JP2003249859A - Encoding method and its device, and decoding method and its device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem of strengthen the noise resistance of encoded data and of reducing bit error rate, when it is decoded. <P>SOLUTION: A GS-APP decoding portion 52 of a GS-decoding portion 50 receives apriori probability related to an information bit u, a parity bit p1 and a parity bit p2 sent from a turbo-decoding portion 60, calculates outside informations and sends them to the turbo-decoding portion 60, when it does a soft decision decoding of a receiving bit y. The outside informations are divided into the information bit u and the first parity bit p1 which are necessary for decoding at a first APP decoding portion 62 and into the second parity bit p2 necessary for decoding at a second APP decoding portion 64 by a demultiplexer 102. A normal turbodecoding is conducted at the turbo-decoding portion 60, and the outside informations calculated by the second APP decoding portion 64 is sent to the GS-decoding portion 50. By cooperation between the GS decoding portion 50 and the turbo-decoding portion 60, the reliability is increased and the bit error rate of the information bit u is decreased. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a coding method and apparatus, and a decoding method and apparatus, and more particularly to error correction coding and decoding technology.

[0002]

2. Description of the Related Art For receiving digital information on a transmission line,
When reproducing digital information from a magnetic recording medium, an error may occur in the bit sequence of the digital information. An error correction coding technique is one of the techniques for preventing such errors in digital information. The error correction coding is a technique that enables digital information to be transmitted or recorded with redundancy to be coded so that correct information can be restored even if an error occurs in the digital information.

As an error correction coding technique, an error correction code such as a block code or a convolutional code has been conventionally used. Recently, a turbo code which is said to have a transmission characteristic close to the Shannon limit has been attracting attention. Came. Turbo codes are detailed in the following two documents. Branka Vucetic and
Jinhong Yuan, “Turbo codes: principles and appli
cations ”, Kluwer Academic Publishers (2000). Zin
ing Wu, “Coding and iterative detection for magnet
ic recording channels ”, Kluwer AcademicPublishers
(1999).

In turbo coding, two convolutional encoders are used on the encoding side when encoding an information sequence, and a second convolutional encoder is used.
An information sequence whose data order is mixed by an interleaver is given to the encoder of (1) and an encoding result different from that of the first encoder is output. On the decoding side, two decoders corresponding to these two encoders are used, and the soft-value output of one decoder is fed back to the other decoder as information on the prior probability, thereby iteratively decoding the encoded sequence. I do. The reliability information about the information bits calculated from the parity bits is used as the soft value output given to the other decoder described above. This reliability information is called external information because it does not use any information about the information bit itself and has nothing to do with the information bit itself. This extrinsic information is transmitted to the other decoder as the a priori probability of the information bit sequence, and posterior probability (APP) decoding is performed using the a priori probability and the information bit sequence. It is possible to improve the reliability of the decoding result and reduce the bit error rate by the two decoders cooperating and repeatedly performing this decoding process.

[0005]

However, even if such a decoding method having a high error correction capability is used at the time of decoding, if the coded sequence has low tolerance to the noise on the transmission line in the first place, it is received. Sometimes it contains many errors, and the bit error rate cannot be reduced only by improving the error correction capability at the time of decoding. Therefore, a combination is required in which the tolerance of the encoded sequence is sufficiently increased at the encoding stage and the detection rate is improved by iterative decoding at the decoding stage.

In particular, when a digital watermark is embedded in content data such as images and sounds, the content data is
In the distribution process and usage process, signal processing such as compression encoding and various filtering is added, processed by the user, watermark information is tampered with, etc.
Receive various operations. Some of the embedded digital watermark data may be changed or lost by these operations, and the digital watermark requires resistance not only to noise on the transmission path but also to such content data operations. To be done. Therefore, it is required to encode highly resistant watermark data and embed the watermark data in the content data to improve the detection accuracy of the watermark data when extracting the watermark.

The present invention has been made in view of such circumstances, and an object thereof is to provide an encoding technique capable of generating an encoded sequence suitable for an error characteristic and a decoding technique having a high error correction capability. is there. Another object is to provide a technique capable of reducing a detection error of a digital watermark at the time of decoding by encoding and embedding a digital watermark having strong resistance.

[0008]

One aspect of the present invention relates to an encoding method. This method scrambles a turbo-encoded information sequence by a convolutional operation to generate a plurality of scramble sequences, evaluates the suitability of each of these scramble sequences, and finally determines the scramble sequence with a good evaluation. It is acquired as an encoded sequence. For example, when the encoded sequence is transmitted and received on the transmission line,
For each scrambled sequence, whether it is suitable for the conditions and performance of the transmission line, whether it is suitable for the purpose of transmission, whether it is immune to the noise of the transmission line, the suitability for transmission characteristics, and the transmission application Aptitude for
The suitability for noise resistance is evaluated. When the coded sequence is recorded on a magnetic recording medium or the like and reproduced, suitability for conditions and performance of the reproducing system, reproducing application, error characteristics, etc. is evaluated.

Another aspect of the present invention relates to a decoding method. In this method, extrinsic information output from a soft-output decoder that performs soft-decision decoding of a received sequence is given to a turbo decoder as input data, and extrinsic information output from the turbo decoder is prior probability in the soft-decision decoding. Iterative decoding of the received sequence is performed by feeding back to the soft output decoder as information on the above.

Yet another aspect of the present invention relates to an encoding device. This apparatus includes a turbo coding unit that turbo-codes an information sequence to generate a turbo code sequence, an interleaver that interleaves the turbo code sequence, and a plurality of scrambled scrambled convolution operations of the interleaved turbo code sequence. A scramble unit that generates a scramble code sequence, an evaluation unit that evaluates suitability for each of the plurality of scramble code sequences, and selects and outputs one of the plurality of scramble code sequences based on the evaluation value of the suitability. And a selection unit for Since a plurality of scramble sequences are generated from the turbo-encoded information sequence, the turbo encoding process only needs to be performed once, and thus the encoding device of this aspect can be realized with a simple configuration.

Yet another aspect of the present invention relates to a decoding device. This apparatus includes a soft-decision decoding unit that performs soft-decision decoding on a reception sequence, a deinterleaver that deinterleaves the reception sequence that has been soft-decision decoded, and a turbo that performs turbo decoding using the deinterleaved reception sequence as input data. And a decoding unit. When the received sequence is scrambled by a convolution operation, the soft decision decoding unit may descramble the received sequence by performing a convolution operation. The soft decision decoding unit may perform APP (A Posteriori Probability) decoding as the soft decision decoding. APP decoding is performed by, for example, MAP (Maximum A
posteriori Probability) Decoding method and Soft Output Viterbi Algorithm.

The received information may be iteratively decoded by feeding back the external information output from the turbo decoding unit to the soft decision decoding unit as information about the prior probability of the received sequence. External information about the information bits of the received sequence output by the turbo decoding unit may be fed back to the soft-decision decoding unit as information about the a priori probability of the information bits. The turbo decoding unit may perform soft-decision decoding of the parity bit of the reception sequence, and the external information about the parity bit of the information sequence output by the turbo decoding unit is the soft information as the information about the prior probability of the parity bit. It may be fed back to the determination decoding unit. External information (extrinsic info
rmation) is a log-likelihood ratio (log-
It is a value obtained by subtracting a value called a channel value to which the information bit to be decoded contributes from the likelihood ratio) and a value related to the a priori probability.

Yet another aspect of the present invention also relates to an encoding device. This apparatus includes a scramble unit that scrambles an information sequence by a convolution operation to generate a plurality of scramble code sequences, an interleaver that interleaves each of the plurality of scramble sequences, and each of the interleaved plurality of scramble code sequences. A turbo coding unit that turbo-encodes a plurality of turbo code sequences, an evaluation unit that evaluates suitability for each of the plurality of turbo code sequences, and a plurality of turbo codes based on the suitability evaluation value. And a selector for selecting and outputting one of the sequences.

Yet another aspect of the present invention also relates to a decoding device. This apparatus includes a turbo decoding unit that performs turbo decoding using a reception sequence as input data, a deinterleaver that deinterleaves the turbo-decoded reception sequence, and a soft-decision decoding that performs soft-decision decoding of the deinterleaved reception sequence. And part. Iterative decoding of the received sequence may be performed by feeding back external information about the information bits of the received sequence output by the soft-decision decoding unit to the turbo decoding unit as information about the prior probability of the information bits.

Yet another aspect of the present invention also relates to an encoding device. This device is a turbo coding unit for turbo coding digital watermark data to be embedded in host data,
An interleaver that interleaves the turbo-encoded digital watermark data, a scramble unit that scrambles the interleaved digital watermark data by a convolution operation to generate a plurality of watermark data candidates, and a plurality of the watermark data candidates Embedded in the host data to generate a plurality of embedded host data candidates, an evaluation unit that evaluates the resistance of the digital watermark for each of the plurality of embedded host data candidates, and an evaluation of the resistance. A selection unit for selecting and outputting one of the plurality of embedded host data candidates based on a value.

Yet another aspect of the present invention also relates to an encoding device. This apparatus includes a scramble unit that scrambles digital watermark data to be embedded in host data by a convolution operation to generate a plurality of watermark data candidates, an interleaver that interleaves each of the plurality of watermark data candidates, and A turbo encoder that turbo-encodes each of a plurality of interleaved watermark data candidates, and embeds each of the turbo-encoded watermark data candidates in the host data.
An embedding unit that generates a plurality of embedded host data candidates, an evaluation unit that evaluates the resistance of the digital watermark for each of the plurality of embedded host data candidates, and the plurality of embeddings based on the resistance evaluation value. And a selector for selecting and outputting one of the host data candidates.

The host data is original data into which a digital watermark is to be embedded and is, for example, data such as a still image, a moving image, a voice, or the like. The digital watermark to be embedded includes identification information of original data, creator information, user information, and the like. In addition, for the purpose of authentication, it is possible to embed digest data of host data, that is, data that briefly represents the characteristics of host data as a digital watermark. Tolerance of digital watermark means, for example, when the host data embedded with the digital watermark is attacked, or when the embedded host data is subjected to signal processing such as compression encoding or filtering.
The robustness of the digital watermark data when some operation is applied to the embedded host data.

According to the encoding device of these aspects, it is possible to embed the watermark data after converting the watermark data into a highly resistant data sequence according to the host data, and reduce the detection error of the digital watermark.

Still another aspect of the present invention also relates to a decoding device. This apparatus includes an extraction unit that extracts watermark data from host data in which a digital watermark is embedded, a soft-decision decoding unit that performs soft-decision decoding of the extracted watermark data, and deinterleave the watermark data that has been soft-decision decoded. And a turbo decoding unit that performs turbo decoding using the deinterleaved watermark data as input data. Iterative decoding of the watermark data may be performed by feeding back external information output by the turbo decoding unit to the soft decision decoding unit as information regarding the prior probability of the watermark data.

Yet another aspect of the present invention also relates to a decoding device. This apparatus includes an extracting unit that extracts watermark data from host data in which a digital watermark is embedded, a turbo decoding unit that performs turbo decoding using the extracted watermark data as input data, and an inverse unit that reverses the turbo decoded watermark data. It includes a deinterleaver for interleaving, and a soft decision decoding unit for performing soft decision decoding on the deinterleaved watermark data. Iterative decoding of the watermark data may be performed by feeding back external information output by the soft-decision decoding unit to the turbo decoding unit as information about the prior probability of the watermark data.

In the decoding device of these aspects, when the watermark data is scrambled, the descrambling of the watermark data is performed by performing a convolution operation on the watermark data finally obtained by the iterative decoding. It may further include a scramble part.

On the side of embedding a digital watermark, when scrambling the digital watermark data, a one-to-many mapping is used to associate the original digital watermark data with a plurality of candidates for the watermark data. On the side of extracting the digital watermark, reverse mapping is performed to obtain the original digital watermark data from the scrambled watermark data. Therefore, on the side that extracts the digital watermark, the correspondence table of the original digital watermark data and the plurality of watermark data candidates may be used. Further, on the side of embedding the digital watermark, a scramble function for generating a plurality of watermark data candidates from the original digital watermark data based on a predetermined initial value may be used. In this case, on the side of extracting the digital watermark, descrambling of the extracted digital watermark is performed based on the initial value used for scrambling and the scramble function.

It should be noted that any combination of the above components, the expression of the present invention can be applied to a method, an apparatus, a system, a recording medium,
A program converted between computer programs and the like is also effective as an aspect of the present invention.

[0024]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiment 1 In the embodiment, a "turbo-encoded GS system" in which a turbo code and a scramble encoding by a GS (Guided Scramble) system are concatenated is used as a channel coding system. First, a general configuration of a turbo code will be described as a premise of the embodiment, and then the configuration of the embodiment will be described.

FIG. 1 shows a general configuration of the turbo encoder 70. The user bit string UB input as an encoding target is input to the first encoder 72, and the parity bit string P ₀ is generated. The order of the user bit string UB is rearranged by the interleaver 74 and is input to the second encoder 76, and the parity bit string P ₁ is generated.
Two parity bit strings P ₀ and P ₁ are multiplexers
The puncturing 78 multiplexes while puncturing, and further multiplexes with the user bit string UB and outputs the coded bits CB from the multiplexer 80. The puncture process is an optional process used when it is desired to increase the coding rate, and may be omitted if not necessary.

FIG. 2 shows a general structure of the turbo decoding unit 90. The received channel output CO is distributed by the demultiplexer 102 to the information bit string IB and the two parity bit strings P ₀ and P ₁ to obtain the information bit string IB and the first bit string IB.
Parity bit stream P ₀ of the input to the first soft decoder 92 corresponding to the first encoder 72 of FIG. 1, the order Sorting altered information bit sequence IB and second by the interleaver 96 parity bit stream P ₁ Is the second encoder 76 of FIG.
Is input to the second soft decoder 98 corresponding to. The first soft decoder 92 outputs external information that gives information on the reliability of the decoding result of the information bit string IB. Second soft decoder 98
Uses the extrinsic information obtained from the first soft decoder 92 via the interleaver 94 as the prior probability to perform the decoding process of the information bit string IB whose order is rearranged by the interleaver 96, and the extrinsic information for the decoding result To the first soft decoder 92 via the deinterleaver 100. First
The soft decoder 92 uses the extrinsic information from the second soft decoder 98 as a priori probability to perform the decoding process of the information bit string IB. By repeating this series of operations, the turbo decoding unit 90 outputs the final determination FD. This final judgment FD
Is the posterior probability of the information bit string IB, not the external information, and the hard decision using this posterior probability provides the decoding result of the hard value of the information bit string IB. Interleaver 9
The sorting by 4, 96 is performed by the interleaver 74 of FIG.
Is the same as sorting. Also deinterleaver 100
Performs a process of restoring the rearrangement of the interleavers 94 and 96.

The turbo decoding unit 90 includes a first soft decoder 92.
The second soft decoder 98 and the second soft decoder 98 can sequentially improve the decoding result of the APP decoding by utilizing the prior information provided from the other one. This makes it possible to further reduce the bit error rate (BER) when detecting watermark bits.

FIG. 3 shows an encoding device 1 according to the first embodiment.
The structure of 0 is shown. In terms of hardware, this configuration can be realized by a CPU, a memory, and other LSIs of an arbitrary computer, and in terms of software, it is realized by a program having a scramble encoding function loaded in the memory. Then, the functional blocks realized by those collaborations are drawn. Therefore, it will be understood by those skilled in the art that these functional blocks can be realized in various forms by only hardware, only software, or a combination thereof.

The coding device 10 receives the input information sequence u.
Is encoded into a channel sequence x suitable for the communication path. The turbo encoding unit 70 performs turbo encoding using the information sequence u as input data and outputs a turbo encoded information sequence u _c . The GS encoding unit 200 scrambles the turbo encoded information sequence u _c by the GS method. The turbo-coded information sequence u _c is interleaved by the interleaver π ₂ 32 and then scrambled by the GS encoder 200. This interleaving process is performed in order to eliminate mutual interference between the turbo coded sequence generated by the turbo coding unit 70 and the scramble coded sequence generated by the GS coding unit 200.

Each of the L multiplexers 20 has an initial data C at the beginning of the interleaved information sequence u _c.
L- type bit sequence u _b with _{0 to} C _L-1 inserted is generated. The L scramblers 22 scramble the L types of bit sequences respectively to generate L types of scrambled information sequences u ′ _b .

The channel restrictions comparator unit 34, for each of the L types of scrambled information sequence u _'b, to assess suitability for channel. The selector 30 selects the information sequence u ′ _b having the best evaluation value of suitability and outputs it as the final channel sequence x.

The GS system is a scramble system used in transmission and digital modulation in magnetic recording. G
In the S method, L types of code sequences are generated for an information sequence having a certain data block length, and these are treated as candidates for a code sequence to be transmitted next. From these candidates, the most suitable one is selected according to the property of the transmission medium, and the final code sequence is selected. With this GS method, a variety of code sequence candidates can be generated by a simple method.

As a typical GS system, a GS system based on convolution calculation will be introduced. The GS encoder adds L kinds of r-bit redundant words c _i (i = 0, ..., L−1) immediately before the n-bit information sequence D (x), and L
Generate a code sequence c _i x ⁿ + D (x) of a type. The code length of this code sequence is (n + r) bits. Thus, the quotient T _i (x) is obtained by dividing the code sequence to which the redundant word is added by the N-dimensional scramble polynomial S (x) as in the following equation.

T _i (x) = Q _{S (x)} [(c _i x ⁿ + D (x)) x ^N ] (1) where Q _a [b] represents the quotient of b divided by a. The quotient set {T ₀ (x), ..., T _L-1 (x)} is a candidate for the code sequence after scramble. For each of these candidates, the performance when the code sequence is actually used is evaluated, and the one with the best evaluation value is selected as the final code sequence.

At the time of demodulation, the original information sequence D (x) is obtained by multiplying the code sequence by S (x) and discarding the conversion information of the lower N bits and the upper r bits.

Here, a case where S (x) = x ^r +1 is used as the scramble polynomial S (x) will be described. n
When mod r = 0, the equation (1) can be expressed by the convolution operation shown in the following equation.

T _j = d _j (+) c _i (j = 0) t _j = d _j (+) t _j-1 (j = 1, ..., N / r)
−1) However, i = 0, ..., L−1, d _j is a bit string obtained by dividing the original information sequence D (x) by r bits, and t _j
Is a bit string in which the r-bit redundant word c _{i and} the beginning of the converted code sequence T _i (x) are separated by r bits. Further, (+) indicates an exclusive OR (EX-OR) operation.

FIG. 4 is a diagram for explaining the convolution operation during this encoding. For example, consider the case where n = 6 and r = 2. Original information sequence D (x) = (1,0,1,0,0,
A redundant word c ₀ = (0,0) is added to ₁ ) to generate a code sequence T ₀ (x) after conversion. By the convolution operation at the time of encoding, t ₀ = d ₀ (+) c ₀ = (1,
0) (+) (0,0) = (1,0), t ₁ = d ₁ (+)
t ₀ = (1,0) (+) (1,0) = (0,0), t ₂
= D ₂ (+) t ₁ = (0,1) (+) (0,0) =
(0,1), and the code sequence after conversion T ₀ = (0,0,
1,0,0,0,0,1) is obtained. Note that the first 2 bits of the code sequence T ₀ after conversion are the redundant word c ₀ .

Similarly, the redundant word c ₁ = (0,1), c
_{For 2} = (1,0) and c ₃ = (1,1), the code sequence after conversion T ₁ = (0,1,1,1,0,1,1)
0,0), T ₂ = (1,0,0,0,1,0,1,
1), T ₃ = (1,1,0,1,1,1,1,0) is obtained.

At the time of decoding, the original information sequence D (x) is obtained by performing a convolution operation as in the following equation.

D _j = t _j (+) c _i (j = 0) d _j = t _j (+) t _j-1 (j = 1, ..., N / r)
-1)

FIG. 5 illustrates the convolution operation during this decoding.
It is a figure that shows. In the above example, the encoded sequence T after conversion
₀= (0,0,1,0,0,0,0,1) is given
And the redundant word c from the first 2 bits₀= (0,0) is obtained
By the convolution operation at the time of decoding, d₀= T
₀(+) C₀= (1,0) (+) (0,0) = (1,
0), d₁= T₁(+) T₀= (0,0) (+) (1,
0) = (1,0), d_Two= T _Two(+) T₁= (0,1)
(+) (0,0) = (0,1), and the original information sequence D
(X) = (1,0,1,0,0,1) is obtained. other
Coded sequence T after conversion ₁, T_Two, T_ThreeAbout this tatami
The original information series D (x) is obtained by the convolution operation.

FIG. 6 is a decoding device 40 according to the first embodiment.
Shows the configuration of. In the decoding device 40, the GS decoding unit 50 and the turbo decoding unit 60 are connected in this order to form a feedback system. GS-APP decoding unit 52 of GS decoding unit 50
Performs soft decision decoding on the received bit y by the MAP decoding method or the like. The GS-APP decoding unit 52 has the prior probability of the information bit u fed back from the turbo decoding unit 60 and the prior probability Λ _a (x) of the first and second parity bits p ₁ and p ₂ (where x = {u , P ₁ , p ₂ }) in the state in which the order is rearranged by the interleaver π ₂ 56, and the external information Λ _e (x) is calculated from the received prior probability and the received bit y. External information Λ
_The deinterleaver π ₂ ^-1 54 deinterleaves _e (x) to restore the order, and the _e (x) is sent to the turbo decoding unit 60. The deinterleaved data of the deinterleaved data of the information bit u and the first parity bit p ₁ required for decoding by the first APP decoding unit 62 is sent to the first APP decoding unit 62 and the second APP decoding unit 64 of the deinterleaved data.
Information bit u and second parity bit p required for decoding in
_The data regarding ₂ is sent to the second APP decoding unit 64 and used as a reception sequence of the turbo decoding unit 60. However, the information bits u sent to the second APP decoding unit 64 are sent after being rearranged in order by the interleaver π ₁ 66b.

In the turbo decoding unit 60, turbo decoding is performed as follows. The first APP decoding unit 62 uses the GS
The posterior probability of the information bit u is calculated based on the information bit u ^ ₁ and the first parity bit p ^ ₁ provided from the decoding unit 50, and the external information Λ _e1 (u) is obtained from the posterior probability,
The second APP decoding unit 6 via the interleaver π ₁ 66a
Send to 4. The second APP decoding unit 64 uses the external information Λ
_{Using e1} (u) as the prior probability Λ _a2 (u) of the information bit u, the information bit u provided from the GS decoding unit 50 is used.
Information bit u based on ^ ₂ and the second parity bit p ^ _2.
Calculates the posterior probability of the external information Λ _e2 from the posterior probability
(U) is obtained and sent to the first APP decoding unit 62 via the deinterleaver π ₁ ^-1 68. First APP decoding unit 62
Uses the extrinsic information Λ _e2 (u) as the prior probability Λ _a1 (u) of the information bit u in the next decoding.

Further, the external information Λ _e2 of the information bit u calculated by the second APP decoding unit 64 of the turbo decoding unit 60.
(U) is the GS decoding unit 5 via the interleaver π ₂ 56.
0 is fed back to the GS-APP decoding unit 52.
However, in this case, since there is no extrinsic information regarding the first parity bit p ₁ and the second parity bit p ₂ , these values are zero, that is, the prior probability of being bit 0 or bit 1 is 1/2. , GS decoding section 50
Entered in.

By repeating the above process, the external information output from the GS decoding unit 50 is given as an input to the turbo decoding unit 60, and the external information with increased reliability from the turbo decoding unit 60 is further input. And the GS decoding is further performed, the reliability of the external information is increased, and the posterior probability of the information bit u is increased. The second APP decoding unit 64 of the turbo decoding unit 60 outputs the posterior probability of the finally obtained information bit u as the final determination FD.

The turbo decoding unit 60 performs not only soft-decision decoding of the information bit u but also soft-decision decoding of the first and second parity bits p ₁ and p ₂ respectively as a first APP decoding unit 62 and a second APP decoding unit. 64, and the resulting extrinsic information Λ of the first and second parity bits p ₁ and p _2.
(P ₁ ), Λ (p ₂ ) together with the external information Λ (u) of the information bits via the multiplexer 104, the GS decoding unit 50.
It is also possible to send to. This can further reduce the bit error rate.

According to the present embodiment, it is possible to enhance the noise resistance of the coded sequence by combining turbo coding with the GS system. At the time of decoding, the GS decoding unit 50 and the turbo decoding unit 60 are connected in cascade to form a feedback loop, so that reliability information regarding the information sequence is repeatedly transmitted and received between the GS decoding unit 50 and the turbo decoding unit 60. , Can be decoded, and the bit error rate can be reduced.

Embodiment 2 FIG. 7 shows the configuration of encoding apparatus 10 according to Embodiment 2. This embodiment is different from Embodiment 1 shown in FIG. 3 in that turbo coding is performed after scrambling by the GS method.

The L multiplexers 20 have information sequences u.
Respectively to generate the initial data _C 0 _{-C L-1} the inserted L type bit sequence _{u b} at the beginning of. The L scramblers 22 scramble the L types of bit sequences respectively to generate L types of scrambled information sequences u ′ _b .

L kinds of scrambled information sequence u '
_bIs the interleaver π_TwoInterly by 32
After being stored, the L turbo coding units 70
Each is turbo encoded. The turbo coding unit 70
Rumbled information sequence u ' _bAs input data
Turbo-encoded information sequence u '_cTo
Output.

[0052] channel restrictions comparator unit 34, for each of the L types of turbo encoded information sequence u _'c, to evaluate the suitability for channel. The selector 30 selects the information sequence u ′ _c having the best evaluation value of suitability and outputs it as the final channel sequence x.

FIG. 8 shows a decoding device 40 according to the second embodiment.
Shows the configuration of. The decoding device 40 includes a turbo decoding unit 60 and a G
The S decoding unit 50 is connected in this order to form a feedback system. The received bit y is the information bit u necessary for decoding by the first APP decoding unit 62 by the demultiplexer 102 and the data related to the first parity bit p ₁ is the first bit A.
Data regarding the second parity bit p ₂ required for decoding by the second APP decoding unit 64 is sent to the PP decoding unit 62 as the _second A data.
It is sent to the PP decoding unit 64 and used as a reception sequence of the turbo decoding unit 60. However, the information bits u sent to the second APP decoding unit 64 are sent after being rearranged in order by the interleaver π ₁ 66b.

In the turbo decoding unit 60, turbo decoding is performed as follows. The first APP decoding unit 62 calculates the posterior probability of the information bit u based on the received information bit u ^ ₁ and the first parity bit p ^ ₁ and obtains the external information Λ _e1 (u) from the posterior probability. , Interleaver π ₁
It is sent to the second APP decoding unit 64 via 66a. Second AP
The P decoding unit 64 uses the extrinsic information Λ _e1 (u) as the prior probability Λ _a2 (u) of the information bit u, and based on the received information bit u ^ ₂ and the second parity bit p ^ _2. The posterior probability of the information bit u is calculated, the external information Λ _e2 (u) is obtained from the posterior probability, and the deinterleaver π ₁ ⁻¹
It sends to the 1st APP decoding part 62 via 68. First APP
The decoding unit 62 uses the external information Λ _e2 in the next decoding.
Use (u) as the prior probability Λ _a1 (u) of the information bit u.

Further calculated by the second APP decoding unit 64.
External information Λ of the information bit u_e2(U) is the deinter
Lever π₁ ^-168 and deinterleaver π_Two ^-15
4 to the GS decoding unit 50. GS-APP recovery
The issue section 52 uses this external information Λ _e2(U) to the GS decoding unit 5
It is used as a reception sequence of 0 and soft decision decoding is performed. However,
Regarding the prior probability input to the GS-APP decoding unit 52
The value is zero. Calculated by the GS-APP decoding unit 52
External information Λ_e(U) is the interleaver π _TwoTo 56
The first AP of the turbo decoding unit 60 is rearranged according to the order.
It is sent to the P decoding unit 62. The first APP decoding unit 62 includes
The switch causes the GS decoding unit 50 to output the external information Λ
_e(U) and the external information Λ output from the second APP decoding unit 64.
_e2Any one of (u) is switched and information bit u
Prior probability Λ_a1It is input as (u). The first
The APP decoding unit 62 outputs the data output from the GS decoding unit 50.
Department information Λ_eOnly (u) is the prior probability Λ of the information bit u._a1
When used as (u), the second APP decoding unit 64
External information Λ output from_e2(U) is unnecessary,
Interleaver π₁ ^-1Second APP decoding unit 6 via 68
4 to the first APP decoding unit 62 feedback loop
The configuration of is omitted.

By repeating the above processing, the external information output from the turbo decoding unit 60 is given as an input to the GS decoding unit 50, and the external information with increased reliability from the GS decoding unit 50 is further received. Feedback processing to further turbo decoding is possible, the reliability of external information is increased, and the information bit u
The posterior probability of increases. GS-APP of GS decoding unit 50
The decoding unit 52 outputs the posterior probability of the finally obtained information bits as the final determination FD.

Embodiment 3 FIG. 9 shows the configuration of encoding apparatus 10 according to Embodiment 3. The encoding device 10 adds the watermark information I to the host data V.
Is embedded, and embedded host data W is output. The host data V is data such as voice, still image, and moving image. The watermark information I is identification information of the host data V, information about copyright such as creator information, user information, authentication information for detecting falsification of host data, and time stamp.

The encryption unit 12 encrypts the watermark information I to be embedded in the host data V with the secret key K and outputs the watermark data X. If the encryption function is f ₀ , this process is represented by the conversion formula X = f ₀ (I, K). If the watermark information is not encrypted, the configuration of the encryption unit 12 may be omitted.

The changing unit 14 scrambles the watermark data X using the watermark data X and the host data V, and outputs the scrambled watermark data X '. Assuming that the scramble function is f ₂ , this process is performed by the conversion formula X ′ =
It is represented by f ₂ (X, V).

The embedding unit 16 embeds the scrambled watermark data X'into the host data V using the secret key K and outputs the embedded host data W. If the embedding function is f ₁ , this process is performed by the conversion formula W = f
It is represented by ₁ (V, X ', K). In the case of the embedding method that does not depend on the secret key K, W = f ₁ (V, X ′).

The changing unit 14 and the embedding unit 16 cooperate to
Generate a plurality of scrambled watermark data X ′,
Each has a function of embedding each in host data V, generating a plurality of candidates for embedded host data W, and selecting one of these candidates.

FIG. 10 is a functional block diagram of the changing unit 14 and the embedding unit 16. The turbo encoding unit 70 performs turbo encoding with the watermark data X as input data, and outputs the turbo encoded watermark data X _c . GS encoder 2
00 scrambles the turbo-encoded watermark data _Xc by the GS method and embeds it in the host data V. The turbo-encoded watermark data X _c is interleaved by the interleaver π ₂ 32 and then GS
Scramble coding is performed by the coding unit 200.

The L multiplexers 20 generate L kinds of bit sequences _{Xb in} which the initial data C _{0 to} C _L-1 are inserted at the beginning of the interleaved watermark data X _c . The L scramblers 22 scramble the L kinds of bit sequences, respectively, to generate the L kinds of scrambled watermark data X ′ _b .

The L embedding sections 26 embed each of the L types of scrambled watermark data X ′ _{b in} the host data V to generate L types of embedded host data W candidates. The L SNR calculators 28 evaluate the tolerance of the watermark data X for each of the L types of embedded host data W candidates. The selector 30 selects a candidate of the embedded host data W having the best tolerance evaluation value and outputs it as the final embedded host data W.

FIG. 13 is a flow chart for explaining a procedure for embedding a digital watermark by the encoding device 10 having the above configuration. In explaining the flowchart, FIGS. 11 and 12 are referred to as appropriate. The multiplexer 20 is turbo-encoded by the turbo encoding unit 70, and the interleaver π
_The L kinds of initial data are inserted at the head of the watermark data X _c interleaved by the H ₂ 32 to generate L pieces of code sequences (S10), and the scrambler 22 scrambles the code series to generate L kinds of code sequences. The scrambled watermark data _Xb 'is generated (S12).

FIG. 11 shows the relationship between the watermark data _Xc and the L types of scrambled watermark data _Xb '.
An r-bit redundant word is added as identification data ID [0] to ID [L-1] to the beginning of the n-bit watermark data X to create L types of watermark data candidates. Up to 2 ^r
A type candidate is created. The bit string of the watermark data _Xc included in these candidates is scrambled by the above-mentioned GS method.

The L kinds of scrambled watermark data _Xb 'generated by the scrambler 22 are embedded in the host data V by the embedding unit 26 (S1).
4).

12 (a) and 12 (b) show scrambling.
Watermark data X_b’In the figure explaining the embedding method
is there. L types of scrambled watermark data X_b’
X ⁰, X¹, ・ ・ ・, X^L-1And Each watermark day
The bit sequence of the data candidate is expressed by the following equation. lead
R bits are identification data. Also, scramble processing
Bit 0 after processing is replaced with -1 and the following processing is performed.
U

X ⁰ = {-1, ..., -1, -1, x ⁰
₀ , x ⁰ ₁ , ..., X ⁰ _n-1 } x ¹ = {-1, ..., -1, 1, x ¹ ₀ , x ¹ ₁ , ...
_{^{··, x 1 n-1}}} ··· x L-1 = {1, ···, 1,1, x L-1 0, x
^L-1 ₁ , ..., X ^{L- 1} _n-1 }

A pair (V ⁺ , V ⁻ ) of a set of samples of the host data V selected as a target for embedding n-bit watermark data is defined as follows. Each of the sample sets V ⁺ and V ⁻ has n elements as follows. The host data V is represented by samples on the space axis, samples on the time axis, samples on the frequency axis, for example, samples after processing such as DCT conversion, FFT conversion, and DWT conversion.

[0071] ^{_{V + = {v + 0,}} v + 1, ···, v + n-1} V - = {v - 0, v - 1, ···, v - n-1} where the sample Each of the subsets v ⁺ _i , v ⁻ _i which are elements of the set V ⁺ , V ⁻ of
It consists of individual sample data.

V ⁺ _i = {v ⁺ _{i, 0} , v ⁺ _{i, 1} , ...
_{^{·, V + i, m-}} 1} v - i = {v - i, 0, v - i, 1, ···, v -
_{i, m-1} }

Watermark data candidate x^k(K = 0, ...
, L-1) is a pair of sample sets (V⁺, V⁻) To
Embedding as follows, of L types of embedded host data
Candidate W ^kTo generate.

[0074] w^{+ K} _{i, j}= V⁺ _{i, j}+ Α⁺ _{i, j}・ X^k _i w^-K _{i, j}= V⁻ _{i, j}-Α⁻ _{i, j}・ X^k _i Where α⁺ _{i, j}And α⁻ _{i, j}Is the human visual model
Scale to reduce the perceived noise based on
It is a ring parameter, and both are positive values. is there
I, α⁺ _{i, j}And α⁻ _{i, j}Is a probability distribution,
For example, to follow Gaussian distribution, uniform distribution, etc.
It may be a positive value generated by the secret key K. This
, The watermark embedding strength decreases, but
The confidentiality of the generated watermark is improved. In this way, k
Each bit x of the eye watermark data candidate^k _iIs each sub-set
Tov⁺ _i, V⁻ _iOf each of the m samples
Embedded. The larger the number of overlaps m, the more watermark bits
Is less likely to be lost and detection error is reduced.
A watermark that can be embedded in the host data.
The number of bits decreases. α⁺ _{i, j}And α⁻ _{i, j}Is a sight
It is set for each pixel so that visual deterioration cannot be detected.
The number of embedded pixels m is increased in principle.
Even if it does, the deterioration of the image quality is not detected by human vision.
However, the number of pixels used to embed 1 bit increases.
Adding means that the embedded area is limited.
Therefore, the number of bits that can be embedded is reduced.
Taste, and thus, the embedding rate is lowered.

Each of the subsets v ⁺ _i and v ⁻ _i is, for example, a discrete cosine transform (Discrete Cosin Transform) of the host data V.
m sample data selected as a watermark bit embedding target is m DC data blocks included in the DCT block.
It is a CT coefficient. 12A and 12B show 8 × 8 D
M pairs of CT block pairs v ⁺ _i and v ⁻ _i , respectively.
It shows that watermark data x ^k _i is embedded in the CT coefficient. Block pairs v ⁺ _i , v ⁻ _i and m D
The CT coefficient is selected based on the secret key K.

Returning to FIG. 13, the SNR calculation unit 28 uses the L type
W of embedded host data of class ^kAgainst watermark de
Data x^kResistance, that is, the embedding strength is evaluated (S1
6), the selector 30 is the embedding that maximizes the embedding strength.
Only host data candidate W^kThe final embedded host device
It is selected as the data W (S18).

Before giving the embedding strength evaluation formula, how the watermark data X'is detected when the embedded host data W is modified by signal processing, image processing, or the like will be examined. . The deformation applied to the embedded host data W is treated as noise N, and the embedded host data W added with the noise N is referred to as an embedded host signal W ′. A method of extracting the watermark data X'from the embedded host signal W'will be described. Embedding pair of a set of host signal ^{(W '+, W' -} ) is defined as follows. Collection of watermarked host signal W ^'+, W' ^- has the respective n elements as follows.

W '⁺= {W '⁺ ₀, W ’⁺ ₁・ ・ ・ ・ ・ ・
w ’⁺ _n-1} W ’⁻= {W '⁻ ₀, W ’⁻ ₁・ ・ ・ ・ ・ ・
w ’⁻ _n-1} Where the set of embedded host signals
W ’⁺, W ’⁻Each subset w ′ that is an element of ⁺ _i,
w ’⁻ _iCorresponds to the position where the digital watermark is embedded.
M sample data of the embedded host signal W '.
It consists of w ’⁺ _i= {W '⁺ _{i, 0}, W ’⁺ _{i, 1}・ ・ ・ ・ ・ ・
w ’⁺ _{i, m-1}} w ’⁻ _i= {W '⁻ _{i, 0}, W ’⁻ _{i, 1}・ ・ ・ ・ ・ ・
w ’⁻ _{i, m-1}}

In order to detect the watermark bit x ^k _i , the next decision value z _i is calculated. ^{_{z i = Σ j = 0 m}} -1 (w '+ i, j -w' - i, j) = Σ j = 0 m-1 [(w + i, j + n + i, j) - (w
_{^{- i, j + n - i}} , j)] = Σ j = 0 m-1 [(v + i, j -v - i, j) + (α
⁺ _{I, j} + α ⁻ _{i, j} ) · x ^k _i + (n ⁺ _{i, j} −n ⁻
_{i, j} )] Here, Σ _{j = 0} ^m−1 (v ⁺ _{i, j−} v ⁻ _{i, j} ) is m.
When is sufficiently large, it generally follows a Gaussian distribution and approaches zero. Also, the noise term Σ _{j = 0} ^m-1 (n ⁺ _{i, j-} n
^- _{i, j)} as well closer to 0 also. Therefore,
z _i is Σ _{j = 0} ^{m −1} [(α ⁺ _{i, j} + α ⁻ _{i, j} ) ·
x ^k _i ] can be approximated. (Α ⁺ _{i, j} +
alpha ^- _{i, j)} since is positive, the watermark bit ^x _{k i} is 1
Then z _i is positive and if watermark bit x ^k _i is −1 then z _i is negative. Therefore, the value of the watermark bit x ^k _i can be determined according to the sign of z _i .

The evaluation of the embedding strength is performed by regarding the host data V as noise with respect to the watermark data X and calculating the variance of the watermark data detected for the embedded watermark data x ^k . It can be considered that the smaller the dispersion is, the stronger the resistance is. The SN ratio is evaluated by the following equation for a pair of embedded host data candidates (W ^{+ k} , W ^−k ), and the optimum candidate K is selected.

[0081] _{K = argmax k (P k /} (2σ k 2)) P k = Σ i = 0 n-1 | Σ j = 0 m-1 (w + k i, j
^{_{-W -k i, j) | 2}} / n σ k 2 = Σ i = 0 n-1 | Σ j = 0 m-1 (w + k
^{_{i, j -w -k i, j}} ) -P k 1/2 · x k i | 2 / n

Watermark bit x^k _iIs either {1, -1}
The above-mentioned judgment value z for judging whether it is_iIs filled
In the state before noise is added to the host data W
Is z _i= Σ_{j = 0} ^m-1(W^{+ K} _{i, j}-W^-K
_{i, j}), The variance σ
_k ^TwoIs the judgment value z_iWatermark bits detected by
Watermark bit embedded at the time x^k _iI =
0, ..., n-1 are evaluated and averaged.
Can be said. On the other hand, P_kIs the judgment value z_iI = 0, ...
The square root mean square for n-1. Therefore,
Embedded watermark data x^kWatermark data extracted with
Euclidean distance between
The larger the absolute value of the judgment value for issuing, the more P_k/ (2
σ_k ^TwoThe value of) becomes large. In other words, P_k/ (2
σ_k ^Two) Is the best choice for the watermark
Means to select the candidate with the smallest detection error
To do.

For the judgment value z _i , v ⁺ _{i, j} > v ⁺
_{If i, j} and x ^k _i = 1 then z _i >> 0 and v ⁺
_{If i, j} <v ⁺ _{i, j} and x ^k _i = -1, then z _i << 0
Becomes Therefore selecting the candidates of the optimum watermark x ^k by the evaluation described above, in order to improve the detection performance of the watermark bit x ^k _i according to the determination value z _i, v ⁺
_{If i, j} > v ⁺ _{i, j,} then x ′ _i = 1 and v ⁺
_{If i, j} <v ⁺ _{i, j,} then x ′ _i = −1,
This means changing the original watermark bits x _i to x ′ _i . This is the guiding rule of the GS method, which improves the response of the determination value z _i .

In the above description, as shown in FIG.
In order to generate L types of watermark data candidates, L multiplexers 20, scramblers 22, and embedding units 2 are used.
6 and the SNR calculation unit 28 are provided in parallel, but these members may be formed into a single configuration, L-type watermark data candidates may be sequentially generated and evaluated, and the optimum candidate may be selected. .

FIG. 14 shows such a sequential digital watermark.
6 is a flowchart illustrating a procedure for embedding the above. variable
i is initialized to 1 (S20). Multiplexer 20
Are turbo-encoded by the turbo encoding unit 70, and
Thalever π_TwoWatermark interleaved with 32
Data X_cInsert the i-th initial data at the beginning of the code
The sequence is generated (S22), and the scrambler 22 uses the code.
The scrambled sequence is the i-th scrambled
Watermark data X_b'Is generated (S24). Scrunch
I th scrambled generated by bra 22
Watermark data X _b'Is a host by the embedding unit 26
It is embedded in the data V (S26). SNR calculation unit 28
Is a candidate W of the i-th embedded host dataⁱAgainst
Watermark data xⁱResistance, that is, embedding strength S_iTo
Evaluate (S28). The selector 30 has an embedding strength S
_iIs greater than the reference value T that guarantees the lowest evaluation value
It is determined (S30). If the embedding strength S_iIs the standard
If it is larger than the value T (Y in S30), the current value is changed to the variable K.
Substitute the value of number i (S32), and the Kth embedded host
Data candidates are used as the final embedded host data W
Select (S40). Embedding strength S_iIs below the reference value T
In case of (N of S30), the current value of the variable i is equal to L
If so (Y of S34), the embedding strength S examined so far_k
The subscript k that maximizes the value of is assigned to the variable K (S3
8) Finalize the Kth embedded host data candidate
The embedded host data W is selected (S40). Present
If the value of the existing variable i is smaller than L (N in S34), the variable
Increment the number i by 1 (S36), then step
Return to S22.

By this iterative process, when a candidate whose embedding strength is equal to or higher than the desired reference value is obtained, that candidate is selected as the final embedding host data W, and if such a candidate is not generated, It is possible to generate L embedded host data candidates and select the one having the maximum embedded strength as the final embedded host data W from the candidates.

15 (a), 15 (b) and 15 (c) show different types of configurations of the decoding apparatus 40 according to the third embodiment. The embedded host data W in which the electronic watermark is embedded by the encoding device 10 is distributed on the network and used in a computer. In the process, the embedded host data W undergoes operations such as compression encoding and falsification. For image data, useful operations such as signal processing such as JPEG compression, filtering, quantization, color correction, and geometric conversion such as scaling, cropping, rotation, and parallel movement are performed, and digital watermarking is performed. Unauthorized attacks such as removing or altering are added.
The deformation caused by such an operation is regarded as noise N with respect to the embedded host data W, and the embedded host data W added with the noise N is embedded host signal W ′ (= W + N).
And The decoding device 40 performs a process of extracting the embedded watermark data X from the embedded host signal W ′.

FIG. 15A shows the first type decoding device 4
It is 0. When receiving the embedded host signal W ′ with noise, the extraction unit 42 extracts the watermark data using the secret key K, calculates the judgment value z _i of the watermark bit as follows, and judges the judgment value z _i Is sent directly to the GS-APP decoding unit 52 without making a hard decision to {-1, 1}.

[0089] ^{_{z i = Σ j = 0 m}} -1 (w '+ i, j -w' - i, j) = Σ j = 0 m-1 [(w + i, j + n + i, j) - (W
_{^{- i, j + n - i}} , j)] = Σ j = 0 m-1 [(v + i, j -v - i, j) + (α
⁺ _{I, j} + α ⁻ _{i, j} ) · x ′ _i + (n ⁺ _{i, j} −n ⁻
_{i, j} )]

The GS-APP decoding unit 52 performs the soft decision decoding of the watermark bit using the decision value z _i as the reception sequence. The extrinsic information output from the GS-APP decoding unit 52 is deinterleaved by the deinterleaver π ₂ ^-1 54 and then provided as input data to the turbo decoding unit 90 having the general configuration shown in FIG. The turbo decoding unit 90 improves the detection accuracy of watermark bits by iterative decoding, and finally outputs the watermark data X.

FIG. 15B shows the second type decoding device 4
It is 0. When the extraction unit 42 receives the embedded host signal W ′, the extraction unit 42 calculates the watermark bit determination value z _i and sends it to the GS-APP decoding unit 52 without performing a hard determination. The GS-APP decoding unit 52 performs soft decision decoding using the watermark bit decision value z _i as a reception sequence. GS-
The external information output from the APP decoding unit 52 is deinterleaved by the deinterleaver π ₂ ^-1 54 and then given as input data to the feedback-type turbo decoding unit 60 shown in FIG. The turbo decoding unit 60 feeds back the extrinsic information regarding the information bits by the second APP decoding unit 64 of FIG. 6 to the GS-APP decoding unit 52 as reliability information. The GS-APP decoding unit 52 uses this reliability information as a priori probability of information bits to perform soft decision decoding. By repeating this operation, the watermark bits are repeatedly decoded by the connection between the GS-APP decoding unit 52 and the turbo decoding unit 60, the accuracy of detecting the watermark bits is improved, and the watermark data X is finally output. .

FIG. 15C shows a third type decoding device 4
It is 0. The turbo decoding unit 60 uses GS- as external information regarding the information bits and the second parity bits by the second APP decoding unit 64 in FIG. 6 and external information regarding the first parity bits by the first APP decoding unit 62 in FIG. 6 as reliability information. Feedback to the APP decoding unit 52, and GS-A
The PP decoding unit 52 uses this reliability information as a priori probability of the information bit and the parity bit to perform soft decision decoding. The difference from the second type is that the first APP decoding unit 62 and the second APP decoding unit 64 of the turbo decoding unit 60 calculate extrinsic information regarding parity bits. Although this processing increases the calculation amount, it is possible to further reduce the error rate.

As described above, according to the embodiment,
At the time of embedding a digital watermark, the watermark bit sequence is embedded after being converted into a bit sequence that is easy to embed in host data by the GS method using turbo coding with high error correction capability. At the time of extracting a digital watermark, the watermark bit is iteratively decoded by a feedback system in which soft decision decoding of the watermark bit and turbo decoding using the soft decision output as input data are connected. Therefore, it is possible to enhance the resistance of the digital watermark against signal processing, geometric transformation, compression, tampering with data, and the like, and at the same time, to greatly improve the detection accuracy of the digital watermark.

[Embodiment 4] An encoding apparatus 10 according to Embodiment 4 has the same configuration as the encoding apparatus 10 of Embodiment 3 shown in FIG. 9, but the functional configurations of the changing unit 14 and the embedding unit 16 are the same. Is different. FIG.
[FIG. 10] is a functional configuration diagram of a changing unit 14 and an embedding unit 16 of the encoding device 10 according to the fourth embodiment. This embodiment is different from Embodiment 3 shown in FIG. 10 in that turbo coding is performed after scrambling by the GS method.

The L multiplexers 20 are connected to the watermark data.
Initial data C at the beginning of each ₀~ C_L-1Insert
L types of bit sequence X_bTo generate. L scrub
The umber 22 uses the GS system for each of the L types of bit sequences.
More scrambled, L types of scrambled transparent
Scarecrow data X '_bTo generate.

The scrambled watermark data X ′ _b is interleaved by the interleaver π ₂ 32 and then turbo encoded by the turbo encoding unit 70. The turbo encoding unit 70 performs turbo encoding using the scrambled watermark data X ′ _b as input data and outputs turbo encoded watermark data X ′ _c .

The L embedding sections 26 embed each of the L types of scrambled watermark data X ′ _{c in} the host data V to generate L types of embedded host data W candidates. The L SNR calculators 28 evaluate the tolerance of the watermark data X for each of the L types of embedded host data W candidates. The selector 30 selects a candidate of the embedded host data W having the best tolerance evaluation value and outputs it as the final embedded host data W.

17 (a), 17 (b) and 17 (c) show different types of configurations of the decoding apparatus 40 according to the fourth embodiment. FIG. 17A shows a first type decoding device 40. When the extraction unit 42 receives the embedded host signal W ′ with noise, the extraction unit 42 calculates the determination value z _i of the watermark bit as in the third embodiment, and the determination value z _i is {−1,
1} is sent to the turbo decoding unit 90 having the general configuration shown in FIG.

The turbo decoding unit 90 determines the judgment value z._iThe receiving system
Performs soft-decision decoding of watermark bits using as a sequence. Tar
Watermark bit X ′ input to the decoding unit 90_cIs a soft value
Handled and the result X'after turbo decoding_bHard decision, Dane
Thalever π_Two ^-154. The watermark data is
Interleaver π_Two ^-1Deinterleaved by 54
And then given to the descrambler 46. Descra
The ambler 46 uses descrambling in the GS system.
Unscramble the original watermark data X _bOutput
To do.

FIG. 17B shows the second type decoding device 4
It is 0. Upon receiving the embedded host signal W ′, the extraction unit 42 calculates the watermark bit determination value z _i , and sends it to the turbo decoding unit 90 of the general configuration shown in FIG. The turbo decoding unit 90 performs soft decision decoding using the decision value z _i of the watermark bit as a reception sequence. The external information output from the turbo decoding unit 90 is deinterleaved by the deinterleaver π ₂ ^-1 54 and then given to the GS-APP decoding unit 52 shown in FIG. 8 as input data. However, this GS-APP decoding unit 52
Does not feed back the soft value output to the turbo decoding unit 90, but makes the soft decision decoding self-complete and outputs the final decoded result watermark data X _b .

FIG. 17C shows a third type decoding device 4
It is 0. Unlike the second type, the file shown in FIG.
The feedback type turbo decoding unit 60 is used, and G of FIG.
The S-APP decoding unit 52 outputs the external information regarding the information bits.
It is fed back to the turbo decoding unit 90 as reliability information.
It The turbo decoding unit 60 uses this reliability information as information bit information.
Perform soft decision decoding using as a priori probability. Repeat this operation
By returning, the turbo decoding unit 60 and the GS-APP recovery unit
Iterative decoding of watermark bits is performed by the cooperation of the encoder 52.
The accuracy of watermark bit detection is improved and finally
Watermark data X _bIs output.

The present invention has been described above based on the embodiments. It is understood by those skilled in the art that these embodiments are mere examples, and that various modifications can be made to the combinations of the respective constituent elements and the respective processing processes, and that such modifications are also within the scope of the present invention. By the way. For example,
In the present embodiment, a turbo encoder having a configuration in which two convolutional encoders are connected in parallel is used, but a turbo encoder having a configuration in which three or more convolutional encoders are connected in parallel is used. You may use. A turbo encoder having a configuration in which two or more convolutional encoders are serially connected may be used.

In the first and second embodiments, the turbo-encoded GS system has been described on the premise of the communication channel, but this system can also be applied to the reproducing system of the recording apparatus. In the third and fourth embodiments, the application of the turbo-encoded GS method has been described by taking the embedding and extraction of the digital watermark as an example. However, the embodiment can be applied to other fields such as wireless image and It can be effectively applied to communication in an environment where noise is large and the SN ratio is low, such as voice communication.

Especially when the turbo coded GS system is used in the recording / reproducing system, the suitability of the scrambled coded sequence is evaluated in consideration of the following points. The reproducing system of magnetic recording has a differential characteristic, and the DC component is not reproduced. In optical recording, the low frequency component near DC of the modulated signal leaks into the servo signal of the optical disk device, which causes a problem such as malfunction, and it is desirable to suppress the low frequency component near DC. Therefore, when evaluating the suitability of a plurality of scramble sequences, the channel constraint comparison unit 34 of FIGS. 3 and 7 evaluates the DC component of each scramble sequence,
The scrambled sequence with the smallest DC component may be selected as the final encoded sequence. Further, in magnetic recording and optical disks, PLL (Phase Locked Loop) is used to read and write channel bits correctly. By using the PLL, it is possible to absorb the rotation irregularity of the disk and to achieve bit synchronization. PL
In order to prevent malfunction of L, it is preferable that bits having the same polarity do not occur consecutively in the channel bits.
Therefore, the channel constraint comparison unit 34 of FIGS.
In each scramble sequence, the maximum continuous length of bits having the same polarity may be evaluated and compared, and the scramble sequence having the smallest maximum continuous length may be selected as the final encoded sequence.

In order to generate a plurality of watermark data candidates or embedding position candidates, the GS method capable of generating a variety of candidates has been used, but other scrambling methods may be applied. The candidate data may be randomly generated by some method. Further, in the embodiment, the original watermark data is reproduced from the generated watermark data candidates by descrambling, but a table is provided in which the generated watermark data candidates are associated with the original watermark data. The original watermark data may be obtained by referring to.

The identification data used as the initial data at the time of scrambling was inserted at the beginning of the watermark data and provided to the decoding side. However, this identification data is not embedded in the watermark but is kept secret on the encoding side. It may be held and managed as a key. In that case, the decryption side acquires the secret key and then descrambles the watermark data.

As a modification of the coding apparatus 10 of the third embodiment, the configuration and operation for generation and evaluation of sequential candidates have been described. However, the same sequential configuration and operation are used in other embodiments. It goes without saying that the present invention can also be applied to the encoding device 10 of the form.

According to the present invention, the bit error rate is reduced by iterative decoding.

[Brief description of drawings]

FIG. 1 is a diagram illustrating a general configuration of a turbo encoder.

FIG. 2 is a diagram illustrating a general configuration of a turbo decoder.

FIG. 3 is a configuration diagram of an encoding device according to the first embodiment.

FIG. 4 is a diagram illustrating a convolution operation during encoding.

FIG. 5 is a diagram for explaining a convolution operation at the time of decoding.

FIG. 6 is a configuration diagram of a decoding device according to the first embodiment.

FIG. 7 is a configuration diagram of an encoding device according to a second embodiment.

FIG. 8 is a configuration diagram of a decoding device according to a second embodiment.

FIG. 9 is a configuration diagram of an encoding device according to a third embodiment.

FIG. 10 is a functional configuration diagram of a changing unit and an embedding unit in FIG.

FIG. 11 is a diagram illustrating a relationship between original watermark data and L types of scrambled watermark data.

12A and 12B are diagrams illustrating a method of embedding scrambled watermark data.

FIG. 13 is a flowchart illustrating a procedure of embedding a digital watermark by the encoding device according to the third embodiment.

FIG. 14 is a flowchart illustrating another procedure for embedding a digital watermark by the encoding device according to the third embodiment.

15 (a), (b), and (c) are diagrams showing configurations of different types of decoding apparatuses according to Embodiment 3, respectively.

[Fig. 16] Fig. 16 is a functional configuration diagram of an encoding device according to a fourth embodiment.

17 (a), (b), and (c) are diagrams showing configurations of different types of decoding apparatuses according to Embodiment 4, respectively.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 encoding device, 12 encryption part, 14 change part, 16 embedding part, 20 multiplexer,
22 scrambler, 26 embedded part, 28 S
NR calculation unit, 30 selector, 34 channel constraint comparison unit, 40 decoding device, 42 extraction unit, 46 descrambler, 50 GS decoding unit, 52 GS-AP
P decoding unit, 60 turbo decoding unit, 62 1st APP
Decoding unit, 64 Second APP decoding unit, 70 Turbo coding unit, 90 Turbo decoding unit, 102 demultiplexer, 104 multiplexer.

   ─────────────────────────────────────────────────── ─── Continued front page    F term (reference) 5B001 AA01 AA10 AB03 AB05 AC01                       AC03 AC05 AD06 AE07                 5J065 AA01 AB01 AC02 AD01 AD10                       AE06 AE07 AF02 AG05 AG06                       AH01 AH02 AH04 AH09 AH21                 5K014 AA01 BA10 FA16

Claims (18)

[Claims] 1. A turbo-encoded information sequence is scrambled by a convolutional operation to generate a plurality of scramble sequences, the suitability of each of these scramble sequences is evaluated, and the scramble sequences with good evaluation are finally determined. Encoding method, which is obtained as a simple encoded sequence. 2. External information output from a soft output decoder that performs soft decision decoding of a received sequence is given to a turbo decoder as input data, and external information output from the turbo decoder is used in advance in the soft decision decoding. A decoding method, wherein the received sequence is iteratively decoded by feeding back to the soft output decoder as information on probability. 3. A turbo coding unit that turbo-codes an information sequence to generate a turbo code sequence, an interleaver that interleaves the turbo code sequence, and scrambles the interleaved turbo code sequence by a convolution operation to generate a plurality of turbo codes. A scramble unit for generating a scramble code sequence, for each of the plurality of scramble code sequences, an evaluation unit for evaluating suitability, and selecting one of the plurality of scramble code sequences based on the evaluation value of the suitability. An encoding device, comprising: an output selecting unit. 4. A soft-decision decoding unit that performs soft-decision decoding on a received sequence, a deinterleaver that deinterleaves the received sequence that has been soft-decision decoded, and turbo decoding using the deinterleaved received sequence as input data. A decoding device including a turbo decoding unit. 5. A soft-decision decoding unit that performs soft-decision decoding of the received sequence, a deinterleaver that deinterleaves the received sequence that has been soft-decision decoded, and turbo decoding using the deinterleaved received sequence as input data. A turbo decoding unit, wherein the external information output from the turbo decoding unit is fed back to the soft decision decoding unit as information about the prior probability of the reception sequence, whereby the iterative decoding of the reception sequence is performed. Decryption device. 6. The soft decision decoding unit according to claim 5, wherein extrinsic information about the information bits of the reception sequence output from the turbo decoding unit is fed back to the soft decision decoding unit as information about a priori probability of the information bits. Decryption device. 7. The turbo decoding unit performs soft-decision decoding of parity bits of the reception sequence, and external information about the parity bits output from the turbo decoding unit is the soft-decision as information about a priori probability of the parity bits. The decoding device according to claim 6, wherein the decoding device is fed back to the decoding unit. 8. A scramble unit that scrambles an information sequence by a convolution operation to generate a plurality of scramble code sequences, an interleaver that interleaves each of the plurality of scramble code sequences, and a plurality of interleaved scramble code sequences. A turbo encoding unit that turbo-encodes each of the plurality of turbo code sequences to generate a plurality of turbo code sequences, an evaluation unit that evaluates suitability for each of the plurality of turbo code sequences, and a plurality of the plurality of turbo code sequences based on the evaluation value of the suitability. An encoding device, comprising: a selecting unit that selects and outputs one of turbo code sequences. 9. The evaluation unit evaluates the maximum continuous length of bits having the same polarity in the scramble code sequence, and the selection unit selects the scramble code sequence having the minimum maximum continuous length. The encoding device according to claim 3 or 8. 10. The evaluation unit evaluates a DC component included in the scramble code sequence, and the selection unit selects the scramble code sequence having the minimum DC component. Encoding device described. 11. A turbo decoding unit that performs turbo decoding using a reception sequence as input data, a deinterleaver that deinterleaves the turbo-decoded reception sequence, and a soft-decision that performs soft-decision decoding of the deinterleaved reception sequence. A decoding device including a decoding unit. 12. A turbo decoding unit that performs turbo decoding using a reception sequence as input data, a deinterleaver that deinterleaves the turbo-decoded reception sequence, and a soft decision that soft-decides the deinterleaved reception sequence. And a decoding unit, wherein the external information output from the soft-decision decoding unit is fed back to the turbo decoding unit as information related to the a priori probability of the received sequence, whereby the iterative decoding of the received sequence is performed. Decoding device. 13. A turbo encoder that turbo-encodes digital watermark data to be embedded in host data, an interleaver that interleaves the turbo-encoded digital watermark data, and a convolution of the interleaved digital watermark data. A scramble unit that scrambles by operation to generate a plurality of watermark data candidates; an embedding unit that embeds each of the plurality of watermark data candidates into the host data to generate a plurality of embedded host data candidates; For each of the embedded host data candidates, an evaluation unit that evaluates the resistance of the digital watermark, and a selection unit that selects and outputs one of the plurality of embedded host data candidates based on the evaluation value of the resistance. An encoding device comprising: 14. An extracting unit that extracts watermark data from host data in which a digital watermark is embedded, a soft-decision decoding unit that performs soft-decision decoding of the extracted watermark data, and an inverse of the watermark data that has been soft-decision decoded. A decoding device, comprising: a deinterleaver for interleaving; and a turbo decoding unit for performing turbo decoding using the deinterleaved watermark data as input data. 15. The watermark data is repeatedly decoded by feeding back external information output from the turbo decoding unit to the soft-decision decoding unit as information about a priori probability of the watermark data. Item 15. The decoding device according to Item 14. 16. A scramble unit that scrambles digital watermark data to be embedded in host data by a convolution operation to generate a plurality of watermark data candidates, and an interleaver that interleaves each of the plurality of watermark data candidates. A turbo encoder that turbo-encodes each of the plurality of interleaved watermark data candidates; a plurality of turbo-encoded watermark data candidates are embedded in the host data; An embedding unit that generates a plurality of embedded host data candidates, an evaluation unit that evaluates the resistance of the digital watermark for each of the plurality of embedded host data candidates, and one of the plurality of embedded host data candidates based on the resistance evaluation value. And a selection section for selecting and outputting Apparatus. 17. An extracting unit that extracts watermark data from host data in which a digital watermark is embedded, a turbo decoding unit that performs turbo decoding using the extracted watermark data as input data, and the turbo-decoded watermark data. A decoding device comprising: a deinterleaver for deinterleaving, and a soft-decision decoding unit for performing soft-decision decoding of the deinterleaved watermark data. 18. The iterative decoding of the watermark data is performed by feeding back external information output from the soft-decision decoding unit to the turbo decoding unit as information about the prior probability of the watermark data. Item 17. The decoding device according to Item 17.