^n,a_1^n)$

$a_1^{n-1}=\max (a_2^n,a_3^n)$

$a_{2^{n-1}-1}^{n-1}=\max (a_{2^n-2}^n,a_{2^n-1}^n)$

For example, the value of the 6th row can be obtained according to following formula by the value of the 7th row:

$a_0^6=\max (a_0^7,a_1^7)$

$a_1^6=\max (a_2^7,a_3^7)$

$a_{63}^6=\max (a_{126}^7,a_{127}^7)$

$a_{127}^6=\max (a_{254}^7,a_{255}^7)$

Can obtain the value of the 5th, 4,3,2,1,0 row then successively.(i.e. 8 information bits: soft-decision X0X1X2X3X4X5X6X7) is output as and n row meta is changed to maximum in 1 all values deducts n row meta and be changed to maximum in 0 all values corresponding bit Xn; Wherein, the n value is 0,1 ... 7 integer.

Claims (3)

1. the implementation method of an encode of nonlinear code is characterized in that:

The first step receives the soft information of code word, utilizes the tree-shaped structure that adds to make it relevant with whole 256 code words in the NR code table, obtains correlated results;

The soft information of described code word comprises the soft information of information bit and the soft information of check bit; Described code word is relevant to be divided into the relevant of the relevant of information bit part and check bit part, and only carries out preceding 128 code words and be correlated with, and the value representation that back 128 code words are relevant is:

Correlation_second_half＝-Correlation_first_half(127:-1:0)，

Wherein, Correlation_first_half is that preceding 128 code words are relevant,

Correlation_second_half is that back 128 code words are relevant;

Being correlated with of the check bit part of preceding 128 code words, computational process is as follows:

(1), utilize the tree-shaped structure that adds to obtain add_first b (0:1:127);

(2), make add_second_b=-add_first_b (127:-1:0);

(3), add_first_b and add_second_b are formed a column vector add_b;

(4), from add_b, take out 128 values according to the sequence number that reorders, as the check bit correlated results partly of preceding 128 code words;

Wherein: add_first_b is be correlated with a last half value before reordering of check bit, and add_second_b is be correlated with a back half value before reordering of check bit, and add_b is be correlated with a all values before reordering of check bit;

Mark X (m:s:n) expression data variable X sequential element call number is from m to n, and step-length is s;

Second step, utilize tree-shaped comparative structure that correlated results is carried out limit relatively, obtain the soft-decision output of each bit; When the soft-decision output of calculating each bit, in the polarity of this bit correlated results is divided into two groups according to 256 code words in the NR code table, maximizing in two groups respectively is with the soft-decision output of the difference between two maximums as this bit.

2. the implementation method of encode of nonlinear code as claimed in claim 1 is characterized in that: the relevant of information bit part of preceding 128 code words directly utilizes the tree-shaped structure that adds to obtain.

3. the implementation method of encode of nonlinear code as claimed in claim 1 is characterized in that: a pair of consecutive value of the value of n-1 row by the n row relatively size obtained;

The soft-decision of corresponding bit Xn is output as and n row meta is changed to maximum in 1 all values deducts n row meta and be changed to maximum in 0 all values; Wherein, the n value is 0,1 ... 7 integer.

Images