CN115831131B - Audio watermark embedding and extracting method based on deep learning - Google Patents
Audio watermark embedding and extracting method based on deep learning Download PDFInfo
- Publication number
- CN115831131B CN115831131B CN202310056386.7A CN202310056386A CN115831131B CN 115831131 B CN115831131 B CN 115831131B CN 202310056386 A CN202310056386 A CN 202310056386A CN 115831131 B CN115831131 B CN 115831131B
- Authority
- CN
- China
- Prior art keywords
- audio
- watermark
- decoder
- encoder
- steps
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 64
- 238000013135 deep learning Methods 0.000 title claims abstract description 11
- 238000001914 filtration Methods 0.000 claims abstract description 14
- 230000002708 enhancing effect Effects 0.000 claims abstract description 4
- 238000000605 extraction Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 8
- 230000004044 response Effects 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 4
- 230000007613 environmental effect Effects 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims 1
- 238000013518 transcription Methods 0.000 abstract description 6
- 230000035897 transcription Effects 0.000 abstract description 6
- 238000012952 Resampling Methods 0.000 abstract description 4
- 230000006835 compression Effects 0.000 abstract description 3
- 238000007906 compression Methods 0.000 abstract description 3
- 238000013139 quantization Methods 0.000 abstract description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000003153 chemical reaction reagent Substances 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses a method for embedding and extracting an audio watermark based on deep learning, which comprises the steps of firstly embedding watermark information into carrier audio by using an encoder to obtain the audio containing the watermark; inserting a distortion layer between the encoder and the decoder for enhancing the robustness of the audio reproduction process before inputting the watermarked audio to the decoder; the audio containing the watermark after being subjected to the distortion layer is input to a decoder, and watermark information therein is extracted by the decoder. According to the method, after the watermark is embedded into the target audio, the watermark information in the target audio can be still extracted after the target audio is subjected to the distortion of noise adding, filtering, compression, resampling, re-quantization, transcription and the like, so that the purposes of audio leakage tracing and copyright protection are achieved.
Description
Technical Field
The invention relates to the technical field of digital watermarking, in particular to a method for embedding and extracting an audio watermark based on deep learning.
Background
Digital watermarking has been widely studied for many years as an effective method of tracking leakage sources and copyright protection. Two of the most important properties that an audio watermark should meet are fidelity, which ensures proper use of the watermarked audio, and robustness, which ensures that the embedded watermark can be extracted without loss even if the audio is subject to distortion (MPEG encoding, noise addition, audio re-recording, etc.). Most conventional audio watermarking methods focus on the robustness of digital distortion in electronic channels, since most audio copying occurs in digital channels. However, with miniaturization of recording devices, audio Re-recording (AR) has become a more convenient and efficient way to copy Audio, when Audio is used as a carrier for information transmission, for many important confidential Audio information (litigation recording, evidence-taking Audio) and paid Audio piracy (network classroom Audio, film piracy), since the recording can effectively retain Audio content and significantly destroy embedded watermark signals, an attacker can easily and implicitly implement Audio content information stealing by using the means of recording, and has difficulty in leaving evidence, as shown in fig. 1 is a schematic diagram of the recording operation on leaking information in the prior art, so how to keep robustness in complex scenes is one of the greatest challenges for Audio watermarking, and ensuring robustness to the recording becomes the current stage of Audio watermarking.
At present, the research field of audio watermarking is mainly based on traditional mathematical algorithms, in an attempt to find the characteristics which are unchanged before and after distortion to embed the watermark, most of the used characteristics are in a transform domain, for example, the transform domain characteristics of audio are obtained by adopting audio frequency domain conversion methods such as Discrete Cosine Transform (DCT), discrete Wavelet Transform (DWT) and Fast Fourier Transform (FFT). However, because of the complexity of the transcription process itself, quantitatively and finely analyzing the distortion and finding robust and invariant features in this process is a very difficult task to achieve, and none of the prior art algorithms is well resistant to transcription distortion.
Disclosure of Invention
The invention aims to provide a method for embedding and extracting an audio watermark based on deep learning, which can be used for extracting watermark information in target audio after the target audio is subjected to noise adding, filtering, compressing, resampling, re-quantizing, copying and the like after the watermark is embedded into the target audio, so that the purposes of audio leakage tracing and copyright protection are realized.
The invention aims at realizing the following technical scheme:
a method of deep learning based audio watermark embedding extraction, the method comprising:
step 1, embedding watermark information into carrier audio by using an encoder to obtain audio containing watermark;
step 2, before inputting the audio containing the watermark into the decoder, inserting a distortion layer between the encoder and the decoder for enhancing the robustness of the audio reproduction process;
and 3, inputting the audio containing the watermark after being subjected to the distortion layer into a decoder, and extracting watermark information in the audio by the decoder.
According to the technical scheme provided by the invention, after the watermark is embedded into the target audio, the watermark information in the target audio can be still extracted after the target audio is subjected to the distortion of noise adding, filtering, compression, resampling, re-quantization, transcription and the like, so that the purposes of audio leakage tracing and copyright protection are realized.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram of a dubbing operation for leaking information in the prior art.
Fig. 2 is a schematic flow chart of a method for audio watermark embedding and extraction based on deep learning according to an embodiment of the present invention.
Detailed Description
The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments of the present invention, and this is not limiting to the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to fall within the scope of the invention.
Embodiments of the present invention will be described in further detail below with reference to the accompanying drawings. What is not described in detail in the embodiments of the present invention belongs to the prior art known to those skilled in the art. The specific conditions are not noted in the examples of the present invention and are carried out according to the conditions conventional in the art or suggested by the manufacturer. The reagents or apparatus used in the examples of the present invention were conventional products commercially available without the manufacturer's knowledge.
Fig. 2 is a schematic flow chart of a method for audio watermark embedding and extraction based on deep learning according to an embodiment of the present invention, where the method includes:
step 1, embedding watermark information into carrier audio by using an encoder to obtain audio containing watermark;
in this step, in step 1, use is made of
To represent a length of +.>
Mono original carrier audio of (a); the original carrier audio is first of all +.>
Transferring to frequency domain to obtain corresponding approximation coefficients +.>
And detail coefficient->
The method comprises the following steps:
wherein the approximation coefficients
And detail coefficient->
Is the length of the original carrier audio +.>
Half of (i.e.)>
；
Inspired by the traditional audio watermarking, watermark information is obtained
Embedded in the original carrier audio->
In the low frequency of (2), i.e. using approximation coefficients +.>
As carrier of watermark information while preserving detail coefficients +.>
For subsequent audio reconstruction;
the encoder is used for watermark information
Embedded in->
In, as shown in fig. 2, the encoder En generates a residual R and marks it further to +.>
Thereby generating approximation coefficients +.>
The method comprises the following steps:
wherein 
Is an intensity factor, default set to 1; en () represents encoder processing.
In addition, to meetFidelity requirements to approximate coefficients containing watermarks
As much as possible->
Keeping in agreement, a basic penalty is introduced in the training of the encoder>
Using mean square error->
As->
The method comprises the following steps:
wherein i represents an index number;
representing the i-th approximation coefficient; />
Representing the ith watermark-containing approximation coefficient;
to further improve fidelity and minimize
and />
The difference of the domains between the two is introduced with an additional discriminator D for creating a countermeasure training with the encoder against losses ∈ ->
For better embedding of watermark information by the encoder, making the discriminator indistinguishable +.>
and />
Thereby minimizing +.>
and />
The domain gap between, namely:
wherein D () represents the discriminator process.
Step 2, before inputting the audio containing the watermark into the decoder, inserting a distortion layer between the encoder and the decoder for enhancing the robustness of the audio reproduction process;
in this step, it is necessary to make the distortion layer tiny, which can prevent gradient interruption in the end-to-end learning process, whereas the transcription process is a complex non-differential process, so the inserted distortion layer is set as a differential audio re-recording operation DAR, including ambient reverberation, band-pass filtering, and gaussian noise; in a specific implementation, in order to realize robustness to the dubbing, the dubbing process is analyzed from the influence of sound propagation in the air and the processing of a microphone and a loudspeaker; from the analysis, this example models the transcription distortion finely by several differential operations (ambient reverberation, band-pass filtering, and gaussian noise) and uses these operations as a distortion layer with the proposed framework;
since DAR is a process running in the time domain, it cannot be directly applied to approximation coefficients containing watermarks
The approximation coefficients containing the watermark are therefore scaled using inverse DWT, IDWT (Inverse Discrete Wavelet Transform)
And the corresponding detail coefficients->
Transform back to watermarked audio->
The method comprises the following steps:
the environmental reverberation specifically includes:
the impulse response is the response of the environment upon receipt of a brief input signal describing the acoustic properties of the environment, in particular the spatial reverberation behavior, the impulse response reproducing the reverberation in the environment by convolution, collecting different base impulse responses from different microphones, room environments and loudspeakers to form a set
Given target audio +.>
From the collection->
A basic impulse response is selected randomly>
And by means of the collection->
Audio of the object->
Convolving->
Operate to simulate ambient reverberation ER (), i.e.:
the band-pass filtering specifically comprises the following steps:
because of the limited frequency band of human hearing, the widely used normal frequency range is 500Hz to 2000 Hz, based on which the conventional speaker does not play audio in too high or too low a frequency band, while the microphone also processes the played audio, typically cuts off the frequency band outside the normal range, to reduce noise, i.e. a basic de-noising process, thus, to simulate distortion caused by the inherent characteristics of the speaker and microphone, the audio containing the watermark
Applying frequency band-pass filtering
Operation, given target Audio->
Execute +.>
：
wherein 
and />
Representing low-pass filtering and high-pass filtering, respectively; />
and />
Representation->
and />
A corresponding threshold;
the Gaussian noise is specifically:
in addition to the two components, random noise caused by uncertainty factors in the reproduction process is simulated by introducing Gaussian noise, which is an additive noise widely used in current automatic speech recognition schemes to enhance robustness to random environmental noise, in particular by directly superimposing Gaussian noise
In the target audio +.>
Upper implementation of Add Gaussian noise operation->
The method comprises the following steps:
wherein ,
representing gaussian noise; />
Mean value 0, variance ++>
Is a gaussian distribution of (c).
And 3, inputting the audio containing the watermark after being subjected to the distortion layer into a decoder, and extracting watermark information in the audio by the decoder.
In this step, the processing procedure of the distortion layer DAR is as follows:
for watermarked audio
Ultimately obtaining a watermark-containing after being subjected to a distortion layer DAR processingAudio frequency
The method comprises the following steps:
obtaining the discrete wavelet transform DWT
Corresponding approximation coefficients>
And detail coefficient
And approximate coefficient +.>
Inputting into decoder De, extracting watermark from the decoder De>
The method comprises the following steps:
where De (-) represents the decoder processing.
In specific implementation, watermark loss is further introduced
I.e. watermark information->
And watermark extracted by decoder>
The mean square error loss MSE (Mean Square Error) between:
when binary watermarking is employed
Rather than +.>
This is more advantageous for watermark embedding and extraction of the model, in which case the watermark-containing audio is +.>
Watermark extracted by decoder>
The distribution of (2) should be as close as possible to-1 and 1; for watermarking-free audio, the watermark extracted by the decoder is +.>
The distribution should be close to 0, which facilitates constrained operation based on MSE.
It is noted that what is not described in detail in the embodiments of the present invention belongs to the prior art known to those skilled in the art.
In order to illustrate the effect of the scheme of the embodiment of the present invention, the following detailed description is made through experiments:
1) Fidelity testing
First, the fidelity of the method described herein is compared to the existing baseline method, as shown in table 1, which achieves an SNR of 25.86, which is superior to the existing baseline method.
Quantitative comparison of Table 1 with baseline method
| Index (I) | The method | Baseline 1 | Baseline 2 |
| SNR(dB) | 25.86 | 25.81 | 24.94 |
| ACC(%) | 99.18 | 77.09 | 56.0 |
2) Robustness test for audio reproduction
The robustness of audio re-recordings was compared in this experiment and quantitative results are provided in table 2, with significant fidelity over the baseline method (over 20% and 40%, respectively). In addition to the default distance (5 cm), a control comparison was further made with the baseline method under different conditions. As shown in table 2, the method of the embodiment of the present application performs better over a long distance, and as the distance increases, the robustness to dubbing decreases correspondingly, but is still acceptable (both are above 90%).
Table 2 robustness comparison of dubbing at different distances
| Distance (cm) | 5 | 20 | 50 | 100 |
| The method | 99.18 | 98.55 | 93.40 | 92.68 |
| Baseline 1 | 77.09 | 82.64 | 74.76 | 66.02 |
3) Robustness testing against other common distortions
To more fully compare robustness, further evaluations were made under other common distortions in the digital transmission process, namely gaussian noise at different signal-to-noise ratios (20 dB, 30 dB, 40 dB, 50 dB), MP3 compression (64 kbps, 128 kbps), bandpass (1 kHz high pass, 4 kHz), resampling, clipping, amplitude modification, re-quantization and median filtering. As shown in table 3, the method employed in the present application is robust against all types of distortion.
Table 3 robustness to other common distortions, default/enhanced ACC
The experimental results show that: the method of the embodiment of the invention can automatically realize the embedding of the audio watermark and the robust extraction under various distortions, and can achieve higher extraction accuracy than the existing method.
In summary, after watermark information is embedded in Audio, the method of the embodiment of the invention can realize robust extraction of the watermark under common Audio processing distortion scene, watermark attack scene and Audio Re-recording (AR) distortion, thereby achieving the purposes of disclosure tracing and copyright protection.
In addition, it will be understood by those skilled in the art that all or part of the steps in implementing the methods of the above embodiments may be implemented by a program to instruct related hardware, and the corresponding program may be stored in a computer readable storage medium, where the storage medium may be a read only memory, a magnetic disk or an optical disk, etc.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions easily contemplated by those skilled in the art within the scope of the present invention should be included in the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims. The information disclosed in the background section herein is only for enhancement of understanding of the general background of the invention and is not to be taken as an admission or any form of suggestion that this information forms the prior art already known to those of ordinary skill in the art.
Claims (3)
1. A method for deep learning-based audio watermark embedding and extraction, the method comprising:
step 1, embedding watermark information into carrier audio by using an encoder to obtain audio containing watermark;
in step 1, using
To represent a length of +.>
Mono original carrier audio of (a); first by means of a differentiable discrete wavelet transform +.>
Original carrier audio +.>
Transferring to frequency domain to obtain corresponding approximation coefficients +.>
And detail coefficient->
The method comprises the following steps:
wherein the approximation coefficients
And detail coefficient->
Is the length of the original carrier audio +.>
Half of (i.e.)>
；
Watermark information
Embedded in the original carrier audio->
In the low frequency of (2), i.e. using approximation coefficients +.>
As carrier of watermark information while preserving detail coefficients +.>
For subsequent audio reconstruction;
the encoder is used for watermark information
Embedded in->
Specifically: the encoder generates a residual and marks it further +.>
Thereby generating approximation coefficients +.>
The method comprises the following steps:
wherein 
Is an intensity factor, default set to 1; en () represents encoder processing;
in order to make the approximation coefficient containing watermark
As much as possible->
Keeping in agreement, a basic penalty is introduced in the training of the encoder>
Using mean square error->
As->
The method comprises the following steps:
wherein i represents an index number;
representing the i-th approximation coefficient; />
Representing the ith watermark-containing approximation coefficient;
to minimize
and />
The difference of the domains is introduced with an additional discriminator for creating an countermeasure training with the encoder, a countermeasure loss ∈ ->
For better embedding of watermark information by the encoder, making the discriminator indistinguishable +.>
and />
Thereby minimizing +.>
and />
The domain gap between, namely:
wherein D () represents a discriminator process;
step 2, before inputting the audio containing the watermark into the decoder, inserting a distortion layer between the encoder and the decoder for enhancing the robustness of the audio reproduction process;
in step 2, the inserted distortion layer is a differential audio re-recording operation DAR, including ambient reverberation, band-pass filtering, and gaussian noise;
since DAR is a process running in the time domain, it cannot be directly applied to approximation coefficients containing watermarks
Therefore, inverse DWT, i.e. IDWT will contain the approximation coefficients of the watermark +.>
And the corresponding detail coefficients->
Transforming back to watermarked audio
The method comprises the following steps:
the environmental reverberation specifically includes:
the impulse response reproduces reverberation in the environment by convolution, collecting different base impulse responses from different microphones, room environments and speakers to form a set
Given target audio +.>
From the collection->
A basic impulse response is selected randomly>
And by means of the collection->
Audio of the object->
Convolving->
Operate to simulate ambient reverberation ER (), i.e.:
the band-pass filtering specifically comprises the following steps:
to simulate distortion caused by inherent characteristics of speakers and microphones, audio with watermark is recorded
Applying frequency band-pass filtering->
Operation, given target Audio->
Execute +.>
：
wherein 
and />
Representing low-pass filtering and high-pass filtering, respectively; />
and />
Representation->
and />
A corresponding threshold;
the Gaussian noise is specifically:
simulating random noise caused by uncertainty factors during reproduction by introducing Gaussian noise, in particular by directly superimposing Gaussian noise
In the target audio +.>
Upper implementation of Add Gaussian noise operation->
The method comprises the following steps:
wherein ,
representing gaussian noise; />
Mean value 0, variance ++>
Is a gaussian distribution of (c);
and 3, inputting the audio containing the watermark after being subjected to the distortion layer into a decoder, and extracting watermark information in the audio by the decoder.
2. The method for deep learning based audio watermark embedding extraction of claim 1, wherein in step 3, the distortion layer DAR is processed as follows:
for watermarked audio
Ultimately obtaining audio containing a watermark after being subjected to a distortion layer DAR processing
The method comprises the following steps:
obtaining the discrete wavelet transform DWT
Corresponding approximation coefficients>
And detail coefficient
And approximate coefficient +.>
Inputting into decoder De, extracting watermark from the decoder De>
The method comprises the following steps:
where De (-) represents the decoder processing.
3. The method for deep learning based audio watermark embedding extraction as claimed in claim 2, wherein in step 3, watermark loss is further introduced
I.e. watermark information->
And watermark extracted by decoder>
The mean square error between MSE, namely:
when binary watermarking is employed
Rather than +.>
In this case, for watermarked audio
Watermark extracted by decoder>
The distribution of (2) should be as close as possible to-1 and 1; for watermarking-free audio, the watermark extracted by the decoder is +.>
The distribution of (2) should be close to 0.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310056386.7A CN115831131B (en) | 2023-01-15 | 2023-01-15 | Audio watermark embedding and extracting method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310056386.7A CN115831131B (en) | 2023-01-15 | 2023-01-15 | Audio watermark embedding and extracting method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115831131A CN115831131A (en) | 2023-03-21 |
| CN115831131B true CN115831131B (en) | 2023-06-16 |
Family
ID=85520711
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310056386.7A Active CN115831131B (en) | 2023-01-15 | 2023-01-15 | Audio watermark embedding and extracting method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115831131B (en) |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE19852805A1 (en) * | 1998-11-15 | 2000-05-18 | Florian M Koenig | Telephone with improved speech understanding, several microphones, and special speech signal processing |
| AU2001231109A1 (en) * | 2000-01-24 | 2001-07-31 | Businger, Peter A. | Transform domain allocation for multimedia watermarking |
| CN101115124B (en) * | 2006-07-26 | 2012-04-18 | 日电（中国）有限公司 | Method and apparatus for identifying media program based on audio watermark |
| CN101290772B (en) * | 2008-03-27 | 2011-06-01 | 上海交通大学 | Embedding and extracting method for audio zero water mark based on vector quantization of coefficient of mixed domain |
| JP2011197619A (en) * | 2010-03-18 | 2011-10-06 | Yasuo Sano | Perception improving means for acoustic signal using electronic watermark |
| EP2431970A1 (en) * | 2010-09-21 | 2012-03-21 | Fraunhofer-Gesellschaft zur Förderung der Angewandten Forschung e.V. | Watermark generator, watermark decoder, method for providing a watermarked signal based on discrete valued data and method for providing discrete valued data in dependence on a watermarked signal |
| CN102074237B (en) * | 2010-11-30 | 2012-07-04 | 辽宁师范大学 | Digital audio watermarking method based on invariant characteristic of histogram |
| CN102074238A (en) * | 2010-12-13 | 2011-05-25 | 山东科技大学 | Linear interference cancellation-based speech secrete communication method |
| CN108962267B (en) * | 2018-07-09 | 2019-11-15 | 成都信息工程大学 | A kind of encryption voice content authentication method based on Hash feature |
| CN111292756B (en) * | 2020-01-19 | 2023-05-26 | 成都潜在人工智能科技有限公司 | Compression-resistant audio silent watermark embedding and extracting method and system |
| CN113808557A (en) * | 2020-06-12 | 2021-12-17 | 比亚迪股份有限公司 | Vehicle-mounted audio processing system, method and device |
-
2023
- 2023-01-15 CN CN202310056386.7A patent/CN115831131B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115831131A (en) | 2023-03-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hua et al. | Twenty years of digital audio watermarking—a comprehensive review | |
| JP3856652B2 (en) | Hidden data embedding method and apparatus | |
| Kirovski et al. | Blind pattern matching attack on watermarking systems | |
| Dhar et al. | A new audio watermarking system using discrete fourier transform for copyright protection | |
| JP2002519916A (en) | Apparatus and method for embedding information into analog signal using duplicate modulation | |
| WO2002091376A1 (en) | Generation and detection of a watermark robust against resampling of an audio signal | |
| Dhar et al. | Digital watermarking scheme based on fast Fourier transformation for audio copyright protection | |
| US20180144755A1 (en) | Method and apparatus for inserting watermark to audio signal and detecting watermark from audio signal | |
| Cvejic et al. | Robust audio watermarking in wavelet domain using frequency hopping and patchwork method | |
| JP4504681B2 (en) | Method and device for embedding auxiliary data in an information signal | |
| Zhang et al. | An audio digital watermarking algorithm transmitted via air channel in double DCT domain | |
| Liu et al. | Dear: A deep-learning-based audio re-recording resilient watermarking | |
| GB2431838A (en) | Audio processing | |
| CN115831131B (en) | Audio watermark embedding and extracting method based on deep learning | |
| US20020184503A1 (en) | Watermarking | |
| KR20030016381A (en) | Watermarking | |
| Subhashini et al. | Robust audio watermarking for monitoring and information embedding | |
| WO2001026110A1 (en) | Embedding and detecting watermarks in one-dimensional information signals | |
| Arnold et al. | Quality evaluation of watermarked audio tracks | |
| Acevedo | Audio watermarking: properties, techniques and evaluation | |
| JP2000209097A (en) | Signal processor, signal processing method, signal recorder, signal reproducing device and recording medium | |
| Dhar et al. | An efficient audio watermarking algorithm in frequency domain for copyright protection | |
| CN114743555A (en) | Method and device for realizing audio watermarking | |
| Cai et al. | A WAV Format Audio Digital Watermarking Algorithm Based on HAS | |
| KR100821349B1 (en) | Method for generating digital watermark and detecting digital watermark |
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 |