CN108010532A - Digital watermark detection method based on Multivariate Gaussian Profile - Google Patents
Digital watermark detection method based on Multivariate Gaussian Profile Download PDFInfo
- Publication number
- CN108010532A CN108010532A CN201711363551.4A CN201711363551A CN108010532A CN 108010532 A CN108010532 A CN 108010532A CN 201711363551 A CN201711363551 A CN 201711363551A CN 108010532 A CN108010532 A CN 108010532A
- Authority
- CN
- China
- Prior art keywords
- watermark
- coefficient
- swt
- represent
- embedded
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
- G10L19/018—Audio watermarking, i.e. embedding inaudible data in the audio signal
Abstract
The invention discloses a kind of digital audio frequency watermark detection method based on Multivariate Gaussian Profile, segment processing is carried out to original audio first, the higher SWT significant coefficient sections of local energy are constructed according to short-time energy computational methods, modified by multiplying property embedded mode to its coefficient, and subband to coefficient after modification and other subbands carry out inverse SWT and convert to obtain containing watermark word tone frequency；Then, using existing correlation between SWT high-frequency sub-bands, statistical modeling is carried out to SWT coefficients using Multivariate Gaussian Profile, and combines MLE methods and carries out parameter Estimation with Newton Raphson alternative manners；Finally, maximum likelihood detector extraction watermark information is gone out using log-likelihood ratio method construct.
Description
Technical field
The invention belongs to the copyright protection technology field of digital audio, it is related to the digital audio frequency watermark side based on statistical model
Method, more particularly to a kind of digital watermark detection method based on Multivariate Gaussian Profile.
Background technology
With the fast development of multimedia technology and the popularization of Internet technology, people obtain various digital multimedia resources
Mode become more and more convenient, so as to cause the behavior for illegally copying, distorting and propagating multimedia resource to become to be becoming increasingly rampant,
Therefore, how safely and effectively to store and transmit multimedia messages and also become focus of concern.Digital audio frequency watermark
Technology is come into being as the important branch of information security field, and providing one kind for solution audio-frequency information intellectual property protection has
The means of effect, have very big research and application value in the content authentication of digital audio and copyright protection field.
Digital Audio Watermarking Techniques are that watermark information is embedded into audio using data embedding strategy, it is hidden in original
It is not detectable in beginning voice data, the integrality of audio had not only been remained, but also the legal protection to audio proprietary owner is realized,
Digital watermark can be still utilized to prove the ownership issue of audio production after digital audio is attacked or is decrypted.Closely
Over a little years, Digital Audio Watermarking Techniques have the development advanced by leaps and bounds, and many outstanding Audio Watermarking Algorithms occur in succession, resonable
All it is extremely important in terms of by research and practical application, but still has that many shortcomings are urgently to be resolved hurrily, wherein, robust
Property and the equilibrium problem between sentience is not still a problem in the field.
Transform domain audio-frequency water mark method based on statistical model can optimize not sentience and robustness at the same time, convert
Embedded watermark can improve the robustness of watermark on domain coefficient, and multiplying property embedded mobile GIS can make the embedment strength of watermark with load
Body signal strength changes in proportion, therefore can effectively solve the problem that robustness with that can not feel with reference to multiplying property embedding grammar in the transform domain as illustrated
Intellectual equilibrium problem between the two, receives the extensive concern of researcher.Method of detecting watermarks based on transform domain statistical modeling
It is proposed successively, mainly include two classes：A kind of method is to be detected whether in test side there are watermark, and another kind of method is can
Specific watermark information is extracted in test side, it is clear that second method is more with practical value.However, based on statistical model
Water mark method remains lower following deficiency：First, the distribution that statistical modeling is selected at present is cell distribution, just with scale
Correlation is without fully taking into account correlation between scale；Second, when being modeled to coefficient in transform domain, not to the distributed mode of selection
Whether type is analysed in depth, and optimal to the transform domain of selection without established model is fully proved；3rd, most base
It is image watermark direction in the digital watermark method of statistical modeling, and does not have too many concern in audio frequency watermark and grind
Study carefully.
The content of the invention
The present invention is to solve the above-mentioned technical problem present in the prior art, it is proposed that one kind is high based on Multivariate
The digital watermark detection method of this distribution.
The present invention technical solution be：A kind of digital watermark detection method based on Multivariate Gaussian Profile, bag
Include watermark insertion and watermark extracting, it is characterised in that：
Agreement：Represent host audio signal；Watermark length is represented,；Represent per segment length
Spend and beAudio,The second scale high-frequency sub-band is represented,Represent the 3rd scale high-frequency sub-band；Represent length of window；Represent the home position of important SWT coefficients；RepresentSignificant coefficient section,
RepresentSignificant coefficient section；Represent collision matrix；Represent form parameter；Represent scale parameter；Represent dimension；
The watermark insertion carries out as follows：
A. initial setting up
Obtain host audio signalAnd Initialize installation；
B. watermark is embedded in
B.1 it is rightSegment processing is carried out, is per segment length：；
B.2 it is rightThree-level SWT conversion is carried out, is obtainedWith, calculateThe short-time energy of each SWT coefficients and according to big
It is small to do descending arrangement：
,
Wherein,；
B.3 the coefficient near the end of audio end and adjacent excessively near coefficient are filtered out, before selectionA important SWT systems
Number is recorded after choosing as watermark insertion initial positionA coefficient is, willThe same conduct in position of embedded watermarkWatermark embedded location；
B.4 two kinds of multiplying property intensity function of embedding are constructed to be used to be embedded in watermark bit " 1 " or " 0 "：
,；
B.5 changeWithMiddle coefficient of correspondence obtainsWith：
；
B.6 willWithInverse SWT is with reference to other subbands, obtains containing watermark word tone frequency；
The watermark extracting carries out as follows：
C. coefficient Multivariate Gaussian Profile statistical modeling
C.1 important SWT coefficients are determined according to record, determines watermark SWT coefficient segments to be extracted；
C.2 the noise standard deviation of SWT subbands is estimated using median method：
；
C.3 tried to achieve respectively by numerical methodInverse function；
C.4 the Multivariate in the case where two kinds of embedded watermark " 1 " and embedded watermark " 0 " are assumed in each coefficient segments is calculated respectively
Gaussian Profile probabilityWith：
,
,,,
Wherein,,,RepresentThe not coefficient of Noise,
RepresentThe not coefficient of Noise；
D. construct maximum likelihood detector and carry out watermark extracting
D.1 to containing watermarked audioSegment processing is carried out, it is rightThree-level SWT conversion is carried out, is obtainedWith；
D.2 it is rightWithCarry out initialization process：
；
D.3 utilizeWithIt is first right according to maximum likelihood method as parameter Estimation sampleEstimated, in conjunction with
Newton-Raphson method pairEstimated, finally obtained according to estimationWithIt is rightEstimation：
,,
,
Wherein,,It is double gamma functions；
D.4 maximum likelihood detector is constructed, extracts the specific watermark information of each watermark coefficient section to be extracted：
；
D.5 optimal watermark sequence is obtained using Voting principle.
The present invention carries out segment processing to original audio first, and it is higher to construct local energy according to short-time energy computational methods
SWT significant coefficient sections, modified by multiplying property embedded mode to its coefficient, and subband to coefficient after modification and other sons
Band carries out inverse SWT and converts to obtain containing watermark word tone frequency；Then, using existing correlation between SWT high-frequency sub-bands, using more
First generalized Gaussian distribution carries out SWT coefficients statistical modeling, and combines MLE methods and carried out with Newton-Raphson alternative manners
Parameter Estimation；Finally, maximum likelihood detector extraction watermark information is gone out using log-likelihood ratio method construct.Experimental result table
Bright, method of the invention is effectively improved detection essence due to constructing more accurate model using Multivariate Gaussian Profile
Degree, while also maintain not well balanced between sentience and robustness.
Compared with prior art, the invention has the advantages that：
First, SWT coefficients are modeled using Multivariate Gaussian Profile, take full advantage of father and son's correlation and ruler between scale
Brother dependencies in degree, can more precisely fit the characteristic distributions of coefficient；
Second, parameter Estimation is carried out with reference to MLE methods and Newton-Raphson alternative manners, further increases the inspection of model
Survey precision.
Brief description of the drawings
Fig. 1 is original audio, ripple containing watermarked audio and difference audio in wedding.wav audios of the embodiment of the present invention
Shape figure.
Fig. 2 is the robustness test result figure of conventional attack of the embodiment of the present invention.
Fig. 3 is the robustness test result figure of desynchronization attack of the embodiment of the present invention.
Fig. 4 is that the watermark of the embodiment of the present invention is embedded in flow chart.
Fig. 5 is the watermark extracting flow chart of the embodiment of the present invention.
Embodiment
The method of the present invention mainly includes three phases：Watermark is embedded, coefficient Multivariate Gaussian Profile statistical modeling and
Construct maximum likelihood detector and carry out watermark extracting.
Agreement：Represent host audio signal；Watermark length is represented,；Represent every
Segment length isAudio,The second scale high-frequency sub-band is represented,Represent the 3rd scale high-frequency sub-band；Represent window length
Degree；Represent the home position of important SWT coefficients；RepresentSignificant coefficient section,RepresentSignificant coefficient section；Represent collision matrix；Represent form parameter；Represent ruler
Spend parameter；Represent dimension；
Watermark insertion as shown in figure 4, carry out in accordance with the following steps：
A. initial setting up
Obtain host audio signalAnd Initialize installation；
B. watermark is embedded in
B.1 it is rightSegment processing is carried out, is per segment length：；
B.2 it is rightThree-level SWT conversion is carried out, is obtainedWith, calculateThe short-time energy of each SWT coefficients and according to big
It is small to do descending arrangement：
,
Wherein,；
B.3 the coefficient near the end of audio end and adjacent excessively near coefficient are filtered out, before selectionA important SWT systems
Number is recorded after choosing as watermark insertion initial positionA coefficient is, willThe same conduct in position of embedded watermarkWatermark embedded location；
B.4 two kinds of multiplying property intensity function of embedding are constructed to be used to be embedded in watermark bit " 1 " or " 0 "：
,；
B.5 changeWithMiddle coefficient of correspondence obtainsWith：
；
B.6 willWithInverse SWT is with reference to other subbands, obtains containing watermark word tone frequency；
Watermark extracting as shown in figure 5, carry out in accordance with the following steps：
C. coefficient Multivariate Gaussian Profile statistical modeling
C.1 important SWT coefficients are determined according to record, determines watermark SWT coefficient segments to be extracted；
C.2 the noise standard deviation of SWT subbands is estimated using median method：
；
C.3 tried to achieve respectively by numerical methodInverse function；
C.4 the Multivariate in the case where two kinds of embedded watermark " 1 " and embedded watermark " 0 " are assumed in each coefficient segments is calculated respectively
Gaussian Profile probabilityWith：
,
,,,
Wherein,,,RepresentThe not coefficient of Noise,
RepresentThe not coefficient of Noise；
D. construct maximum likelihood detector and carry out watermark extracting
D.1 to containing watermarked audioSegment processing is carried out, it is rightThree-level SWT conversion is carried out, is obtainedWith；
D.2 it is rightWithCarry out initialization process：
；
D.3 utilizeWithIt is first right according to maximum likelihood method as parameter Estimation sampleEstimated, in conjunction with
Newton-Raphson method pairEstimated, finally obtained according to estimationWithIt is rightEstimation：
,,
,
Wherein,,It is double gamma functions；
D.4 maximum likelihood detector is constructed, extracts the specific watermark information of each watermark coefficient section to be extracted：
；
D.5 optimal watermark sequence is obtained using Voting principle.
Experiment test and parameter setting：
Experiment performs on MATLAB R2011a platforms, and audio carrier is that sample frequency is 41kHz, and resolution ratio is 16 bits, long
Spend the monaural digital audio signal for 20 seconds.
Fig. 1 is original audio, ripple containing watermarked audio and difference audio in wedding.wav audios of the embodiment of the present invention
Shape figure.
Fig. 2 is the robustness test result figure of conventional attack of the embodiment of the present invention.
Fig. 3 is the robustness test result figure of desynchronization attack of the embodiment of the present invention.
Test result indicates that method of the invention is using Multivariate Gaussian Profile due to constructing more accurate model,
Accuracy of detection is effectively improved, while is also maintained not well balanced between sentience and robustness.
Claims (1)
1. a kind of digital watermark detection method based on Multivariate Gaussian Profile, including watermark insertion and watermark extracting, it is special
Sign is：
Agreement：Represent host audio signal；Watermark length is represented,；Represent per segment length
Spend and beAudio,The second scale high-frequency sub-band is represented,Represent the 3rd scale high-frequency sub-band；Represent length of window；Represent the home position of important SWT coefficients；RepresentSignificant coefficient section,RepresentSignificant coefficient section；Represent collision matrix；Represent form parameter；Represent ruler
Spend parameter；Represent dimension；
The watermark insertion carries out as follows：
A. initial setting up
Obtain host audio signalAnd Initialize installation；
B. watermark is embedded in
B.1 it is rightSegment processing is carried out, is per segment length：；
B.2 it is rightThree-level SWT conversion is carried out, is obtainedWith, calculateThe short-time energy of each SWT coefficients and according to big
It is small to do descending arrangement：
,
Wherein,；
B.3 the coefficient near the end of audio end and adjacent excessively near coefficient are filtered out, before selectionA important SWT systems
Number is recorded after choosing as watermark insertion initial positionA coefficient is, willThe same conduct in position of embedded watermarkWatermark embedded location；
B.4 two kinds of multiplying property intensity function of embedding are constructed to be used to be embedded in watermark bit " 1 " or " 0 "：
,；
B.5 changeWithMiddle coefficient of correspondence obtainsWith：
；
B.6 willWithInverse SWT is with reference to other subbands, obtains containing watermark word tone frequency；
The watermark extracting carries out as follows：
C. coefficient Multivariate Gaussian Profile statistical modeling
C.1 important SWT coefficients are determined according to record, determines watermark SWT coefficient segments to be extracted；
C.2 the noise standard deviation of SWT subbands is estimated using median method：
；
C.3 tried to achieve respectively by numerical methodInverse function；
C.4 the Multivariate in the case where two kinds of embedded watermark " 1 " and embedded watermark " 0 " are assumed in each coefficient segments is calculated respectively
Gaussian Profile probabilityWith：
,
,,,
Wherein,,,RepresentThe not coefficient of Noise,Table
ShowThe not coefficient of Noise；
D. construct maximum likelihood detector and carry out watermark extracting
D.1 to containing watermarked audioSegment processing is carried out, it is rightThree-level SWT conversion is carried out, is obtainedWith；
D.2 it is rightWithCarry out initialization process：
；
D.3 utilizeWithIt is first right according to maximum likelihood method as parameter Estimation sampleEstimated, in conjunction with
Newton-Raphson method pairEstimated, finally obtained according to estimationWithIt is rightEstimation：
,,
,
Wherein,,It is double gamma functions；
D.4 maximum likelihood detector is constructed, extracts the specific watermark information of each watermark coefficient section to be extracted：
；
D.5 optimal watermark sequence is obtained using Voting principle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711363551.4A CN108010532B (en) | 2017-12-18 | 2017-12-18 | Digital watermark detection method based on multivariate generalized Gaussian distribution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711363551.4A CN108010532B (en) | 2017-12-18 | 2017-12-18 | Digital watermark detection method based on multivariate generalized Gaussian distribution |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108010532A true CN108010532A (en) | 2018-05-08 |
| CN108010532B CN108010532B (en) | 2021-09-24 |
Family
ID=62059793
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711363551.4A Active CN108010532B (en) | 2017-12-18 | 2017-12-18 | Digital watermark detection method based on multivariate generalized Gaussian distribution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108010532B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2005006326A2 (en) * | 2003-07-11 | 2005-01-20 | Koninklijke Philips Electronics N.V. | Watermark embedding and detection |
| CN101039371A (en) * | 2006-03-18 | 2007-09-19 | 辽宁师范大学 | Novel method of digital watermarking for protecting literary property of presswork |
| CN102074237A (en) * | 2010-11-30 | 2011-05-25 | 辽宁师范大学 | Digital audio watermarking method based on invariant characteristic of histogram |
| CN102147912A (en) * | 2011-03-30 | 2011-08-10 | 北京航空航天大学 | Adaptive difference expansion-based reversible image watermarking method |
-
2017
- 2017-12-18 CN CN201711363551.4A patent/CN108010532B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2005006326A2 (en) * | 2003-07-11 | 2005-01-20 | Koninklijke Philips Electronics N.V. | Watermark embedding and detection |
| CN101039371A (en) * | 2006-03-18 | 2007-09-19 | 辽宁师范大学 | Novel method of digital watermarking for protecting literary property of presswork |
| CN102074237A (en) * | 2010-11-30 | 2011-05-25 | 辽宁师范大学 | Digital audio watermarking method based on invariant characteristic of histogram |
| CN102147912A (en) * | 2011-03-30 | 2011-08-10 | 北京航空航天大学 | Adaptive difference expansion-based reversible image watermarking method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108010532B (en) | 2021-09-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cannons et al. | Design and statistical analysis of a hash-aided image watermarking system | |
| Bhowal et al. | A GA based audio steganography with enhanced security | |
| CN101556686B (en) | Robust image watermark method based on local Tchebichef moment | |
| CN100559466C (en) | A kind of audio-frequency watermark processing method of anti-DA/AD conversion | |
| CN108198563A (en) | A kind of Multifunctional audio guard method of digital copyright protection and content authentication | |
| CN102496367A (en) | Audio blind watermark algorithm based on combination of discrete wavelet transformation (DWT), discrete cosine transformation (DCT) and singular value decomposition (SVD) | |
| CN105632506A (en) | Robust digital audio watermark embedding and detection method based on polar harmonic transform | |
| CN103456308B (en) | A kind of recoverable ciphertext domain voice content authentication method | |
| CN102867513B (en) | Pseudo-Zernike moment based voice content authentication method | |
| CN103050120A (en) | High-capacity digital audio reversible watermark processing method | |
| CN103886864A (en) | Method for detecting quantitatively-embedded secret information for DWT domain fidelity compensation | |
| CN106205627B (en) | Digital audio reversible water mark algorithm based on side information prediction and histogram translation | |
| CN101847411A (en) | MIDI (Musical Instrument Digital Interface) audio hidden information analysis method and device | |
| Wu et al. | Robust and blind audio watermarking algorithm in dual domain for overcoming synchronization attacks | |
| CN108010532A (en) | Digital watermark detection method based on Multivariate Gaussian Profile | |
| Liu et al. | Pseudo-zernike moments-based audio content authentication algorithm robust against feature-analysed substitution attack | |
| CN103745725B (en) | A kind of audio frequency watermark embedding grammar based on constant Q transform | |
| CN107910010A (en) | Digital watermark detection method based on multi-parameter Weibull statistical modelings | |
| Chowdhury | A Robust Audio Watermarking In Cepstrum Domain Composed Of Sample's Relation Dependent Embedding And Computationally Simple Extraction Phase | |
| Wu et al. | BCH Code-Based Robust Audio Watermarking Algorithm in the DWT Domain | |
| Chang et al. | Image authentication with tampering localization based on watermark embedding in wavelet domain | |
| CN103871015B (en) | Watermark method aiming at computer graphic image | |
| KR20060123280A (en) | Method and apparatus for detecting a watermark in a signal | |
| Yue et al. | Rights protection for trajectory streams | |
| CN105374361A (en) | Audio frequency watermark algorithm based on chaos system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |