JP5051051B2 - Apparatus, method and program for embedding and extracting digital watermark information - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To embed a symbol for indicating electronic watermark information in a form suitable for a band where the information is embedded, when transmitting the watermark information embedded in various kinds of bands of a carrier signal. <P>SOLUTION: A band dividing section 110 divides the carrier signal xp(n) into a plurality of band carrier signals. An embedding section 120-0 etc. performs modulation using a modulation signal that a modulation signal creation section 140 outputs to a low component of an amplitude sequence of a band carrier signal X(&omega;00) etc. A band synthesizing section 130 synthesizes the modulated band carrier signal where the information is embedded, with other band carrier signals which is output by the band dividing section 110. The modulation signal creation section 140 outputs a modulation signal with a frequency or an amplitude dependent on each carrier signal where the information is embedded, the modulation signal being applied to various kinds of carrier signals where the information is embedded. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to a digital watermark information embedding technique, and more particularly to a method and apparatus suitable for embedding digital watermark information in an acoustic signal propagating in the air and extracting the digital watermark information embedded in the acoustic signal.

Various techniques have been proposed in which an audio signal such as music is used as a carrier signal that bears the digital watermark information, and the digital watermark information is embedded in a part of the band component of the carrier signal (see, for example, Patent Documents 1 to 3). According to the embedding techniques such as Patent Document 1 and Patent Document 2, for example, compressed encoded data of an audio signal in which digital watermark information is embedded is decoded and reproduced as sound, and this sound is collected by a microphone. Even in the process in which the obtained audio signal is recorded and propagated, digital watermark information can be retained in the recorded audio signal. Therefore, this embedding technique can be used for a wider range of applications than the technique of embedding digital watermark information in compression-encoded data.
Japanese Patent No. 3659321 JP 2006-251676 A Special table 2007-535699

  By the way, when digital watermark information is embedded and transmitted in a part of the band of the carrier signal, a symbol string indicating the digital watermark information is embedded in a plurality of bands of the carrier signal, or the digital watermark information is embedded. The band may be switched depending on the situation. Here, the robustness against the disturbance of the symbol in the transmission process of the carrier signal in which the symbol is embedded is not uniform between the bands. In addition, when modulation corresponding to a symbol string indicating digital watermark information is performed on a part of a band of a carrier signal, depending on the intensity of the modulation, a signal processing system that generates a carrier signal embedded with digital watermark information may be adversely affected. May bring. Specifically, the signal processing system may not satisfy the so-called complete reconstruction condition, and a carrier signal containing a large amount of imaging components may be generated. Here, as is known in many documents related to subband coding and the like, the perfect reconstruction condition is that the carrier signal is band-divided into a plurality of band carrier signals by, for example, analysis filter back and the like, In a signal processing system that performs band synthesis of a band carrier signal using, for example, a synthesis filter bank, the imaging component generated in the band carrier signal in the band division process is canceled in the band synthesis process, and the original carrier signal is completely reproduced. Therefore, this is a condition that the signal processing system should satisfy. In the apparatus for embedding digital watermark information, a part of the signal obtained by the band division is modulated according to the symbol string indicating the digital watermark information. Depending on the intensity of the modulation, the signal processing system of the embedding apparatus May not satisfy the complete reconstruction condition, and a carrier signal in which an imaging component remains may be generated. Then, at what level of modulation, whether the signal processing system of the embedding device does not satisfy the complete reconstruction condition also differs depending on the embedding destination band.

  The present invention has been made in view of the circumstances described above, and when digital watermark information is embedded and transmitted in various bands of a carrier signal, the digital watermark information is displayed in a mode suitable for the band to be embedded. It is an object of the present invention to provide a technique capable of embedding symbols, preventing the occurrence of harmful effects caused by embedding digital watermark information, and enhancing the robustness of the embedded digital watermark information.

The present invention provides band dividing means for performing band division on a carrier signal and outputting a plurality of band carrier signals belonging to different bands, and a symbol indicating digital watermark information for at least one band carrier signal of the plurality of band carrier signals Each embedding destination band carrier signal is an embedding destination band carrier signal indicating each symbol to be embedded in each embedding destination band carrier signal, and each modulation signal having an amplitude or frequency corresponding to the band of each embedding destination band carrier signal is output. Modulation signal generating means for embedding, and embedding for applying low-frequency components of the amplitude sequence of each embedding destination band carrier signal with a modulation signal corresponding to the embedding destination band carrier signal output by the modulation signal generating means And an embedding band after the processing by the embedding means Band combining means for combining a carrier signal and a band carrier signal other than the embedded band carrier signal output from the band dividing means and outputting a carrier signal in which digital watermark information is embedded. An electronic watermark information embedding apparatus is provided.
The present invention also provides band dividing means for performing band division on a carrier signal and outputting a plurality of band carrier signals belonging to different bands, and at least one band carrier signal of the plurality of band carrier signals as digital watermark information. The low frequency component of the amplitude sequence of each embedding destination band carrier signal is calculated from each embedding destination band carrier signal, and the low frequency band of the amplitude sequence of each embedding destination band carrier signal. Each part of the component column indicates various symbols that may be embedded in each embedded band carrier signal, and each modulation signal having a frequency corresponding to the band of each embedded band carrier signal. Correlation is calculated and embedded in each embedded band carrier signal based on the cross-correlation calculation result. Providing an extraction device for an electronic watermark information, characterized by comprising a determining extraction means symbols Mareta electronic watermark information.
According to this invention, when a symbol indicating digital watermark information is embedded and transmitted in various bands of the carrier signal, the modulation signal generating means of the embedding device indicates each symbol to be embedded in each embedded band carrier signal, In addition, each modulation signal having an amplitude or frequency corresponding to the band of each embedding destination band carrier signal is output, and the embedding means generates a modulation signal for the low frequency component of the amplitude sequence of each embedding destination band carrier signal. Modulation is performed with a modulation signal corresponding to the embedded band carrier signal output by the means. Therefore, according to the present invention, when a symbol indicating digital watermark information is embedded and transmitted in various bands of the carrier signal, the symbol indicating digital watermark information can be embedded in a mode suitable for the band to be embedded. In addition, it is possible to prevent an adverse effect caused by embedding the digital watermark information and to improve the robustness of the embedded digital watermark information.

Embodiments of the present invention will be described below with reference to the drawings.
<First Embodiment>
An embedded transmission technique in which digital watermark information is embedded in a carrier signal and transmitted is a blind watermark embedded transmission technique that does not require an original carrier signal in which digital watermark information is not embedded in the extraction of digital watermark information. It is roughly classified into a non-blind watermarking embedded transmission technique that requires a single carrier signal. The first embodiment described below belongs to the former blind watermarking embedded transmission technique. FIG. 1 is a block diagram showing a configuration of a digital watermark embedding device 100 according to the first embodiment of the present invention, and FIG. 2 is a block diagram showing a configuration of a digital watermark extracting device 200 according to the first embodiment. Note that each of the embedding device 100 and the extraction device 200 may be realized as dedicated hardware for executing processing for embedding digital watermark information in a carrier signal or processing for extracting digital watermark information from a carrier signal, or embedding processing. Or a computer program that causes a computer to execute extraction processing. This also applies to other embodiments after the second embodiment.

  First, the digital watermark embedding apparatus 100 will be described. As shown in FIG. 1, the embedding device 100 includes a band dividing unit 110, a band synthesizing unit 130, a plurality of embedding units 120-0, 120-1,. Part 140. In FIG. 1, only two embedding parts 120-0 and 120-1 of the plurality of embedding parts 120-0, 120-1,... Are shown to prevent the drawing from becoming complicated.

  FIG. 3 is a diagram for explaining the processing contents of the band dividing unit 110. In the present embodiment, the carrier signal serving as the digital watermark information is an audio sample xp (n) obtained by sampling an audio waveform at a constant sampling rate. The band dividing unit 110 receives the audio sample xp (n), which is the carrier signal, and divides it into band carrier signals belonging to different M bands.

  More specifically, the band dividing unit 110 includes a windowing unit and a band dividing filter bank (not shown). Here, the windowing unit performs processing for dividing the column of audio samples xp (n) into blocks xp (n) (n = 0 to N−1) of N samples (N = M / 2) per block, Until the latest two blocks of audio samples xp (n) (n = 0 to 2N−1) are multiplied by the window function, the processing is repeated and supplied to the band division filter bank.

Each of the band division filter banks includes M complex coefficients BPF (Band Pass Filter) whose passband center angular frequency is ω _i (i = 0 to M−1). Each complex coefficient BPF corresponding to each passband center angular frequency ω _i (i = 0 to M−1) is two blocks of audio samples xp (n) (n = 0) multiplied by the window function from the windowing unit. ˜2N−1) for each received audio sample sequence xp (n) (n = 0 to 2N−1), BPF using a complex coefficient corresponding to each passband center angular frequency ω _i By performing the processing, the complex spectrum coefficient X (ω _i ) indicating the component of the passband center angular frequency ω _i included in the audio sample sequence xp (n) (n = 0 to 2N−1) is calculated, and the band Output as a carrier signal. In the present embodiment, a symbol string in which each band carrier signal of a part or all of the band carrier signals for M bands obtained from the band division filter bank of the band dividing unit 110 in this way indicates digital watermark information. This is an embedded band carrier signal to be embedded. Further, in the present embodiment, a plurality of embedding destination band carrier signals are divided into two embedding destination band carrier signal pairs whose bands are adjacent to each other, and individual symbols constituting the symbol sequence of the digital watermark information are Each of the two embedded band signals is embedded in a distributed manner. At this time, in the present embodiment, one symbol is embedded in two embedded band carrier signals that form one pair. The symbol sequence embedded in each pair of embedded band carrier signals may be a symbol obtained by decomposing a symbol sequence indicating the same type of digital watermark information, or one type of different types of digital watermark information. It may be a symbol string indicating digital watermark information. As to which of the plurality of band carrier signals obtained by the band division is to be the embedding destination band carrier signal, it is only necessary to agree between the embedding apparatus 100 and the extracting apparatus 200. The selection method is arbitrary.

In FIG. 1, the embedding units 120-0 and 120-1 embed an embedded band carrier signal composed of a sequence of complex spectrum coefficients X (ω ₀ ) and a sequence of complex spectrum coefficients X (ω ₁ ). Is a means for embedding a symbol indicating digital watermark information in the pair. Other embedding units 120-2, 120-3,..., Not shown, are also divided into two pairs. Each pair of embedding units uses two-phase modulated signals indicating symbols to be embedded with respect to the low frequency component (or rough shape) of the amplitude sequence of the embedding destination band carrier signal for two bands given to each of the embedding units. By performing the modulation process, the symbol string of the series is embedded.

  The modulation signal generation unit 140 generates a two-phase modulation signal that indicates a symbol to be embedded, has the same amplitude and frequency, and has an opposite phase relationship with each other, for embedding the symbol in the carrier signal. . The amplitude or frequency of the two-phase modulation signal is determined depending on the band of the embedded band carrier signal pair to which the modulation signal is applied. Details of the modulation signal generated by the modulation signal generation unit 140 will be described later.

Next, the configuration of each embedding unit will be described using the embedding unit 120-0 as an example among the embedding units constituting the pair of embedding units. In the embedding unit 120-0, the absolute value detection unit 121 converts the absolute value | X (ω ₀ ) | shown in the following equation into a real part Re (X (ω ₀ )) and an imaginary part of the complex spectrum coefficient X (ω ₀ ). Calculated from Im (X (ω ₀ )). A sequence of absolute values | X (ω ₀ ) | sequentially calculated by the absolute value detection unit 121 is an amplitude sequence of the embedded band carrier signal.

Further, the declination detection unit 122 converts the declination Arg (X (ω ₀ )) represented by the following expression into the real part Re (X (ω ₀ )) and the imaginary part Im (X (X ()) of the complex spectrum coefficient X (ω ₀ ). ω ₀ )).

The band dividing unit 123 performs band division on the amplitude sequence of the embedded band carrier signal output from the absolute value detecting unit 121. In a preferred aspect, the band dividing unit 123 has a configuration similar to that of the band dividing unit 110 described above, and a sequence of absolute values | X (ω ₀ ) | of complex spectrum coefficients obtained from the absolute value detecting unit 121. In other words, the amplitude sequence of the embedded band carrier signal is divided into sections of a predetermined length, subjected to filter processing or FFT processing by a filter bank for each section, and outputs a plurality of complex spectrum coefficients belonging to different bands. In this case, the lowest Lm complex spectrum coefficient sequence obtained sequentially by band division is a unit for embedding one symbol.

この離散ウェーブレット変換により各種のｊ、ｋに対応したウェーブレット係数Ｗｆ（ｊ，ｋ）が得られると、元の時間関数ｆ（ｔ）は、各種のｊ，ｋに対応したウェーブレット基底ψ_ｊ，ｋ（ｔ）に対して、ウェーブレット係数Ｗｆ（ｊ，ｋ）を重み付け加算した一次結合として近似することができる。 In another preferred embodiment, the band dividing unit 123 performs complex processing on the sequence of absolute values | X (ω ₀ ) | of complex spectrum coefficients obtained from the absolute value detecting unit 121, that is, the amplitude sequence of the embedded band carrier signal. This is a means for performing a discrete wavelet transform. As is well known, the discrete wavelet transform is an arbitrary time function as shown in equation (4) when a discrete wavelet basis ψ _{j, k} (t) as shown in equation (3) is given. This is a conversion equation for calculating a wavelet coefficient Wf (j, k), which is the inner product of f (t) and this wavelet basis ψ _{j, k} (t).

When wavelet coefficients Wf (j, k) corresponding to various kinds of j and k are obtained by this discrete wavelet transform, the original time function f (t) is converted into wavelet bases ψ _{j, k} corresponding to various kinds of _{j and k.} (T) can be approximated as a linear combination obtained by weighted addition of wavelet coefficients Wf (j, k).

The band dividing unit 123, which is a complex wavelet transform unit, is a filter bank as shown in FIG. 4, for example. The complex wavelet transform unit performs a discrete wavelet transform on a sequence of absolute values | X (ω ₀ ) | of the complex spectrum coefficients, and the complex wavelet coefficients are assumed to correspond to the wavelet coefficients Wf (j, k) in Expression (4). Is calculated.

As shown in FIG. 4, the complex wavelet transform unit, which is a filter bank, includes a tree TRa that calculates a real part of a complex wavelet coefficient and a tree TRb that calculates an imaginary part. In the tree TRa, the portion composed of LPF 1231a, HPF 1232a, downsampler 1233a, and 1234a is at level 1 with respect to the absolute value | X (ω ₀ ) | of the continuous L complex spectrum coefficients given from the absolute value detector 121. Perform band division. More specifically, the LPF 1231a includes the wavelet base ψ _{0, k} (t) of j = 0 and the wavelet coefficient Wf (0, k) from the column of absolute values | X (ω ₀ ) | of L complex spectrum coefficients. ) To select a low frequency component that can be approximated by linear combination, and output L samples indicating the low frequency component. Downsampler 1233a, by thinning out the L samples output from the LPF1231a into two at a ratio of one, and outputs the real part _{x 0a} of the wavelet coefficients _{x 0} of L / 2 pieces of level 1. On the other hand, the HPF 1232a has a wavelet basis ψ _{0, k} (t) of j = 0 and a wavelet coefficient Wf (0, k) out of a column of absolute values | X (ω ₀ ) | of L complex spectrum coefficients. A high frequency component that does not belong to the linear combination is selected, and L samples indicating the high frequency component are output. The downsampler 1234a outputs the real part x _1a of the L / 2 level 1 wavelet coefficients x ₁ by thinning out the L samples output from the HPF 1232a at a ratio of one to two.

Further, in the tree TRa, LPF1235a, HPF1236a, parts made of down sampler 1237a and 1238a, compared real part _{x 0a} of L / 2 pieces of complex wavelet coefficients supplied from the down-sampler 1233a, performs band division of level 2. More specifically, the LPF 1235a includes a wavelet basis ψ _{1, k} (t) of j = 1 and a wavelet coefficient Wf (1, k) from the sequence of the real part x _0a of L / 2 complex wavelet coefficients. A low frequency component that can be approximated by linear combination is selected, and L / 2 samples indicating the low frequency component are output. Downsampler 1237a, by thinning out the L / 2 samples output from the LPF1235a into two at a ratio of one, and outputs the real part _{x 00a} of the wavelet coefficients of L / 4 pieces of level 2. On the other hand, the HPF 1236a includes a wavelet basis ψ _{1, k} (t) of j = 1 and a wavelet coefficient Wf (1, k) from the sequence of the real part x _0a of L / 2 complex wavelet coefficients x ₀₀ . A high frequency component that does not belong to the primary combination is selected, and L / 2 samples indicating the high frequency component are output. Downsampler 1238a, by thinning out the L / 2 samples output from the HPF1236a into two at a ratio of one, and outputs the real part _{x 01a} of the wavelet coefficients _{x 01} of L / 4 pieces of Level 2 .

Although the tree TRa has been described above, the tree TRb is also composed of a portion composed of LPF 1231b, HPF 1232b, down samplers 1233b and 1234b, and a portion composed of LPF 1235b, HPF 1236b, down samplers 1237b and 1238b, similar to those of the tree TRa. Yes. Then, by the operation of each of these units, L / 2 level 1 wavelet coefficients x are obtained from a sequence of absolute values | X (ω ₀ ) | of consecutive L complex spectrum coefficients given from the absolute value detection unit 121. ₁ imaginary part x _1b , L / 4 level 2 wavelet coefficients x ₀₀ imaginary part x _00b and L / 4 level 2 wavelet coefficients x ₀₁ imaginary part x _01b are calculated and output.

As illustrated in FIG. 4, when the band dividing unit 123, which is a complex wavelet transform unit, performs band division up to level 2, the absolute value | X (ω ₀ ) | of the complex spectrum coefficient output from the absolute value detecting unit 121 Thus, the real part and the imaginary part of three kinds of wavelet coefficients x ₁ , x ₀₀ and x ₀₁ are obtained. Among these, the absolute value | x ₁ | of the wavelet coefficient x ₁ corresponding to the wavelet base ψ _{0, k} (t) of j = 0 is the amplitude sequence of the embedded band carrier signal (specifically, the absolute value of the complex spectrum coefficient). This represents a detailed change in the value | X (ω ₀ ) | On the other hand, the absolute value | x ₀₀ | of the wavelet coefficient x ₀₀ corresponding to the wavelet basis ψ _{1, k} (t) of j = 1 indicates the amplitude sequence of the embedded band carrier signal most roughly, It shows the outline or low-frequency component of the same amplitude sequence.

  The band dividing unit 123 may have the same configuration as the band dividing unit 110 or may be a complex wavelet transform unit as described above. The other parts will be described by taking the case of a part as an example.

The absolute value detection unit 124 in FIG. 1 has each band component of the amplitude sequence of the embedded band carrier signal output from the band division unit 123, that is, the real part and imaginary part of each wavelet coefficient x ₁ , x ₀₀ and x _01. Based on the above, absolute values | x ₁ |, | x ₀₀ | and | x ₀₁ | of the wavelet coefficients x ₁ , x ₀₀ and x ₀₁ are calculated according to the following equations.

Further, deviation angle detecting unit 125, based on the real and imaginary part of the wavelet coefficients _{x 1, x 00} and _{x 01} obtained from the band division unit 123, the argument of the wavelet coefficients _{x 1, x 00} and _{x 01} Arg (x ₁ ), Arg (x ₀₀ ) and Arg (x ₀₁ ) are calculated according to the following formula.

The multiplier 126 and the adder 127 are absolute values of each band component of the amplitude sequence of the embedded band carrier signal output from the absolute value detection unit 124, that is, absolute values | x ₁ |, | x ₀₀ | of wavelet coefficients. And | x ₀₁ | constitutes an amplitude modulator that applies amplitude modulation to the absolute value | x ₀₀ | that belongs to the lowest band and is considered to best indicate the outline of the amplitude sequence. In more detail, the multiplier 126 receives the modulated signal Am ₀ generated based on the symbol sequence modulating signal generator 140 is an embedded object, the modulated signal Am ₀ absolute value of the wavelet coefficients | x _{00 |} to Multiply. The adder 127 adds the multiplication result and the absolute value | x ₀₀ | of the wavelet coefficient, and outputs an amplitude-modulated absolute value (1 + Am ₀ ) | x ₀₀ |. Here, the modulation signal Am ₀ is a signal obtained by multiplying the basic modulation signal m ₀ having an amplitude from −1 to +1 by the modulation intensity A. The modulation intensity A is selected from a range from 0 to 1. When A = 1, maximum amplitude modulation is performed, and when A = 0, no modulation is performed. How the modulation intensity A is determined will be described later.

The phase recombination unit 128a recombines the amplitude-modulated absolute value (1 + Am ₀ ) | x ₀₀ |) obtained from the adder 127 and the deviation angle Arg (x ₀₀ ) obtained from the deviation angle detection unit 125. Return to the complex wavelet coefficient x ₀₀ indicating the low frequency component of the amplitude sequence of the embedded band carrier signal. In addition, the phase recombination unit 128a quantizes the absolute values | x ₁ | and | x ₀₁ | of the wavelet coefficients obtained from the absolute value detection unit 124 in a predetermined quantization step, and biases the absolute values after the quantization. The declination angles Arg (x ₁ ) and Arg (x ₀₁ ) of each wavelet coefficient obtained from the angle detection unit 125 are recombined, respectively, and returned to the complex wavelet coefficients x ₁ and x ₀₁ . Each wavelet coefficient which is each band component (complex band component) of the amplitude sequence of the embedded band carrier signal obtained in this way is processed by the phase recombining unit 128b and the band synthesizing unit 129.

Here, the reason why the phase recombining unit 128a performs the quantization process on the absolute values | x ₁ | and | x ₀₁ | of the wavelet coefficients that are not subjected to amplitude modulation is as follows. First, wavelet coefficients in adjacent bands are not separated to the extent that they can be said to be practically complete in the frequency domain, but overlap. Therefore, when wavelet coefficients that are not subject to amplitude modulation (embedding destination of digital watermark) among adjacent wavelet coefficients are used for inverse complex wavelet transformation described later without being processed, the wavelet coefficients are converted into audio sample sequences. It will act as a disturbance when extracting a digital watermark from. However, on the other hand, if wavelet coefficients that are not subject to amplitude modulation (embedding destination of the digital watermark) are completely removed, and the complex wavelet inverse transform described later using only the amplitude-modulated wavelet coefficients, This leads to deterioration of the sound quality of the embedded audio sample sequence. Therefore, in consideration of the tradeoff between the extraction accuracy and sound quality of the digital watermark at the time of extraction, the absolute value | x ₁ | of the wavelet coefficient that is not subject to amplitude modulation is set so that both the extraction accuracy and sound quality are appropriately satisfied. | X ₀₁ | is subjected to quantization processing. In addition, when the band dividing unit 123 is not a complex wavelet transform unit but has a configuration like the band dividing unit 110, such a quantization process is not necessary.

The complex wavelet coefficients x ₁ , x _00, and x ₀₁ that have undergone the processing of the phase recombination unit 128 a described above are delivered to the band synthesis unit 129. In the aspect in which the band dividing unit 123 is a complex wavelet transform unit, the band synthesis unit 129 is a complex wavelet inverse transform unit. This complex wavelet inverse transform unit performs the complex wavelet inverse transform on the three types of complex wavelet coefficients x ₁ , x ₀₀ and x ₀₁ delivered from the phase recombination unit 128a, thereby obtaining the absolute value | X ′ (ω ₀ ) | Is output. This complex wavelet inverse transform is the inverse of the process of FIG. That is, for example, focusing on the real part, after performing double upsampling in which zero samples are inserted between each of the real parts x _00a of the L / 4 complex wavelet coefficients, a transfer function opposite to that of the filter 1235a is obtained. After performing filter processing and performing upsampling twice to insert zero samples between each of the real parts _x01a of L / 4 complex wavelet coefficients, filter processing having a transfer function opposite to that of the filter 1236a is performed. And add the results of both filter processes. Thereby, the real part x _{0a of} L / 2 complex wavelet coefficients is obtained. Such a process is repeated from the tip of the tree TRa toward the root and from the tip of the tree TRb toward the root, thereby returning to the state before the complex discrete wavelet transform.

The phase recombination unit 128b uses the deviation angle Arg (X (X ()) output from the deviation angle detection unit 122 with respect to the absolute value | X ′ (ω ₀ ) | of the complex spectral coefficient obtained from the band synthesis unit 129 in this way. ω ₀ )) are recombined to calculate the complex spectral coefficient X ′ (ω ₀ ).
The above is the details of the embedding unit 120-0.

The embedding unit 120-1 has basically the same configuration as the embedding unit 120-0, and the embedding unit 120-0 is a rough shape of the amplitude sequence of the complex spectral coefficient X (ω ₁ ) obtained from the band dividing unit 110. Embeds the same symbol sequence that embeds. However, the equivalent to the multiplier 126 of the embedding portion 120-1, the modulated signal Am ₀ given to the multiplier 126 of the embedding portion 120-0, the frequency and amplitude are the same, and reverse phase The modulation signal Am ₁ having the relationship is supplied from the modulation signal generation unit 140.

Here, the timing relationship between the modulation signals Am ₀ and Am ₁ and the amplitude modulation target using them will be described, and the relationship between the symbol to be embedded and the modulation signals Am ₀ and Am ₁ will be described. 5A and 5D show the absolute values | X (ω ₁ ) | and | X (ω ₀ ) | of the complex spectrum coefficients output from the band dividing unit 110, respectively. 5B and 5E show the absolute value | x ₀₀ (ω ₁ ) | obtained by the embedding unit 120-1 out of the absolute value | x ₀₀ | of the complex wavelet coefficient as the symbol embedding destination. The absolute value | x ₀₀ (ω ₀ ) | obtained in the embedding unit 120-0 is shown. 5C and 5F show the modulation signals Am ₀ and Am ₁ output from the modulation signal generation unit 140, respectively.

As shown in FIGS. 5A and 5B, the embedding unit 120-1 uses a complex type in units of absolute values | X (ω ₁ ) | of L complex spectrum coefficients output from the band dividing unit 110. Discrete wavelet transform is executed, and the absolute value of L / 4 complex wavelet coefficients | x ₀₀ (ω ₁ ) | obtained from the absolute values | X (ω ₁ ) | of the L complex spectrum coefficients is one symbol. It becomes an embedding destination. Further, as shown in FIGS. 5D and 5E, the embedding unit 120-0 uses the absolute value | X (ω ₀ ) | of the L complex spectrum coefficients output from the band dividing unit 110 as a unit. The complex discrete wavelet transform is executed, and the absolute value | x ₀₀ (ω ₀ ) | of the L / 4 complex wavelet coefficients obtained from the absolute values | X (ω ₀ ) | of the L complex spectrum coefficients is 1. This is the symbol embedding destination. As illustrated in FIGS. 5C and 5F, the modulation signal generation unit 140 determines the absolute value | x ₀₀ (ω ₁ ) | or | x ₀₀ (ω ₀ ) | of L / 4 complex wavelet coefficients. Are output as triangular wave modulation signals Am ₁ and Am ₀ having a period of 1 as one period. Here, when the symbol b to be embedded is bit “0”, the modulation signal Am ₁ has a peak in the first half and a valley in the second half, whereas the modulation signal Am ₀ has a valley in the first half and a peak in the second half. When the symbol b to be embedded is bit “1”, the modulation signal Am ₁ has a valley in the first half and a peak in the second half, whereas the modulation signal Am ₀ has a peak in the first half and a valley in the second half. In this way, the modulation signals Am ₁ and Am ₀ representing the same symbol are in opposite phase to each other, and the individual modulation signals Am ₁ or Am ₀ have different waveforms depending on the type of symbol represented by each.

Embedding unit 120-1 and 120-0 receives a modulation signal Am ₁ and Am ₀ is generated based on the common symbol as described above, the complex amplitude modulation using each respective modulated signals Am ₁ and Am ₀ The absolute values | x ₀₀ (ω ₁ ) | and | x ₀₀ (ω ₀ ) | of the complex wavelet coefficients indicating the outline of the amplitude sequence of the spectrum coefficients X (ω ₁ ) and X (ω ₀ ) It generates complex spectral coefficients X ′ (ω ₀ ) and X ′ (ω ₁ ) with embedded symbols.

1 performs processing of the complex spectrum coefficients X ′ (ω ₀ ) and X ′ (ω ₁ ) obtained from the embedding units 120-0 and 120-1 and the embedding destination band carrier signal. The complex spectrum coefficient obtained from each pair of other embedding units to be performed and the band carrier signal (complex spectrum coefficient) of the embedding destination band carrier signal output from the band dividing unit 110 are added back to the time domain signal and added to the window. Multiplication processing is performed, and an audio sample sequence xp ′ (n) in which digital watermark information is embedded is output.

Now, the band dividing unit 110 described above performs band division by dividing the carrier signal into overlapping blocks on the time axis. On the other hand, in the aspect in which the band dividing unit 123 in the embedding units 120-0 and 120-1 is a complex wavelet transform unit, the band dividing unit 123 calculates the absolute value | X of the complex spectrum coefficient that is the embedding destination band carrier signal. Each column of (ω ₀ ) | and | X (ω ₁ ) | is divided into blocks each consisting of L absolute values that do not overlap on the time axis, and complex discrete wavelet transform is performed. In this block unit, 1-bit modulation signals Am ₁ and Am ₀ are generated to perform amplitude modulation. The reason why the block division is different between band division and symbol embedding is as follows.

  First, since the band dividing unit 110 requires high frequency resolution, band division is performed assuming a complex exponential function sequence such as an FFT basis function. For this reason, when band division / synthesis is performed without providing overlap between blocks, signal level discontinuity or phase discontinuity occurs at both ends of the block during synthesis. Therefore, in order to avoid the occurrence of such a problem, the band dividing unit 110 performs band division by dividing the carrier signal into overlapping blocks on the time axis.

On the other hand, in symbol embedding, each block in which symbols are embedded (in this example, absolute values of complex wavelet coefficients for L / 4 | x ₀₀ (ω ₁ ) | or | x ₀₀ (ω ₀ ) |) on the time axis If they are overlapped, interference between symbols embedded at the time of symbol extraction may occur. It is not preferable that such interference occurs. Further, in the aspect using the complex wavelet transform unit as the band dividing unit 123, it is sufficient to embed symbols in the “rough shape (contained to a certain degree from a low frequency to a high frequency)” of the amplitude sequence of the embedded band carrier signal. The complex discrete wavelet transform does not require strict frequency division. Therefore, the embedded band carrier signal is divided into blocks that do not overlap on the time axis, and complex discrete wavelet transform is performed.

  Next, the modulation signal generation unit 140 will be described. Modulation signal generation section 140 generates, for each pair of embedded band carrier signals, two-phase modulation signals that indicate symbols to be embedded, have the same amplitude and frequency, and are in opposite phase relation to each other. The feature of this embodiment is that the frequency or amplitude of the modulation signal generated by the modulation signal generation unit 140 depends on the band of the pair of embedded band carrier signals to which the modulation signal is applied.

  FIGS. 6A to 6C show examples of modulation signals generated by the modulation signal generation unit 140, respectively. 6A to 6C, the horizontal axis represents time, the vertical axis represents the frequency of the embedded band carrier signal to which the modulation signal is applied, and the triangular wave shown in the figure represents each embedded band carrier signal. It is a modulation signal waveform used for modulation of a low frequency component of an amplitude sequence.

In the example shown in FIG. 6A, the frequency of the modulation signal applied to the pair (angular frequencies ω ₀ and ω ₁ ) of the embedded band carrier signal having the lowest frequency is fm, and the embedded band carrier having the middle frequency. The frequency of the modulation signal applied to the signal pair (angular frequencies ω _i and ω _{i + 1} ) is 2 fm, and the modulation signal applied to the pair of the embedded band carrier signals of the highest frequency (angular frequencies ω _j and ω _{j + 1} ) The frequency is 3fm. That is, in this example, the frequency of the modulation signal is increased as the frequency band of the embedded band carrier signal pair is increased.

  In this embodiment, the length of one period of the modulation signal is the symbol frame length. Therefore, in this manner in which the frequency of the modulation signal is changed for each frequency band of the pair of embedded band carrier signals, the symbol frame length is between each pair of the embedded units that process each pair of embedded band carrier signals. It will be different. That is, it is as follows.

First, a modulation signal having the lowest frequency fm is given to the pair of embedding units 120-0 and 120-1 that processes the pair of the embedded band carrier signals with the lowest frequency (angular frequencies ω ₀ and ω ₁ ). It is done. Therefore, in the pair of embedding units 120-0 and 120-1, band division by the band division unit 123 is performed in units of a complex spectrum coefficient sequence (sequence in the time axis direction) corresponding to the symbol frame length 1 / fm. Then, a sequence of a predetermined number of complex wavelet coefficients (or complex spectrum coefficients) is generated, and one symbol is embedded by multiplying each absolute value of the coefficient sequence by a modulation signal for one period.

Other embedded frequencies (for example, 2fm, 3fm, etc.) are included in each embedded portion pair for processing a pair of embedded band carrier signals of other frequencies (such as angular frequencies ω _i and ω _{i + 1} and angular frequencies ω _j and ω _{j + 1} ). Are provided. Therefore, in each of these embedding unit pairs, the band division by the band dividing unit 123 and the modulation processing by the modulation signal for one symbol are performed with the reciprocal of the frequency of the modulation signal given to each being one symbol frame length.

When the frequency of the modulation signal is changed in accordance with the band of the embedded band carrier signal as in the example of FIG. First, the low frequency component of the amplitude sequence of the embedded band carrier signal has an amplitude variation, and the frequency that dominates the amplitude variation is actually different depending on the band of the embedded band carrier signal. Specifically, the dominant frequency of the amplitude fluctuation that appears in the low frequency component of the amplitude sequence of the embedded band carrier signal becomes higher as the band of the embedded band carrier signal becomes higher. Therefore, as in the example of FIG. 6A, for each pair of embedded band carrier signals, modulation is performed at a frequency near the dominant frequency of the amplitude variation of the low frequency component of the amplitude sequence, so that the extraction device 200 side The digital watermark information can be extracted more easily. Also, in this example, the higher the embedded band carrier signal, the higher the frequency of the modulation signal used for modulation processing, the shorter the symbol frame length, and the greater the number of symbols per unit time of the embedded symbol string . Therefore, when a large number of symbol groups are dispersed and embedded in a plurality of pairs of embedded band carrier signals, the overall data rate can be increased.

  In the example of FIG. 6B, the modulation intensity A of the modulation signal used for the processing of the pair is reduced as the pair of the embedded band carrier signal becomes higher. This is merely an example, and the modulation intensity A may be increased as the pair of embedded band carrier signals becomes higher, or may be changed in other modes. Here, when the modulation intensity A is large, it is easy to cause a sense of incongruity in the sense of hearing when a carrier signal in which symbols are embedded is output as sound, but resistance to noise is enhanced. Therefore, according to this example, there is an advantage that auditory influence and noise resistance can be controlled according to the band of the embedded band carrier signal.

The example shown in FIG. 6C is a mode of generating a modulation signal having the characteristics of the examples of FIGS. 6A and 6B. In other words, in this example, both the frequency and the modulation intensity of the modulation signal are changed depending on the band of the embedded band carrier signal pair. According to this example, both advantages obtained by the example of FIG. 6A and advantages obtained by the example of FIG. 6B can be obtained.
The above is the details of the embedding device 100 shown in FIG.

Next, the digital watermark extracting apparatus 200 will be described with reference to FIG. In the extraction device 200, the band dividing unit 210 has the same configuration as the band dividing unit 110 of the embedding device 100. The band dividing unit 210 receives a sequence of audio samples xp ′ (n) in which digital watermark information is embedded, and from this sequence of audio samples xp ′ (n), for example, complex spectrum coefficients X ′ (that belong to M bands). ω _k ) (k = 0 to M−1) is generated.

A portion including the absolute value detection unit 221-0, the band division unit 223-0, and the absolute value detection unit 224-0 includes an absolute value detection unit 121, a band division unit 123, and an absolute value in the embedding unit 120-0 of the embedding device 100. This is a part that plays the same role as the part comprising the value detection unit 124. That is, this part obtains the absolute value | X ′ (ω ₀ ) | of the complex spectrum coefficient X ′ (ω ₀ ) obtained from the band dividing unit 210, and enters this absolute value | X ′ (ω ₀ ) | A discrete discrete wavelet transform is performed, and one absolute value of the complex wavelet coefficient obtained as a result, that is, an absolute value (1 + Am ₀ ) | x ₀₀ (ω ₀ ) subjected to amplitude modulation by the modulation signal Am _{0 in} the embedding device 100. | Is calculated. Similarly, the portion including the absolute value detection unit 221-1, the band division unit 223-1, and the absolute value detection unit 224-1 includes an absolute value detection unit 121 and a band division unit in the embedding unit 120-1 of the embedding device 100. 123 and a part that plays the same role as the part that includes the absolute value detection unit 124. That is, this part is the complex spectrum output by the band dividing unit 210 as the absolute value (1 + Am ₁ ) | x ₀₀ (ω ₁ ) | of the complex wavelet coefficient subjected to amplitude modulation by the modulation signal Am ₁ in the embedding device 100. Calculated from the coefficient X ′ (ω ₁ ).

Then, the divider 251 divides the output information of the absolute value detection unit 224-1 by the output information of the absolute value detection unit 224-0 as shown in the following equation, and outputs the modulation waveform data m that is the result of the division. To do.

Here, the complex wavelet coefficient x ₀₀ (ω ₁ ) and the complex wavelet coefficient x ₀₀ (ω ₀ ) are obtained from complex spectrum coefficients in adjacent bands, and thus are approximate to each other. In addition, various disturbances may occur during propagation of a sequence of audio samples xp ′ (n) in which digital watermark information is embedded, and noise based on the disturbances may be mixed into the audio samples xp ′ (n). This noise becomes an in-phase component and is included in the complex wavelet coefficient x ₀₀ (ω ₁ ) and the complex wavelet coefficient x ₀₀ (ω ₀ ). Therefore, in the division of the above equation (11), the complex wavelet coefficients | x ₀₀ (ω ₁ ) | and | x ₀₀ (ω ₀ ) | containing in-phase noise cancel each other, and the modulation waveform data m is ideally As shown in the following equation (12), it can be approximated to a value depending only on the modulation signals Am ₁ and Am ₀ indicating the bit b of the digital watermark information.

  The offset removing unit 252a obtains an average value (that is, a moving average value) of the modulated waveform data m output within a certain past period, and subtracts it from the modulated waveform data m as an offset. As a result, the offset component (DC component) is removed from the modulated waveform data m.

  The normalization unit 252b obtains the maximum value of the absolute value of the modulation waveform data m output from the offset removal unit 252a within the past certain period, and the modulation waveform data m currently output from the offset removal unit 252a is determined by this maximum value. Normalization is performed by division, and normalized modulation waveform data mn is output.

  The digital watermark information extraction unit 260 is a means for extracting the bit b of the digital watermark information from the normalized modulation waveform data mn, and includes a buffer 261, a modulation signal generation unit 262, a correlation coefficient calculation unit 263, a correlation And a determination unit 264. The buffer 261 is a buffer for storing the latest L / 4 pieces of modulation waveform data mn output from the normalization unit 252b. The operations of the modulation signal generation unit 262, the correlation coefficient calculation unit 263, and the correlation determination unit 264 are the synchronization acquisition period until the symbol delimiter position in the modulation waveform data mn output from the normalization unit 252b is found, and subsequent synchronizations. It will be different depending on the maintenance period.

First, during the synchronization acquisition period, every time the buffer 261 takes in the new modulation waveform data mn from the normalization unit 252b and stores the new modulation waveform data mn, the bit “0” on the modulation signal generation unit 140 side of the embedding device 100 is stored. Modulation signal having the same waveform as the modulation signal m ₁ corresponding to “a (modulation signal having an amplitude of ± 1 before multiplication by the modulation intensity A)” and the modulation having the same waveform as the modulation signal m ₁ corresponding to the bit “1”. Signal. Further, the correlation coefficient calculation unit 263 has a correlation coefficient R0 indicating the correlation between the L / 4 modulation waveform data mn in the buffer 261 and the modulation signal corresponding to the bit “0” generated by the modulation signal generation unit 262. And a correlation coefficient R1 indicating the correlation between the L / 4 modulation waveform data mn in the buffer 261 and the modulation signal corresponding to the bit “1” generated by the modulation signal generation unit 262 is calculated. Then, the correlation determination unit 264 determines the break position of the bit b in the modulation waveform data mn based on the change state of the correlation coefficients R0 and R0 obtained each time new modulation waveform data mn is stored in the buffer 261. judge. Specifically, when it is determined that the peak of the correlation coefficient R0 or R1 occurs at a period of L / 4 samples, it is determined that the peak generation point is the break position of the bit b, and then The operation shifts to the operation of the synchronization maintaining period described.

  During the synchronization maintaining period, the modulation signal generation unit 262 writes the modulation signal and the bit “0” corresponding to the bit “0” every time L / 4 pieces of modulation waveform data mn representing one bit b are written to the buffer 261. A modulation signal corresponding to 1 ″ is generated. The correlation coefficient calculation unit 263 calculates a correlation coefficient R0 indicating the correlation between the L / 4 modulation waveform data mn in the buffer 261 and the modulation signal corresponding to the bit “0”, and also stores the correlation coefficient R0 in the buffer 261. The correlation coefficient R1 indicating the correlation between the L / 4 modulation waveform data mn and the modulation signal corresponding to the bit “1” is calculated. Then, when R0> R1, the correlation determination unit 264 outputs “0” as the bit b of the digital watermark information that is the extraction result, and when R1> R0, the digital watermark that is the extraction result “1” is output as the bit b of the information.

The above is the details of the configuration for extracting the digital watermark information symbol from one of the embedded band carrier signal pairs. When the embedding device 100 side embeds a symbol string of digital watermark information in a plurality of pairs of embedding destination band carrier signals by embedding them, the extracting device 200 embeds the embedding destination band carrier signal obtained from the band dividing unit 210. With respect to each pair, symbols are extracted with the above-described configuration, and the extracted digital symbols are reproduced by collecting the extracted symbols.
The above is the details of the present embodiment.

  According to the present embodiment described above, the mode of the modulation signal used for modulation of the low frequency component of the amplitude sequence of the pair is changed depending on the band of the embedded band carrier signal pair. There is an effect that the symbol can be embedded in a mode suitable for the band to be embedded. More specifically, when the frequency of the modulation signal is changed depending on the band of the embedded band carrier signal pair, the amplitude fluctuation dominant in the low frequency component of the amplitude sequence of the embedded band carrier signal is dominant. By performing amplitude modulation according to the symbol in accordance with the frequency, there is an effect that it is possible to more easily extract digital watermark information on the extraction device 200 side. Also, in this case, the higher the frequency of the embedded band carrier signal, the higher the frequency of the modulation signal used for modulation processing and the shorter the symbol frame length. The data rate as a whole can be increased when transmitting by embedding in pairs. Also, when changing the modulation intensity of the modulation signal depending on the band of the embedded band carrier signal, the auditory influence and noise resistance can be controlled according to the band of the embedded band carrier signal. There is an effect that can be done.

Second Embodiment
In this embodiment, the first embodiment belonging to the blind watermarking embedded transmission technique is changed to a non-blind watermarking method. Although not shown, the digital watermark information embedding device according to the present embodiment omits, for example, the embedding unit 120-1 from the embedding device 100 according to the first embodiment, and replaces only the embedding unit 120-0 with the band dividing unit 110 and The band composition unit 130 is interposed.

  FIG. 7 is a block diagram showing the configuration of the digital watermark information extraction apparatus 200A in this embodiment. This extraction device 200A adds a band dividing unit 210 ′ to the extraction device 200 in the first embodiment, and an absolute value detection unit 221-1, a band division unit 223-1 and an absolute value detection of the extraction device 200. The unit 224-1 is replaced with an absolute value detection unit 221-0 ′, a band division unit 223-0 ′, and an absolute value detection unit 224-0 ′. In FIG. 7, the illustration of the portion after the offset removing unit 252a is omitted.

In this embodiment, the audio sample sequence xp ′ (n) in which the digital watermark information is embedded is synchronized with the supply of the audio sample sequence xp ′ (n) to the band dividing unit 210, and the original audio sample sequence xp (n) in which the digital watermark information is not embedded. ) Is supplied to the band dividing unit 210 ′. The band division section 210 'performs band division of the original audio sample sequence xp (n), complex spectrum coefficient band dividing section 210 outputs X' (ω ₀₎ in the same band as the complex spectrum coefficients X (ω ₀₎ Is output. A portion including the absolute value detection unit 221-0 ′, the band division unit 223-0 ′, and the absolute value detection unit 224-0 ′ includes an absolute value detection unit 221-0, a band division unit 223-0, and an absolute value detection. The same processing as the processing performed by the part 224-0 for the complex spectrum coefficient X ′ (ω ₀ ) is performed for the complex spectrum coefficient X ′ (ω ₀ ).

As a result, as already described in the first embodiment, the absolute value detection unit 224-0 outputs from the absolute value (1 + Am ₀ ) | of the complex wavelet coefficient subjected to amplitude modulation by the modulation signal Am ₀ corresponding to the symbol. x ₀₀ (ω ₀ ) | is output. On the other hand, the absolute value detection unit 224-0 ′ outputs the absolute value | x ₀₀ (ω ₀ ) | of the complex wavelet coefficient not subjected to amplitude modulation by the modulation signal Am ₀ . In the divider 251, the output data (1 + Am ₀ ) | x ₀₀ (ω ₀ ) | of the absolute value detection unit 224-0 is converted into the output data | x ₀₀ (ω ₀ ) | of the absolute value detection unit 224-0 ′. The division result (1 + Am ₀ ) is output as modulation waveform data. The contents of the process of extracting symbols from the modulated waveform data are the same as those in the first embodiment.

<Third Embodiment>
In the first embodiment, amplitude modulation using a modulation signal corresponding to a symbol is performed on a complex wavelet coefficient indicating an outline of an amplitude sequence of a complex spectrum coefficient in a predetermined band of the audio sample sequence xp (n). The symbol was embedded. On the other hand, in this embodiment, phase modulation using a modulation signal corresponding to a symbol is performed on the complex wavelet coefficient indicating the outline of the amplitude sequence of the complex spectrum coefficient in a predetermined band of the audio sample sequence xp (n). By embedding, symbols are embedded. More specifically, in this embodiment, each complex wavelet coefficient of the amplitude sequence of each complex spectrum coefficient of two adjacent bands of the audio sample sequence xp (n) has a phase difference corresponding to the digital watermark information. Perform phase modulation.

  FIG. 8 is a block diagram showing the configuration of the digital watermark information embedding device 100B according to this embodiment. In this embedding device 100B, the embedding units 120-0B and 120-1B have a configuration in which a part of the embedding units 120-0 and 120-1 of the embedding device 100 of the first embodiment is changed. In the embedding device 100B, the modulation signal generation unit 140 in the first embodiment is replaced with a modulation signal generation unit 140B. The configuration of the other parts is basically the same as that of the embedding device 100 of the first embodiment.

In the embedding unit 120-0 </ b> B, the declination detecting unit 125 is based on the real part and imaginary part of each wavelet coefficient output from the band dividing unit 123, as in the first embodiment, and each wavelet coefficient x ₁ , x _00. and declination _Arg (x 1) of _{x 01,} calculates the Arg _{(x 00)} and Arg _{(x 01).} The adder 125a adds the modulation signal Δp ₀ given from the modulation signal generation unit 140B to, for example, the deviation angle Arg (x ₀₀ ) among these deviation angles.

The phase recombination unit 128a recombines the absolute value | x ₀₀ | of the wavelet coefficient obtained from the absolute value detection unit 124 and the phase-modulated argument Arg (x ₀₀ ) + Δp ₀ obtained from the adder 125a, back to the complex wavelet coefficient _{x 00.} Further, the phase recombination unit 128a quantizes the absolute value | x ₁ | of the wavelet coefficient obtained from the absolute value detection unit 124 at a predetermined quantization step, and converts the absolute value after the quantization from the declination detection unit 125. The obtained argument Arg (x ₁ ) is recombined and returned to the complex wavelet coefficient x ₁ . Further, the phase recombination unit 128a quantizes the absolute value | x ₀₁ | of the wavelet coefficient obtained from the absolute value detection unit 124 at a predetermined quantization step, and converts the absolute value after the quantization from the declination detection unit 125. The obtained argument Arg (x ₀₁ ) is recombined and returned to the complex wavelet coefficient x ₀₁ . The three types of complex wavelet coefficients x ₁ , x _00, and x ₀₁ obtained in this way are to be processed by the band synthesizer 129 (see FIG. 1).

The embedding unit 120-1B performs the same processing as the embedding unit 120-0B on the complex spectrum coefficient X (ω ₁ ) obtained from the band dividing unit 110 (see FIG. 1). However, the equivalent to the adder 125a of the buried portion 120-1B, in cooperation with the modulating signal Delta] p ₀ applied to the adder 125a of the buried portion 120-0B modulated signal representing a phase difference corresponding to the symbol Δp ₁ is supplied from the modulation signal generation unit 140B.

More specifically, the modulation signal generator 140B outputs complementary symmetrical triangular wave modulation signals Δp ₁ and Δp ₀ similar to the modulation signals m ₁ and m _{0 of the first} embodiment, but for one period (1 While the modulation signal Δp ₁ -Δp _{0 of} bits) is output, the difference Δp ₁ -Δp ₀ of the modulation signal changes from −R to + R (where R is a value equal to or less than π). ). The mode of the change is determined by whether the symbol indicating the digital watermark information is bit “0” or bit “1”.

  FIG. 9 is a block diagram showing the configuration of the digital watermark information extraction apparatus 200B in this embodiment. This extraction device 200B includes absolute value detection units 224-1 and 224-0 and a divider 251 in the subsequent stage of the band dividing units 223-1 and 223-0 in the extraction device 200 in the first embodiment, and a declination detection unit. The configuration is replaced with 225-1 and 225-0 and a subtractor 253.

In this configuration, the absolute value detection unit 221-0, the band division unit 223-0, and the declination detection unit 225-0 include an absolute value detection unit 121 and a band division in the embedding unit 120-0B of the embedding device 100B. This is a part that plays the same role as the part composed of the part 123 and the declination detection part 125. That is, this part obtains the absolute value | X (ω ₀ ) | of the complex spectrum coefficient X (ω ₀ ) obtained from the band dividing unit 210, and the complex type discrete element is arranged in the column of the absolute value | X (ω ₀ ) |. The wavelet transform is performed, and one deviation angle of the complex wavelet coefficient obtained as a result, that is, the deviation angle Arg (x ₀₀ (ω ₀ )) + Δp ₀ subjected to the phase modulation by the modulation signal Δp ₀ in the embedding apparatus 100B is calculated. To do. Similarly, an absolute value detection unit 221-1, a band division unit 223-1, and a declination detection unit 225-1 include an absolute value detection unit 121 and a band division unit in the embedding unit 120-1B of the embedding device 100B. 123 and a portion that plays the same role as the portion that includes the declination detection unit 125. That is, this part includes the complex spectrum coefficient output from the band dividing unit 210, the deviation angle Arg (x ₀₀ (ω ₁ )) + Δp ₁ of the complex wavelet coefficient that has been phase-modulated by the modulation signal Δp ₁ in the embedding device 100B. Calculated from X (ω ₁ ).

Then, the subtractor 253 calculates the difference {Arg (x ₀₀ (ω ₁ )) + Δp ₁ } − {Arg (x ₀₀ (ω ₀ )) + Δp ₀ } between the two declinations. Here, since the argument angle Arg (x ₀₀ (ω ₁ )) and the argument angle Arg (x ₀₀ (ω ₀ )) are in adjacent bands, they are considered to be approximate. Therefore, the subtraction result obtained from the subtracter 253 is a value close to Δp ₁ −Δp ₀ . Then, a process of extracting a symbol indicating digital watermark information from the subtraction result Δp ₁ -Δp ₀ is performed by the part after the normalization unit 252b. Since the contents of this process are basically the same as those in the first embodiment, description thereof will be omitted.

<Fourth embodiment>
In this embodiment, the third embodiment belonging to the blind watermarking embedded transmission technique is changed to a non-blind watermarking method. Although not shown, the digital watermark information embedding device according to the present embodiment omits, for example, the embedding unit 120-1B from the embedding device 100B according to the third embodiment, and replaces only the embedding unit 120-0B with the band dividing unit 110 and The band composition unit 130 is interposed. The modulation signal Δp0 used for phase modulation by the embedding unit 120-0B is a triangular wave that changes from −R to + R (R is a value of π or less) with a phase corresponding to a symbol indicating digital watermark information.

  FIG. 10 is a block diagram showing the configuration of the digital watermark information extraction apparatus 200C in this embodiment. This extraction device 200C adds a band division unit 210 ′ to the extraction device 200B in the third embodiment, and an absolute value detection unit 221-1, a band division unit 223-1, and a declination detection of the extraction device 200B. The unit 225-1 is replaced with an absolute value detection unit 221-0 ′, a band division unit 223-0 ′, and a declination detection unit 225-0 ′. In FIG. 10, the illustration of the portion after the offset removing unit 252a is omitted.

Similar to the second embodiment, in this embodiment, the audio sample sequence xp ′ (n) in which the digital watermark information is embedded is supplied to the band dividing unit 210 and the digital watermark information is embedded. The original original audio sample sequence xp (n) is supplied to the band dividing unit 210 ′. The band division section 210 'performs band division of the original audio sample sequence xp (n), complex spectrum coefficient band dividing section 210 outputs X' (ω ₀₎ in the same band as the complex spectrum coefficients X (ω ₀₎ Is output. A portion including the absolute value detection unit 221-0 ′, the band division unit 223-0 ′, and the declination detection unit 225-0 ′ includes the absolute value detection unit 221-0, the band division unit 223-0, and the declination detection. The same process as the process performed by the part 225-0 for the complex spectrum coefficient X ′ (ω ₀ ) is performed for the complex spectrum coefficient X (ω ₀ ).

As a result, as previously described in the third embodiment, from the argument detector 225-0, argument Arg (x of complex wavelet coefficients phase modulation is performed by the modulation signal Delta] p ₀ corresponding to the digital watermark information ₀₀ (ω ₀ )) + Δp ₀ is output. On the other hand, the deviation angle detection unit 225-0 ′ outputs the deviation angle Arg (x ₀₀ (ω ₀ )) of the complex wavelet coefficient that has not been subjected to phase modulation by the modulation signal Δp ₀ . Then, in the subtracter 253, the output data Arg deflection angle detecting unit _{225-0 (x 00 (ω 0)} ) + Δp output data _{Arg (x 00 (ω 0)} ) of the deviation angle detecting unit 225-0 'from ₀ Subtraction is performed, and the subtraction result Δp0 is output as modulation waveform data. The contents of the process of extracting the symbol indicating the digital watermark information from the modulated waveform data are the same as those in the first embodiment.

<Fifth Embodiment>
FIG. 11 is a block diagram showing the configuration of a digital watermark information embedding device 100D according to the fifth embodiment of the present invention. The present embodiment is a means for determining the modulation intensity of the modulation signal applied to the band of the embedded band carrier signal as in the example of FIG. 6B described above in the first embodiment. It has the characteristics. The embedding device 100D according to the present embodiment has a configuration in which an energy detection unit 150 is added to the embedding device 100 according to the first embodiment. The energy detection unit 150 receives information indicating the low frequency component of the amplitude sequence of the embedding destination band carrier signal to be processed by each of the embedding units 120-0, 120-1,. get. Specifically, when the band dividing unit 123 of each embedding unit performs complex wavelet transform (see FIG. 4) up to level 2, for example, the energy detection unit 150 includes the absolute value detection unit 124 of each embedding unit. The absolute value | x ₀₀ (ω _i ) | (i = 0, 1,...) Of the wavelet coefficient to be output is acquired.

Then, the energy detection unit 150 calculates an index E (bg) represented by the following equation for each group bg of the embedding destination band to which the same modulation intensity is applied. Here, each group bg has one or more pairs of embedding destination bands in which the same symbol is embedded. Various methods can be considered as to how to divide the entire embedded band into a plurality of groups bg. For example, considering the human psychoacoustic model, the entire embedded band is divided by 1/3 octave intervals. A method of forming a plurality of groups bg can be considered.

In the above equation (13), x ₀₀ (ω _bb , k) (k = 0 to (Lm−1)) is one symbol frame obtained from the amplitude sequence of the embedded band carrier signal in the band bb of the angular frequency ωbb. This is the absolute value of (Lm = L / 4) complex wavelet coefficients. The denominator of the above equation (13) is one symbol frame (Lm = L / 4) of complex wavelets for all embedding destination bands bb belonging to the embedding destination band group allbg used for embedding symbols in the embedding device 100D. The sum of the squares of the absolute values of the coefficients, that is, a value corresponding to the sum of the energy of the low frequency components of all the embedded band carrier signals within the period of one symbol frame. Further, the numerator of the above equation (13) is the sum of squares of absolute values of complex wavelet coefficients of one symbol frame (Lm = L / 4) for each embedding destination band bb belonging to a specific embedding destination band group bg. , That is, a value corresponding to the sum of the energy of the low frequency components of each embedded band carrier signal of the embedded band group bg within the period of one symbol frame.

The energy detection unit 150 sends the index E (bg) thus calculated for each group bg of the embedding destination band to the modulation signal generation unit 140. Based on the index E (bg) corresponding to each group bg in the embedding destination band, the modulation signal generation unit 140 embeds each group in such a manner that the modulation intensity A decreases as the index E (bg) increases. The modulation intensity A of each modulation signal used for amplitude modulation of the low frequency component of the amplitude sequence of the band carrier signal is determined. As an example, modulation allocated to a group in which the maximum value of E (bg) is max (E (bg)), the minimum value is min (E (bg)), and E (bg) is maximum through all the groups in the embedded band. When the intensity is Amin and the modulation intensity assigned to the group having the smallest E (bg) is Amax, the modulation intensity A (bg) for each group bg in the embedding destination band is calculated according to the following equation.

  In the present embodiment, the energy detection unit 150 calculates an index E (bg) that is a relative value of the energy of the low frequency component of the amplitude sequence of each embedded band carrier signal for each symbol frame. Then, the modulation signal generation unit 140 calculates each modulation intensity A (bg) to be applied to each group bg of the embedding destination band from each index E (bg), and the symbol to be embedded is calculated for each group bg. The ± 1 modulation signal shown is multiplied by the modulation intensity A (bg) and output.

  Here, in order to obtain the modulation intensity A applied to the low frequency component (Lm complex wavelet coefficients) of the amplitude sequence of the embedded band carrier signal in a symbol frame, Lm complex wavelet coefficients in the symbol frame are obtained. We have to wait for it to be obtained from the band dividing unit 123. For this reason, in the present embodiment, when the band division unit 123 is calculating the complex wavelet coefficient of the symbol frame, the modulation signal generation unit 140 calculates the modulation intensity calculated from the complex wavelet coefficient of the immediately preceding symbol frame. A multiplier 126 and an adder 127 perform amplitude modulation on the absolute value of the complex wavelet coefficient of the immediately preceding symbol frame.

  When the modulation intensity A of the modulation signal is switched in symbol frame units, if a sudden change occurs in the energy distribution of the low frequency component of each group bg in the embedding destination band, the modulation intensity A (bg applied to each group is affected by the influence. ) Is not preferable because a sudden change occurs. Therefore, in a preferred embodiment, the modulation intensity A (bg) corresponding to each group bg calculated for each symbol frame is not applied to the modulation process as it is, but the modulation intensity A (bg) is subjected to LPF processing, After abrupt changes in modulation intensity A (bg) are relaxed, the modulation intensity A (bg) is used for modulation processing.

  According to this embodiment, the following effects can be obtained. First, when each embedded band carrier signal is compared, if the modulation intensity at the time of symbol embedding is increased in the embedded band carrier signal having a large spectral energy, the imaging component remaining in the carrier signal at the time of band synthesis increases. This is not preferable because it causes deterioration of sound quality. On the other hand, the embedded band carrier signal with low spectral energy has little effect on the sound quality associated with symbol embedding, but it is not robust against noise and other disturbances during transmission, so the modulation intensity at symbol embedding is increased. It is necessary to improve robustness. According to the present embodiment, a small modulation strength is applied to the embedded band carrier signal with a large energy in the low frequency component of the amplitude sequence, and a large modulation is applied to the embedded bandwidth signal with a low energy in the low frequency component of the amplitude sequence. Since the intensity is applied, it is possible to realize both suppression of deterioration of sound quality due to symbol embedding and improvement of robustness against disturbance of embedded symbols for all embedding destination bands. Also, according to the present embodiment, the optimum modulation strength A (bg) is calculated for each group bg of the embedding destination band in symbol frame units, and the modulation strength A (bg) used for amplitude modulation is updated. Even in situations where the power spectrum distribution of the signal changes from moment to moment, both the suppression of sound quality degradation due to symbol embedding and the improvement of robustness against disturbance of the embedded symbol are realized for all sections in which the symbol sequence is embedded. Can do.

<Other embodiments>
Although the first to fifth embodiments of the present invention have been described above, various other embodiments are conceivable for the present invention. For example:

(1) In each embodiment corresponding to the blind watermark method, two band carrier signals adjacent to each other in the band are embedded band carrier signals in which the same symbol is embedded, but the two embedded band carrier signals are not necessarily adjacent to each other. It does not have to be in the band to be.

(2) In the above embodiment, one symbol is a bit representing a binary value, but three or more types of modulation waveforms representing one symbol are prepared, and a symbol representing three or more values per symbol is embedded. It may be embedded in a band carrier signal.

(3) In the above embodiment, a plurality of embedding destination band carrier signals or pairs thereof are used, and symbols of digital watermark information are distributed and embedded in each. However, instead of using all of a plurality of embedded band carrier signals or pairs thereof for symbol embedding in this way, a plurality of embedded band carrier signals or pairs that are candidates for symbol embedding are prepared. One embedded band carrier signal or a pair thereof may be selected from these candidates according to a predetermined rule (a rule agreed with the extraction device) and used for symbol embedding. In this case, the frequency or amplitude of the two-phase modulation signal indicating the symbol is set to the band of the embedding destination band carrier signal or the pair selected as the embedding destination, as illustrated in FIGS. 6A to 6C, for example. It may be changed depending on it.

1 is a block diagram showing a configuration of a digital watermark information embedding device 100 according to a first embodiment of the present invention. FIG. It is a block diagram which shows the structure of the electronic watermark information extraction apparatus 200 by the embodiment. It is a figure which shows the processing content of the band division | segmentation part 110 of the embedding apparatus 100 in the embodiment. It is a figure which shows the processing content of the complex discrete wavelet transform by the band division part 123 in the embodiment. In this embodiment, the timing relationship between the modulation signals Am0 and Am1 and the target of amplitude modulation using them will be described, and the time between the symbol b of the digital watermark information to be embedded and the modulation signals Am0 and Am1 will be described. It is a chart. FIG. 4 is a waveform diagram showing an example of a modulation signal generated by a modulation signal generation unit 140 in the same embodiment. It is a block diagram which shows the structure of 200 A of electronic watermark information extraction apparatuses by 2nd Embodiment of this invention. It is a block diagram which shows the structure of the embedding apparatus 100B of the digital watermark information by 3rd Embodiment of this invention. It is a block diagram which shows the structure of the electronic watermark information extraction apparatus 200B by the embodiment. It is a block diagram which shows the structure of the electronic watermark information extraction apparatus 200C by 4th Embodiment of this invention. It is a block diagram which shows the structure of embedding apparatus 100D of the digital watermark information by 5th Embodiment of this invention.

Explanation of symbols

100, 100B, 100D ... Embedding device, 200, 200A, 200B, 200C ... Extraction device, 110, 210, 210 '... Band-splitting unit, 120-0, 120-1, 120-0B, 120-1B ... ... Embedding part, 121, 124, 221-0, 221-1, 224-0, 224-1, 221-0 ', 224-0' ... Absolute value detection part, 122, 125, 225-0, 225 1, 225-0 '... declination detectors, 128a, 128b ... phase recombination units, 123, 223-0, 223-1, 223-0' ... band division units, 129 ... band synthesis units, 140, 140B: Modulation signal generation unit, 126: Multiplier, 127, 125a ... Adder, 251: Divider, 252a ... Offset removal unit, 252b ... Normalization unit, 260 ... Digital watermark information Extraction Part.

Claims (7)

Band dividing means for performing band division on the carrier signal and outputting a plurality of band carrier signals belonging to different bands;
At least one band carrier signal of the plurality of band carrier signals is set as an embedding destination band carrier signal that is an embedding destination of a symbol indicating digital watermark information, each symbol to be embedded in each embedding destination band carrier signal, and each embedding Modulation signal generating means for outputting each modulation signal having an amplitude or frequency corresponding to the band of the previous band carrier signal;
An embedding unit that modulates a low frequency component of an amplitude sequence of each embedding destination band carrier signal with a modulation signal corresponding to the embedding destination band carrier signal output by the modulation signal generating unit;
The embedded band carrier signal that has undergone the processing by the embedding unit and the band carrier signal other than the embedding destination band carrier signal output by the band dividing unit are combined to output a carrier signal in which digital watermark information is embedded. An electronic watermark information embedding apparatus comprising: a band synthesizing unit. Energy detecting means for detecting the energy of the low frequency component of the amplitude sequence of each embedded band carrier signal;
The modulation signal generating means includes the embedded band carrier signal so that the modulation intensity decreases in a group in which the energy of the low frequency component is relatively large among the groups in which the embedded band carrier signals are grouped by band. The modulation intensity of the modulation signal applied to each group of signals is controlled based on the energy of the low frequency component of the amplitude sequence of the embedded band carrier signal detected by the energy detection means. An electronic watermark information embedding device. Band dividing means for performing band division on the carrier signal and outputting a plurality of band carrier signals belonging to different bands;
At least one band carrier signal of the plurality of band carrier signals is set as an embedding destination band carrier signal which is an embedding destination of a symbol indicating digital watermark information, and the amplitude sequence of each embedding destination band carrier signal is reduced from the embedding destination band carrier signal. Calculates the band component, indicates each part of the low band component column of the amplitude sequence of each embedding destination band carrier signal, and indicates various symbols that may be embedded in each embedding destination band carrier signal, and each embedding Calculates the cross-correlation with each modulated signal having a frequency corresponding to the band of the pre-band carrier signal, and determines the symbol of the digital watermark information embedded in each pre-band carrier signal based on the cross-correlation calculation result And a digital watermark information extracting device. A band division process of performing band division on the carrier signal and outputting a plurality of band carrier signals belonging to different bands;
At least one band carrier signal of the plurality of band carrier signals is set as an embedding destination band carrier signal that is an embedding destination of a symbol indicating digital watermark information, each symbol to be embedded in each embedding destination band carrier signal, and each embedding A modulation signal generation process for outputting each modulation signal having an amplitude or frequency corresponding to the band of the previous band carrier signal;
An embedding process for modulating a low frequency component of an amplitude sequence of each embedding destination band carrier signal with a modulation signal corresponding to the embedding destination band carrier signal output in the modulation signal generation process;
The embedded band carrier signal that has undergone the processing in the embedding process and the band carrier signal other than the embedded band carrier signal that was output in the band dividing process are combined to output a carrier signal in which digital watermark information is embedded A method of embedding digital watermark information, comprising: A band division process of performing band division on the carrier signal and outputting a plurality of band carrier signals belonging to different bands;
At least one band carrier signal of the plurality of band carrier signals is set as an embedding destination band carrier signal which is an embedding destination of a symbol indicating digital watermark information, and the amplitude sequence of each embedding destination band carrier signal is reduced from the embedding destination band carrier signal. Calculates the band component, indicates each part of the low band component column of the amplitude sequence of each embedding destination band carrier signal, and indicates various symbols that may be embedded in each embedding destination band carrier signal, and each embedding Calculates the cross-correlation with each modulated signal having a frequency corresponding to the band of the pre-band carrier signal, and determines the symbol of the digital watermark information embedded in each pre-band carrier signal based on the cross-correlation calculation result And a digital watermark information extracting method. On the computer,
A band division process of performing band division on the carrier signal and outputting a plurality of band carrier signals belonging to different bands;
At least one band carrier signal of the plurality of band carrier signals is set as an embedding destination band carrier signal that is an embedding destination of a symbol indicating digital watermark information, each symbol to be embedded in each embedding destination band carrier signal, and each embedding A modulation signal generation process for outputting each modulation signal having an amplitude or frequency corresponding to the band of the previous band carrier signal;
An embedding process for modulating a low frequency component of an amplitude sequence of each embedding destination band carrier signal with a modulation signal corresponding to the embedding destination band carrier signal output in the modulation signal generation process;
The embedded band carrier signal that has undergone the processing in the embedding process and the band carrier signal other than the embedded band carrier signal that was output in the band dividing process are combined to output a carrier signal in which digital watermark information is embedded A computer program characterized by executing a band synthesizing process. On the computer,
A band division process of performing band division on the carrier signal and outputting a plurality of band carrier signals belonging to different bands;
At least one band carrier signal of the plurality of band carrier signals is set as an embedding destination band carrier signal which is an embedding destination of a symbol indicating digital watermark information, and the amplitude sequence of each embedding destination band carrier signal is reduced from the embedding destination band carrier signal. Calculates the band component, indicates each part of the low band component column of the amplitude sequence of each embedding destination band carrier signal, and indicates various symbols that may be embedded in each embedding destination band carrier signal, and each embedding Calculates the cross-correlation with each modulated signal having a frequency corresponding to the band of the pre-band carrier signal, and determines the symbol of the digital watermark information embedded in each pre-band carrier signal based on the cross-correlation calculation result A computer program characterized by causing an extraction process to be executed.

Images