CN101038771A - Novel method of digital watermarking for protecting literary property of music works - Google Patents

本发明属于信息安全，多媒体信息处理领域，具体为一种用于音乐作品版权保护的数字水印新方法，主要包括同步码、水印嵌入，及检测。本发明首先将音乐作品进行分段处理，然后再将每个音乐数据段划分成前后两个部分以进行信息嵌入。在前部分通过修改多个采样值的统计均值嵌入同步码，而后部分用于水印信息的嵌入，其嵌入过程为，首先对音乐数据段进行DWT，再对小波域的低频系数进行DCT，然后利用自适应量化策略将加密后的水印信息嵌入到DCT域的低中频系数中，最后通过逆变换过程及音乐数据段重新组合得到嵌入水印信息的数字音乐作品。数字水印检测是嵌入过程的逆过程。本发明不依赖于对阈值的调整，可应用于不同种类的音乐作品中。

The invention belongs to the fields of information security and multimedia information processing, and specifically relates to a new digital watermark method for copyright protection of music works, which mainly includes synchronization code, watermark embedding, and detection. The invention first divides the music works into segments, and then divides each music data segment into two parts before and after to carry out information embedding. In the first part, the synchronization code is embedded by modifying the statistical mean value of multiple sampling values, and the latter part is used to embed the watermark information. The embedding process is as follows: first, perform DWT on the music data segment, and then perform DCT on the low-frequency coefficients in the wavelet domain, and then use The adaptive quantization strategy embeds the encrypted watermark information into the low intermediate frequency coefficients in the DCT domain, and finally obtains the digital music works with embedded watermark information through the process of inverse transformation and recombination of music data segments. Digital watermark detection is the reverse process of embedding process. The present invention does not depend on the adjustment of the threshold, and can be applied to different kinds of music works.

The novel method of digital watermarking that is used for protecting literary property of music works

Technical field

The invention belongs to Information hiding and digital watermark technology field in the multi-media information security; be specifically related to a kind of digital audio frequency watermark new method; can effectively resist time domain synchronization attack very difficult in the audio frequency watermark technology, have great practical value for the copyright protection of musical works.

Background technology

Along with the develop rapidly of network technology (particularly Internet technology) with the multimedia messages treatment technology, the intellectual property protection of copyright becomes the key issue that presses for solution.Under this background, digital watermark technology has been subjected to people's common concern, and has become a focus of network information security research field.According to range of application, can be divided into image watermark, video watermark and audio frequency watermark etc. to digital watermarking usually.In recent years, a lot of for the research of image watermark and video watermark (especially image watermark), and the research of relevant digital audio frequency watermark seems and relatively lags behind.Especially, it is deficient unusually especially effectively to resist the strong steady audio-frequency water mark method of synchronization attack.

So-called synchronization attack (Synchronization attack) is not that the attack of this kind of finger can be removed watermark information from contain the watermark object, and is meant that it can be deformed to digital watermarking detecting device and detect not come out (promptly changing the watermark embedded location).By the end of at present, people mainly adopt the steady audio-frequency water mark scheme of four kinds of measures design synchronous attack resistants, are respectively exhaustive search, the spread-spectrum watermark spreading code combines, utilizes original audio key character, synchronous code etc.Wherein, exhaustive search scheme has that calculated amount is big, the more high weakness of false alarm rate; The spread-spectrum watermark spreading code scheme of combining can't realize the blind Detecting of watermark information; Utilize original audio key character scheme to have feature point extraction instability, required threshold value and too much be unfavorable for deficiencies such as practical application.By contrast, the synchronous code scheme has more tangible technical advantage.But there is following deficiency in the present watermark embedding method majority based on the synchronous code scheme: the Barker code length that (1) is adopted is too short, causes that easily vacation synchronously, has influenced the robustness of synchronizing signal; (2) only synchronous code is embedded on the single sampled value of DAB, greatly reduces the ability (particularly attack such as resampling, MP3 compression) that opposing is attacked; (3) fail to determine quantization step (promptly having adopted fixing watermark embedding depth), influenced the not sentience of digital watermarking in conjunction with the auditory masking characteristic.

Pertinent literature is as follows:

[1]Huang J W，Wang Yong，Shi Y.Q.A Blind Audio Watermarking Algorithm WithSelf-synchronization.In：Proceedings of IEEE International Symposium on Circuits and System，Arizona，USA，2002，3：627-630.

Summary of the invention

In view of above-mentioned existing in prior technology problem; the present invention has designed a kind of novel method of digital watermarking that is used for protecting literary property of music works; this method is chosen 16 more sane Barker codes as sync mark; and, adopt the adaptive quantizing strategy that watermark information is embedded in the DAB in hybrid domain in conjunction with the auditory masking characteristic simultaneously by revising the average statistical embedding synchronous code of a plurality of sampled values.

Basic functional principle of the present invention is: at first original audio signal is carried out staging treating, and then each audio data section is divided into former and later two parts to carry out the information embedding.Wherein, previous section is used for the spatial domain and embeds synchronizing signal, considers the instability of single sampled value, so in the telescopiny of synchronizing signal, adopted the method for the average statistical of revising a plurality of sampled values, has improved the robustness of synchronous code.Aft section is used for the hybrid domain embed watermark information.And the concrete telescopiny of watermark information is, at first audio data section (aft section) is carried out DWT, and the low frequency coefficient to wavelet field carries out DCT again, utilizes the watermark information after the adaptive quantizing strategy will be encrypted to be embedded in the Low Medium Frequency coefficient in DCT territory at last.Carry out just because of the embedding of watermark is a energy according to audio frequency,, can be applicable in different types of musical works so the present invention does not rely on the adjustment to threshold value.

Audio frequency watermark significant synchronously.When carrying out real-time watermark detection, detecting device must be by detecting the reference position that synchronizing information determine that watermark embeds accurately.And the effect of synchronous code in the isochronous audio watermark is most important, and can synchronous code accurately locate to have determined whether detecting device is effective, so the robustness of synchronous code also just becomes the key of watermark detection system.

The novel method of digital watermarking that is used for protecting literary property of music works of the present invention; this method can be by choosing 16 more sane Barker codes as sync mark, determine quantization step, synchronous code is embedded into the measures such as average statistical of a plurality of sampled values in conjunction with the digital audio content self-adaptation, with digital watermark signal in hybrid domain is embedded in original digital audio.Simultaneously, emulation experiment shows that also audio-frequency water mark method proposed by the invention not only has preferably not sentience, and all has robustness preferably for Audio Processing such as superimposed noise, resampling, re-quantization, shearing, MP3 compression and attack.In addition; the present invention has an outstanding advantage; promptly highly depend on adjustment the method that in former document, proposes to some thresholding or supposition parameter; can not add in the different types of musical works of being applied to of modification; and need not characteristics such as initial carrier audio frequency when extracting watermark, this has strengthened it greatly and has been used for the practicality of DAB Works copyright protection.

Description of drawings

The present invention has four accompanying drawings, wherein,

Fig. 1: the structure that is used for the embedded data of novel method of digital watermarking of protecting literary property of music works

Fig. 2: the embedding flow process that is used for the novel method of digital watermarking of protecting literary property of music works

Fig. 3: the testing process that is used for the novel method of digital watermarking of protecting literary property of music works comprises that synchronous code detects and two key links of watermark signal extraction

Fig. 4 (a): original audio signal

Fig. 4 (b): original watermark image

Fig. 4 (c): the digital audio and video signals behind the embed watermark

Fig. 4 (d): the watermarking images that from contain the watermark sound signal, extracts

Embodiment

Digital watermarking of the present invention embeds strategy: be embedded in synchronizing signal in spatial domain, (DCT and DWT) is embedded in watermark information in hybrid domain.Suppose that the original digital audio signal is A={a (i), 0≤i＜Length}.Wherein, Length is the number of voice data, a (i) ∈ 0,1,2 ..., (2 ^p-1) } be the range value of i voice data, p is the employed bit number of each data of expression.Suppose binary bitmap be W={w (i, j), 0≤i＜M, 0≤j＜N}.Wherein, (i, j) { 0,1} represents capable, the j row pixel value of the i of binary bitmap to ∈ to w.The assumes synchronization sign indicating number is F={f (i), 0≤i＜Lsyn}, wherein, f (i) ∈ 0, and 1} (for the ease of using, former Barker code is carried out simple transformation here, promptly+1 → 0 ,-1 → 1), Lsyn is the length of synchronous code.Then the blind watermark embed process of ADAPTIVE MIXED domain digital audio (committed step) can be described below.

1 pre-service

In order to eliminate the pixel space correlativity of binary bitmap, improve the security performance of whole digital watermarking system, guarantee still can recover watermark whole or in part after a DAB part is damaged, should at first carry out scrambling encryption to binary bitmap.For this reason, the present invention at first introduces the Arnold conversion, is the secure watermark matrix W with binary bitmap W scrambling encryption ₁, W wherein ₁={ w ₁(i, j), 0≤i＜M, 0≤j＜N}.

In order further to improve attacking ability synchronously such as opposing shearing, translation, so that the DAB partial data still can guarantee the effectively reliable of watermark detection after losing, the present invention at first carries out staging treating to the original digital audio signal, and then each audio data section is divided into length is L ₁And L ₂Two parts and embed synchronous code and watermark information respectively.Wherein, i audio data section can be expressed as

Here, L=L ₁+ L ₂, L ₁=Lsyn * n, L ₂=M * N * 2 ^H+1, H is a wavelet transformation progression, n is constant (the present invention is chosen for 5).

If certain audio data section is A ⁰(its length is L ₁And L ₂Two parts be A ₁ ⁰And A ₂ ⁰), then the telescopiny of synchronous code and digital watermark information is as follows.

2 synchronous codes embed

In order to guarantee the well balanced of not sentience and robustness, the present invention adopts the method for revising a plurality of sampled values (n sampled value), synchronizing information is embedded in the average statistical of sampled value, and concrete steps are:

(1) with A ₁ ⁰Syn is divided into the Lsyn section by the synchronous code length L, each section PA ₁ ⁰(m) contain n sampled point, promptly

PA 1 0 ( m ) = { p a 1 0 ( m ) ( i ) = a 1 0 ( i + m &times; n ) , 0 &le; i < n , 0 &le; m < Lsyn }

(2) calculate PA ₁ ⁰(m) mean value, promptly

PA 1 0 ( m ) &OverBar; = 1 n &Sigma; i = 0 n - 1 pa 1 0 ( m ) ( i ) , ( 0 &le; m < Lsyn )

pa 1 0 &prime; ( m ) ( i ) = pa 1 0 ( m ) ( i ) + ( PA 1 0 &prime; ( m ) &OverBar; - PA 1 0 ( m ) &OverBar; )

Wherein, mod is a modulo operation, S ₁Be quantization step.

The embedding of 3 watermark signals

(1) DWT of audio data section (aft section).To each audio data section A ₂ ⁰Carry out H level wavelet transformation, and obtain wavelet coefficient A ₂ ^0H, D ₂ ^0H, D ₂ ^0H-1..., D ₂ ⁰¹Wherein, A ₂ ^0HBe audio data section A ₂ ⁰The H level of wavelet transformation is similar to component, D ₂ ⁰¹..., D ₂ ^0H-1, D ₂ ^0HBe respectively audio data section A ₂ ⁰The details component of the 1st～H level of wavelet transformation.

(2) DCT of approximate component.Consider that the approximate component of wavelet transformed domain can effectively resist various interference, select the approximate component A of H level wavelet transformation here ₂ ^0HCarry out DCT and embed watermark.

A 2 0 HC = DCT ( A 2 0 H ) = { a 2 0 ( t ) HC , 0 &le; t < L 2 / 2 H }

(3) embedding of watermark signal.Be the robustness that guarantees audio frequency watermark and sentience not, the present invention is according to human auditory system's frequency domain masking characteristics, adopts quantization strategy that watermark information is embedded into the Low Medium Frequency part in DCT territory, promptly quantizes M * N DCT coefficient before the modulation, obtains A ₂ ^{0 ' HC}, wherein:

A 2 0 &prime; HC = a 2 0 &prime; ( t ) HC , 0 &le; t < M &times; N a 2 0 ( t ) HC , M &times; N &le; t < L 2 / 2 H

Quantizing rule is as follows:

Wherein, 0≤i＜M, 0≤j＜N, S ₂Be quantization step.

(4) inverse DCT.To the audio data section A behind the embed watermark ₂ ^{0 ' HC}Do inverse DCT.

A 2 0 &prime; H = IDCT ( A 2 0 &prime; HC )

(5) contrary DWT.With A ₂ ^{0 ' H}Replace A ₂ ^0HAnd carry out the contrary DWT of H level, obtain containing the voice data segment signal A of watermark ₂ ^{0 '}

4 circulations embed

The self-adaptation of quantization step is chosen

Concerning embed scheme based on the audio frequency watermark that quantizes, choosing of quantization step is most important.Because quantization step and watermark embed strength are closely related, the quantization step value is big more, digital watermarking robust performance good more (but simultaneously also easier of audio frequency introducing distortion).Choose and determine that quantization step should take into full account DAB own characteristic and visual masking characteristic.Yet the existing audio frequency watermark scheme that embeds has generally adopted the uniform quantization strategy, promptly whole DAB is adopted an identical quantization step.Like this, if the quantization step value is less, then can influence the anti-aggressiveness of digital watermarking; And if the quantization step value is bigger, then can bring considerable influence to digital audio quality.In addition, different carrier audio frequency is only adopted different quantization step values and could be reached reasonable concealment effect separately.

In order effectively to overcome the existing deficiency of uniform quantization strategy, take into account the practical application (all determine a quantization step if embed point to each, calculated amount is too big) of method simultaneously, the present invention will determine quantization step S in conjunction with whole digital audio content in time domain ₁, in frequency domain, determine quantization step S in conjunction with the audio data section content ₂, promptly

At first, determine quantization step S according to the time domain masking characteristic ₁(being used for synchronous code embeds) has

Wherein, s _a=| A| represents the absolute value average (being the average of peak value) of original digital audio A.

Then, determine quantization step S2 (being used for watermark information embeds) in conjunction with frequency domain masking characteristics self-adaptation, concrete steps are

(1) audio data section A ₂ ⁰(A ₂ ⁰Implication the same) DWT.To audio data section A ₂ ⁰Carry out H level wavelet transformation, and obtain wavelet coefficient A ₂ ^0H, D ₂ ^0H, D ₂ ^0H-1..., D ₂ ⁰¹Wherein, A ₂ ^0HBe audio data section A ₂ ⁰The H level of wavelet transformation is similar to component, and D ₂ ^0H, D ₂ ^0H-1..., D ₂ ⁰¹Be respectively audio data section A ₂ ⁰The details component of the 1st～H level of wavelet transformation.

(2) DCT of approximate component.Here select the approximate component A of H level wavelet transformation ₂ ^0HCarry out DCT.

A 2 0 HC = DCT ( A 2 0 H ) = { a 2 0 ( t ) HC , 0 &le; t < L 2 / 2 H }

(3) quantization step S ₂Self-adaptation choose.Because watermark information has been embedded in preceding M * N the coefficient in DCT territory, so disturb in order to reduce digital watermarking, to guarantee that watermark extraction process can effectively calculate quantization step, the present invention is in conjunction with the audio data section content and utilize back M * N the coefficient self-adaptation in DCT territory to determine quantization step S ₂Promptly

The leaching process of digital watermarking

The ADAPTIVE MIXED domain digital audio water mark method of discussion of the present invention belongs to blind watermark method, does not need the original audio carrier when promptly detecting digital watermark signal.Whole digital watermarking testing process comprises that synchronous code detects and two key links of watermark signal extraction, as shown in Figure 3.

The detection of 1 synchronous code

Here said synchronous code detects, and is meant that (size is L to detection window ₁) whether contain embedded synchronous code information in the audio data section that covered.Its job step is:

(1) job step of choosing according to the self-adaptation of quantization step, self-adaptation is determined quantization step S ₁ ^*

PA ^*(m)＝{pa ^*(m)(i)＝a ^*(i+m×n)，0≤i＜n，0≤m＜Lsyn}

(3) extract synchronous code information F '.Fetch strategy is

(4) in order effectively to avoid false synchronia to take place, the frame synchronization code that the present invention has adopted communications field comparative approach is by turn determined synchronous code, if promptly synchronous code F ' that is extracted and former synchronous code F are identical, then think to have detected embedded synchronous code information.

The extraction of 2 watermark signals

Here said watermark signal extracts, and is meant that audio data section behind the Secondary Synchronization Code (makes that this audio data section is A ₂ ^0*, its length is L ₂) the interior watermark information that extracts.Its job step is:

(1) job step of choosing according to the self-adaptation of quantization step is in conjunction with audio data section A ₂ ^0*Content-adaptive is determined quantization step S ₂ ^*

(2) to audio data section A ₂ ^0*Carry out H level DWT, obtain wavelet coefficient A ₂ ^0*H, D ₂ ^0*H, D ₂ ^0*H-1..., D ₂ ^0*1

(3) pairing approximation component A ₂ ^0*HDo dct transform

A 2 0 * HC = DCT ( A 2 0 * H ) = { a 2 0 * ( t ) HC , 0 &le; t < L 2 / 2 H }

(4) extract watermark information, the extraction formula is

(5) the watermark information W ' that is extracted is carried out inverted and disorderly encrypt, just can obtain binary bitmap

W ^*＝{w ^*(i，j)，0≤i＜M，0≤j＜N}。

In addition, influence for subjective and objective factors such as the experience of eliminating the observer, physical qualification, experiment conditions, (Normalized Cross-Correlation NC), carries out quantitative evaluation to the watermark that extracts and the similarity of original watermark must to adopt normalized correlation coefficient.

Below our good effect of bringing in conjunction with experimental data explanation the present invention.

Provided the experimental result that detects performance test, anti-attack ability test respectively, and contrasted with document [1] method.In the experiment, selected initial carrier is that sample frequency is 44.1kHz, and resolution is 16 bits, and length is 9.75 seconds monophony digital audio and video signals.64 * 64 bianry image has been adopted in digital watermarking.Wavelet transformation has adopted common Daubechies-1 wavelet basis.Wavelet transformation progression is chosen for H=2, and to have selected code length for use be that 16 Barker code 1111100110101110 is as synchronizing signal.Because the watermark of adopting is visual sensuously, so the watermark information that extracts is easy to distinguish.

The anti-attack ability test

In order to detect robustness of the present invention, emulation experiment has been carried out a series of attacks to the watermark sound signal that contains of the present invention and document [1] method respectively, comprising: resampling, and re-quantization, Gaussian noise is disturbed, MP3 compression and shearing etc.Wherein

(1) resampling.The sampled audio signal frequency is dropped to 22.05kHz, 11.025kHz, 8kHz, restore and be former sample frequency 44.1kHz.

(2) re-quantization.Be 8 bits with sound signal from 16 bit quantizations earlier, be quantified as 16 bits again.

(3) superimposed noise.To digital audio and video signals add in time domain that average is 0, mean square deviation is 0.01 white Gaussian noise.

(4) MP3 compression.Earlier sound signal is carried out MP3 compression (being respectively 320k, 256k, 112k, 64k, 48k, 32k by frequency), decompress again.

(5) shear.Shear audio frequency front 10% part and random shearing audio frequency middle 10% respectively.

Table 1 has provided the anti-attack ability results of comparison (the Y-PSNR PSNR that comprises the normalized correlation coefficient NC of the watermarking images that extracted, digital watermarking and distortion rate BER, DAB) of the present invention and document [1] method.

Emulation experiment also shows, audio-frequency water mark method proposed by the invention not only has preferably not sentience (referring to the PSNR in the table 1), and all has robustness (referring to NC in the table 1 and BER, particularly attack such as resampling, MP3 compression) preferably with attacking for Audio Processing such as superimposed noise, resampling, re-quantization, shearing, MP3 compressions.In addition, the hybrid domain audio frequency watermark embedding grammar that the present invention proposes also has characteristics such as calculating simply, realize easily, need not when extracting watermark the initial carrier audio frequency, and this has strengthened it greatly and has been used for the practicality of DAB Works copyright protection.

1、一种用于音乐作品版权保护的数字水印新方法，其特征在于包括嵌入水印的预处理，同步码、水印的嵌入和同步码、水印的检测，其中：1. A new digital watermarking method for copyright protection of music works, characterized in that it includes preprocessing of embedding watermark, embedding of synchronous code and watermark and detection of synchronous code and watermark, wherein: 数字水印的预处理：Preprocessing of digital watermarking: 为了消除二值水印图像的象素空间相关性，提高整个数字水印系统的安全性能，确保数字音频某一部分受到破坏后仍能全部或部分地恢复水印，宜首先采用Arnold变换对二值水印图像进行置乱加密。In order to eliminate the pixel spatial correlation of the binary watermark image, improve the security performance of the entire digital watermark system, and ensure that the watermark can still be fully or partially restored after a certain part of the digital audio is damaged, it is advisable to use the Arnold transform to perform binary watermark image firstly. Scramble encryption. 为了进一步提高抵抗剪切、平移等同步攻击能力，以使得数字音频部分数据丢失后仍能保证水印检测的有效可靠，本发明首先对原始数字音频信号进行分段处理，再将每个音频数据段分割成前后两部分并分别嵌入同步码与水印信息。In order to further improve the ability to resist synchronous attacks such as cutting and translation, so that the effective and reliable watermark detection can still be guaranteed after the digital audio data is lost, the present invention firstly performs segment processing on the original digital audio signal, and then divides each audio data segment It is divided into two parts, the front and the back, and the synchronization code and watermark information are embedded respectively. 同步码嵌入：为了保证不可感知性与稳健性的良好平衡，本发明采用修改多个采样值(n个采样值)的方法，将同步信息嵌入到采样值的统计均值中。Synchronization code embedding: In order to ensure a good balance between imperceptibility and robustness, the present invention uses a method of modifying multiple sampling values (n sampling values) to embed synchronization information into the statistical mean of sampling values. 水印信号嵌入：采用量化混合域的低中频系数的方法。Embedding of Watermark Signals: A method using quantized low-IF coefficients in the mixed domain. 水印检测：是水印嵌入方法的逆过程。Watermark detection: It is the reverse process of the watermark embedding method. 2、根据权利要求1的同步码嵌入方法，其特征在于同步码的量化嵌入步骤如下：2, according to the synchronization code embedding method of claim 1, it is characterized in that the quantization embedding step of synchronization code is as follows: 根据时域掩蔽特性确定出同步码嵌入时的量化步长S₁，有

(

为向下取整函数)，s_a＝ |A|.According to the time-domain masking characteristics, the quantization step S ₁ when the synchronization code is embedded is determined, and there is

(

is the rounding down function), s _a = |A|. 其中，s_a＝ |A|表示原始数字音频A的绝对值均值(即峰值的均值)。Wherein, s _a = |A| represents the mean value of the absolute value of the original digital audio A (ie, the mean value of the peak value). 3、根据权利要求1的水印信号嵌入方法，其特征在于水印的嵌入步骤如下：3. The watermark signal embedding method according to claim 1, wherein the embedding steps of the watermark are as follows: (1)音频数据段(后面部分)的DWT变换。(1) DWT transformation of the audio data segment (later part). (2)考虑到小波变换域的近似分量能够有效抵御各种干扰，所以对近似分量进行DCT变换。(2) Considering that the approximate components in the wavelet transform domain can effectively resist various interferences, DCT transformation is performed on the approximate components. (3)为保证音频水印的稳健性及不可感知性，本发明根据人类听觉系统的频域掩蔽特性，采用量化策略将水印信息嵌入到DCT域的低中频部分。(3) In order to ensure the robustness and imperceptibility of the audio watermark, the present invention uses a quantization strategy to embed the watermark information into the low-intermediate frequency part of the DCT domain according to the frequency-domain masking characteristics of the human auditory system. (4)对嵌入水印后的音频数据段做逆DCT变换及逆DWT变换，并得到含有水印的音频数据段。(4) Perform inverse DCT transformation and inverse DWT transformation on the audio data segment embedded with the watermark, and obtain the audio data segment containing the watermark. (5)采用上面的步骤对其它数据段进行同步码与水印信息的循环嵌入。(5) Use the above steps to cyclically embed the synchronization code and watermark information on other data segments. 4、根据权利要求3水印嵌入的量化策略，其特征在于量化步长的选取，结合频域掩蔽特性自适应确定用于水印信息嵌入的量化步长S₂，具体步骤为：4. The quantization strategy for watermark embedding according to claim 3, characterized in that the selection of the quantization step size, combined with the frequency domain masking characteristics, adaptively determines the quantization step size _S2 for watermark information embedding, and the specific steps are: (1)对音频数据段A₂ ⁰进行DWT变换，并得到小波变换的第H级近似分量A₂ ^0H。(1) DWT transform is performed on the audio data segment A ₂ ⁰ , and the H-th order approximate component A ₂ ^0H of the wavelet transform is obtained. (2)对近似分量进行DCT变换，得到(2) Perform DCT transformation on the approximate components to get AA 22 00 HCHC == DCTDCT (( AA 22 00 Hh )) == {{ aa 22 00 (( tt )) HCHC ,, 00 &le;&le; tt << LL 22 22 Hh }} (3)量化步长S₂的自适应选取。由于水印信息已嵌入到DCT域的前M×N个系数内，故为了减少数字水印干扰，以确保水印提取过程能够有效计算出量化步长，本发明结合音频数据段内容利用DCT域的后M×N个系数自适应确定量化步长S₂。即(3) Adaptive selection of quantization step size _S2 . Since the watermark information has been embedded into the first M×N coefficients of the DCT domain, in order to reduce digital watermark interference and ensure that the watermark extraction process can effectively calculate the quantization step size, the present invention uses the last M coefficients of the DCT domain in combination with the content of the audio data segment The ×N coefficients adaptively determine the quantization step size S ₂ . Right now

(

为向下取整函数)， S b = | A ^ 2 0 HC | &OverBar; = &Sigma; t = M &times; N L 2 2 H - 1 | a 2 0 ( t ) HC | M &times; N

(

is the rounding down function), S b = | A ^ 2 0 HC | &OverBar; = &Sigma; t = m &times; N L 2 2 h - 1 | a 2 0 ( t ) HC | m &times; N 5、根据权利要求1数字水印的提取，其特征在于水印的提取步骤如下：5. According to the extraction of claim 1 digital watermark, it is characterized in that the extraction steps of watermark are as follows: (1)同步码检测。(1) Synchronization code detection. (2)按照水印嵌入过程量化步长的自适应选取的工作步骤，结合音频数据段的内容自适应确定出量化步长。(2) According to the working steps of adaptively selecting the quantization step size in the watermark embedding process, the quantization step size is adaptively determined in combination with the content of the audio data segment. (3)对音频数据段进行DWT变换，得到小波系数。(3) Perform DWT transformation on the audio data segment to obtain wavelet coefficients. (4)对小波系数的近似分量做DCT变换，并提取水印信息。(4) Perform DCT transformation on the approximate components of the wavelet coefficients, and extract the watermark information. (5)对所提取出的水印信息进行逆置乱加密，得到二值水印图像。(5) Perform inverse scrambling encryption on the extracted watermark information to obtain a binary watermark image. 6、根据权利要求3的循环嵌入策略，其特征在于：6. The loop embedding strategy according to claim 3, characterized in that: 由于本发明对同步码及水印信息采用了循环嵌入策略，故数字水印检测过程将提取出多个二值水印图像W_r ^*(r＝1，2，…，R)。为此，需进一步按照“多数原则”计算出最终的数字水印

即Since the present invention adopts a circular embedding strategy for the synchronization code and watermark information, a plurality of binary watermark images W _r ^* (r=1, 2, ..., R) will be extracted during the digital watermark detection process. For this reason, it is necessary to further calculate the final digital watermark according to the "majority principle"

Right now

(r＝1，2，…，R)

(r=1, 2, ..., R) 这里， w r * ( i , j ) &Element; W r * , w ^ * ( i , j ) &Element; W ^ * . here, w r * ( i , j ) &Element; W r * , w ^ * ( i , j ) &Element; W ^ * . 7、根据权利要求5的同步码检测，是指检测窗口(大小为L₁)所覆盖的音频数据段内是否含有所嵌入的同步码信息。其特征在于其步骤如下：7. The synchronization code detection according to claim 5, which refers to detecting whether the audio data segment covered by the window (the size is L ₁ ) contains the embedded synchronization code information. It is characterized in that its steps are as follows: (1)按照量化步长的自适应选取的工作步骤，自适应确定出量化步长S₁ ^*。(1) Adaptively determine the quantization step size S ₁ ^* according to the working steps of the adaptive selection of the quantization step size. (2)按照同步码嵌入的工作步骤，分别计算音频数据段(即窗口所覆盖部分)前n×m个采样值的平均值。(2) According to the working steps of synchronous code embedding, calculate the average value of the first n×m sampling values of the audio data segment (that is, the part covered by the window). (3)提取同步码信息F′。(3) Extract the synchronization code information F'. (4)为了有效避免假同步现象发生，本发明采用了通信领域的帧同步码逐位比较方法确定同步码，即如果所提取出的同步码F′和原同步码F完全相同，则认为已检测到所嵌入的同步码信息。(4) In order to effectively avoid the occurrence of false synchronization, the present invention adopts the bit-by-bit comparison method of the frame synchronization code in the communication field to determine the synchronization code, that is, if the extracted synchronization code F' is exactly the same as the original synchronization code F, then it is considered to have Embedded sync code information detected.