A digital audio watermark processing method, device, equipment and storage medium

Abstract

The present application discloses a watermark processing method, device, equipment and storage medium for digital audio, which belongs to the field of artificial intelligence technology. The present application receives a watermark adding instruction, performs Fourier transform on the original host voice, and obtains an audio signal; identifies the watermark information adding bit in the audio signal through the audio watermark model, and embeds the watermark message in the watermark information adding bit; combines the interference audio signal and the audio signal carrying the watermark message. Receive a message extraction signal, separate the interference audio signal from the combined audio signal; identify the watermark information adding bit in the audio signal, and extract the watermark message from the watermark information adding bit. The present application also relates to the field of blockchain technology, and the host voice is stored on the blockchain network. The present application automatically identifies the watermark information adding bit through the audio watermark model, realizes the addition and extraction of watermark information, significantly reduces the dependence on expert knowledge, and improves the speed and efficiency of watermark processing.

Description

Technical Field

[0001] The present application belongs to the field of artificial intelligence technology, and specifically relates to a watermark processing method, device, equipment and storage medium for digital audio. Background Art

[0002] With the widespread adoption and application of internet technology, various audio data are constantly circulating online. The copyright protection of audio data is becoming increasingly serious. Piracy and infringement are common, causing huge losses to creators. Therefore, strengthening copyright protection for audio data and combating infringement have become urgent issues.

[0003] Digital watermarking technology is an effective solution to these problems and can be used for copyright and authenticity authentication. For example, in insurance sales scenarios, digital watermarking can be applied to copyright protection of audio materials such as insurance contracts and product descriptions, ensuring the authenticity and integrity of information during the sales process, preventing tampering or misappropriation, and thus safeguarding the legitimate rights and interests of insurance companies. Current digital audio watermarking implementations typically rely on expert knowledge and experience to design digital watermark embedding and extraction methods. However, this method is highly dependent on expert knowledge and is significantly affected by the expert's knowledge reserves. Furthermore, this method has slow processing speed and low efficiency. Summary of the Invention

[0004] The purpose of the embodiments of the present application is to propose a digital audio watermark processing method, device, computer equipment and storage medium to solve the technical problems of existing digital audio watermark processing methods, such as high dependence on expert knowledge, significant influence of expert knowledge reserves, slow processing speed and low efficiency.

[0005] In order to solve the above technical problems, the present invention provides a method for watermarking digital audio, which adopts the following technical solutions:

[0006] A digital audio watermark processing method, comprising:

[0007] Receive a watermark adding instruction, obtain the original host voice, and perform Fourier transform on the original host voice to obtain the original host audio signal;

[0008] The pre-trained audio watermark model is used to identify the watermark information addition bit in the original host audio signal, and the watermark message is embedded in the watermark information addition bit to obtain an audio signal carrying the watermark message;

[0009] Acquire an interference audio signal, combine the interference audio signal with an audio signal carrying a watermark message to obtain a combined audio signal, and output the combined audio signal;

[0010] Receive a message extraction signal, obtain a combined audio signal, separate the interfering audio signal from the combined audio signal, and obtain an audio signal carrying a watermark message;

[0011] The watermark information addition bits in the audio signal carrying the watermark message are identified, and the watermark message is extracted from the watermark information addition bits.

[0012] Furthermore, the watermark information addition bits include a first addition bit and a second addition bit, the watermark message includes a watermark message body and an audio synchronization code matching the watermark message body, and the steps of identifying the watermark information addition bits in the original host audio signal through a pre-trained audio watermark model, and embedding the watermark message in the watermark information addition bits to obtain an audio signal carrying the watermark message specifically include:

[0013] Divide the original host audio signal to obtain several audio signal segments;

[0014] Counting the frequency range of each audio signal segment, and determining the non-sensitive audio signal segment based on the frequency range of each audio signal segment;

[0015] The watermark information is embedded in the non-sensitive audio signal segment through the audio watermark model;

[0016] Embedding the watermark message body in the first adding position;

[0017] An audio synchronization code is embedded in the second adding position to generate an audio signal carrying a watermark.

[0018] Furthermore, before the step of identifying the watermark information addition bit in the original host audio signal by using the pre-trained audio watermark model and embedding the watermark message in the watermark information addition bit to obtain the audio signal carrying the watermark message, the method further includes:

[0019] Obtain historical host voice and perform Fourier transform on the historical host voice to obtain historical host audio signal;

[0020] Divide the historical host audio signal to obtain several historical audio signal segments;

[0021] Counting the frequency range of each historical audio signal segment, and determining the historical non-sensitive audio signal segment based on the frequency range of each historical audio signal segment;

[0022] Embedding a historical watermark information adding bit in a historical non-sensitive audio signal segment and marking the historical watermark information adding bit;

[0023] The marked historical host audio signal is input into a preset initial audio watermark model for model training to obtain a trained audio watermark model.

[0024] Furthermore, the initial audio watermark model includes a message embedding network, a message extraction network, and an attack simulation layer. The steps of inputting the marked historical host audio signal into the preset initial audio watermark model for model training to obtain the trained audio watermark model specifically include:

[0025] Adding a bit-embedded historical watermark message to the historical watermark information in the historical host audio signal through a message embedding network to obtain an audio signal carrying the historical watermark message, wherein the historical watermark message includes a historical watermark message body and a historical audio synchronization code;

[0026] Use the attack simulation layer to generate a historical interference audio signal, and add the historical interference audio signal to the audio signal carrying the historical watermark message to generate a historical combined audio signal;

[0027] The historical combined audio signal is subjected to audio signal separation and message extraction through a message extraction network to obtain a host audio restoration signal and a restoration watermark message;

[0028] The model prediction error is calculated according to the host audio restoration signal and the restoration watermark message, and the initial audio watermark model is adjusted based on the model prediction error to obtain the trained audio watermark model.

[0029] Furthermore, the steps of calculating the model prediction error according to the host audio restoration signal and the restoration watermark message, and adjusting the initial audio watermark model based on the model prediction error to obtain the trained audio watermark model specifically include:

[0030] Calculating the error between the historical host audio signal and the restored host audio signal based on a preset first loss function to obtain a first prediction error;

[0031] Calculate the error between the historical watermark message and the restored watermark message based on a preset second loss function to obtain a second prediction error;

[0032] The initial audio watermark model is iterated based on the first prediction error and the second prediction error until the model is fitted, thereby obtaining a trained audio watermark model.

[0033] Furthermore, the expression of the first loss function is as follows:

[0034] L integrity =||(X syn ,X msg ),Ext(Emb(X host ,X syn ,X msg ))||2

[0035] Where, L integrity is the first loss function, Xhost , X syn , X msg They represent the historical host audio signal, historical audio synchronization code, and historical watermark message body respectively. Emb and Ext represent the message embedding operation and message extraction operation respectively.

[0036] Furthermore, the expression of the second loss function is as follows:

[0037] L imperceptibility =||X host ,Emb(X host ,X syn ,X msg )||2

[0038] Where, L imperceptibility is the second loss function, X host , X syn , X msg They represent the historical host audio signal, historical audio synchronization code, and historical watermark message body respectively, and Emb represents the message embedding operation.

[0039] In order to solve the above technical problems, the embodiment of the present application further provides a digital audio watermark processing device, which adopts the following technical solution:

[0040] A digital audio watermark processing device, comprising:

[0041] The audio conversion module is used to receive the watermark addition instruction, obtain the original host voice, and perform Fourier transform on the original host voice to obtain the original host audio signal;

[0042] A watermark adding module is used to identify the watermark information adding bit in the original host audio signal through the pre-trained audio watermark model, and embed the watermark message in the watermark information adding bit to obtain an audio signal carrying the watermark message;

[0043] The simulated attack module is used to obtain an interference audio signal, combine the interference audio signal with the audio signal carrying the watermark message to obtain a combined audio signal, and output the combined audio signal;

[0044] An audio separation module is used to receive a message extraction signal, obtain a combined audio signal, separate the interfering audio signal from the combined audio signal, and obtain an audio signal carrying a watermark message;

[0045] The watermark extraction module is used to identify the watermark information added bits in the audio signal carrying the watermark message and extract the watermark message from the watermark information added bits.

[0046] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0047] A computer device includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the digital audio watermark processing method as described in any one of the above are implemented.

[0048] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0049] A computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the digital audio watermark processing method as described in any one of the above.

[0050] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0051] The present application discloses a watermark processing method, device, equipment and storage medium for digital audio, which belongs to the field of artificial intelligence technology. The present application automatically identifies the watermark information addition position through the audio watermark model, realizes the addition and extraction of watermark information, significantly reduces the dependence on expert knowledge, and improves the speed and efficiency of watermark processing. Compared with the traditional method of manually selecting watermark embedding positions based on expert experience, the pre-trained model can more efficiently and accurately select the appropriate embedding position, thereby improving the processing speed and efficiency. At the same time, the audio signal is converted into a frequency domain signal through Fourier transform, which is conducive to embedding the watermark in the frequency domain and improving the concealment and robustness of the watermark. In addition, the present application also combines the interference audio signal to enhance the watermark's anti-attack ability, realizes rapid watermark extraction and signal separation, and improves the overall performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the solutions in this application, a brief introduction will be given below to the drawings required for use in the description of the embodiments of this application. Obviously, the drawings described below are some embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0053] Figure 1 shows an exemplary system architecture diagram in which the present application can be applied;

[0054] Figure 2 A flow chart showing an embodiment of a method for watermarking digital audio according to the present application is shown;

[0055] Figure 3 A schematic structural diagram of an embodiment of a digital audio watermark processing device according to the present application is shown;

[0056] Figure 4 A schematic structural diagram of an embodiment of a computer device according to the present application is shown. DETAILED DESCRIPTION

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the art to which this application belongs. The terms used in the specification of the application are for the purpose of describing specific embodiments only and are not intended to limit this application. The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of this application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0058] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0059] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0060] like Figure 1 As shown, system architecture 100 may include a terminal device 101, a network 102, and a server 103. Terminal device 101 may be a laptop computer 1011, a tablet computer 1012, or a mobile phone 1013. Network 102 is a medium for providing a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0061] The user can use the terminal device 101 to interact with the server 103 via the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0062] The terminal device 101 can be various electronic devices with a display screen and supporting web browsing. In addition to the laptop computer 1011, tablet computer 1012 or mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV) player, a laptop computer and a desktop computer, etc.

[0063] The server 103 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal device 101 .

[0064] It should be noted that the watermark processing method for digital audio provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the watermark processing device for digital audio is generally set in the server / terminal device.

[0065] It should be understood that Figure 1 The numbers of terminal devices, networks and servers in the embodiment are merely illustrative. The above system may have any number of terminal devices, networks and servers according to implementation requirements.

[0066] Continue to refer Figure 2 , shows a flow chart of an embodiment of a method for watermarking digital audio according to the present application. The method for watermarking digital audio includes the following steps:

[0067] S201, receiving a watermark adding instruction, obtaining the original host voice, and performing Fourier transform on the original host voice to obtain the original host audio signal.

[0068] Specifically, the system first obtains the original host voice. The original host voice is the audio file into which the watermark will be embedded. It can be a voice recording, music, or any other type of audio signal. To better process and analyze the audio signal, the system needs to convert the original host voice from a time domain signal to a frequency domain signal. This can be accomplished through a Fourier transform (such as the Short-Time Fourier Transform (STFT)). The Fourier transform is a mathematical tool used to decompose a time-domain audio signal into a superposition of a series of frequency components.

[0069] By performing Fourier transform on the original host voice, the system can represent the audio signal as various frequency components and their corresponding amplitude and phase. This frequency domain representation can not only more intuitively display the frequency characteristics of the audio signal, but also provide a more flexible operating space for watermark embedding. Embedding the watermark in the frequency domain can improve the concealment of the watermark, making it more difficult to detect while maintaining a high sound quality.

[0070] In addition, the frequency domain representation of the audio signal after Fourier transform enables the system to better achieve robustness and anti-interference capabilities through control of the frequency dimension.

[0071] S202: identifying watermark information addition bits in the original host audio signal through a pre-trained audio watermark model, and embedding a watermark message in the watermark information addition bits to obtain an audio signal carrying the watermark message.

[0072] Specifically, the core of this step is to use a pre-trained audio watermark model to automatically identify the most suitable location for embedding the watermark, namely the location where the watermark information is added. This audio watermark model is a reversible network-based digital audio watermarking implementation model that can sequentially embed and extract the audio watermark message. This pre-trained model is typically trained on a large number of audio samples and uses machine learning techniques to learn how to find frequency ranges in the audio spectrum that have a minimal impact on sound quality but are highly robust.

[0073] The choice of watermarking locations is crucial to audio quality and watermark concealment. The model analyzes the Fourier-transformed frequency domain signal and selects locations where the watermark will have the least impact on the original audio characteristics, thus avoiding noticeable degradation of sound quality. These locations also need to be resistant to interference, ensuring that the watermark information cannot be easily modified or deleted.

[0074] After determining the added bits, a watermark message is embedded into the audio signal. The watermark message can be any data that needs to be hidden within the audio, such as copyright information or an encrypted identification code. Watermarking techniques typically utilize tiny frequency or phase shifts, making the embedded audio virtually indistinguishable from the original audio while allowing the watermark information to be accurately extracted. Through this step, the system generates an audio signal carrying the watermark message.

[0075] S203: Acquire an interference audio signal, combine the interference audio signal with the audio signal carrying the watermark message to obtain a combined audio signal, and output the combined audio signal.

[0076] Specifically, obtaining an interfering audio signal is to improve the watermark's anti-attack capability and concealment. In practical applications, watermark signals may be subject to various forms of attacks, such as noise addition, compression processing, or other types of signal modification. The purpose of these attacks is usually to remove or destroy the watermark embedded in the audio, thereby circumventing copyright protection or hiding information.

[0077] Therefore, the introduction of interfering audio signals is a common technical means to improve the robustness of watermarking systems. Interfering audio signals can be natural noise, environmental audio clips, randomly generated audio signals, etc. Its purpose is to cover up or confuse the traces of watermark embedding.

[0078] By combining the interfering audio signal with the audio signal carrying the watermark message, the system generates a combined audio signal. During the combination process, the interfering audio signal and the audio signal carrying the watermark are superimposed according to a specific strategy. This can maintain the sound quality of the original audio while enhancing the concealment effect of the watermark.

[0079] When outputting the combined audio signal, the design of the interference signal should not only ensure the security of the watermark, but also maintain the naturalness of the audio to avoid obvious impact on the auditory experience.

[0080] S204: Receive a message extraction signal, obtain a combined audio signal, separate the interfering audio signal from the combined audio signal, and obtain an audio signal carrying the watermark message.

[0081] Specifically, during the watermark extraction process, the system first receives a message extraction signal, typically sent by the user or the system, instructing the start of watermark extraction. After receiving the extraction signal, the system first needs to obtain the combined audio signal and then separate the interfering audio signal, as the watermark information and the interfering signal have been mixed together during the previous watermark embedding process.

[0082] To accurately extract the watermark, the system uses preset separation rules to effectively separate and remove the interference signal from the combined audio signal. The separation of interference signals typically relies on the system's analysis of their characteristics. Combined with the specific strategy for adding the interference signal, the system can reversely process the interference portion and retain the original audio signal carrying the watermark. During the separation process, the system must also avoid compromising the integrity of the watermark information, ensuring that the watermark message remains accurate and extractable. The separated audio signal should be the original audio signal carrying the watermark message, free of external interference. Through this step, the system removes redundant portions of the audio and retains the core watermark carrier.

[0083] S205: Identify the watermark information addition bits in the audio signal carrying the watermark message, and extract the watermark message from the watermark information addition bits.

[0084] Specifically, after the system obtains the audio signal carrying the watermark message, the next step is to identify the watermark information added bits in the audio signal. Since the watermark is embedded in a specific frequency domain or time domain position during the extraction process, the system needs to rely on the same model algorithm to accurately locate the location information of the embedded watermark. This process is usually reversely inferred through the strategy when the watermark was embedded. These added bits usually have specific characteristics or are automatically identified through pre-trained models. The recognition process needs to consider various deformations of the audio signal during transmission and processing, such as compression, distortion, noise interference, etc., to ensure the robustness of the watermark information. After identifying the watermark information added bits, the system will extract the embedded watermark message from these bits.

[0085] Watermark extraction techniques can be based on amplitude changes at specific frequencies, phase shifts, or other covert information embedding techniques. The extracted watermark message undergoes decoding to reveal the original embedded information, such as copyright identification, authentication information, or other embedded secret data. This completes the watermark extraction process. The watermark information can be used for purposes such as verifying the audio source and protecting copyrights, ensuring that the embedded information can be accurately recovered without compromising sound quality.

[0086] In the above embodiment, the present application automatically identifies the watermark information addition position through the audio watermark model, realizes the addition and extraction of watermark information, significantly reduces the dependence on expert knowledge, and improves the speed and efficiency of watermark processing. Compared with the traditional method of manually selecting the watermark embedding position based on expert experience, the pre-trained model can more efficiently and accurately select the appropriate embedding position, thereby improving the processing speed and efficiency. At the same time, the audio signal is converted into a frequency domain signal through Fourier transform, which is conducive to embedding the watermark in the frequency domain and improving the concealment and robustness of the watermark. In addition, the present application also combines the interference audio signal to enhance the watermark's anti-attack capability, realizes rapid watermark extraction and signal separation, and improves the overall performance of the system.

[0087] Furthermore, the watermark information addition bits include a first addition bit and a second addition bit, the watermark message includes a watermark message body and an audio synchronization code matching the watermark message body, and the steps of identifying the watermark information addition bits in the original host audio signal through a pre-trained audio watermark model, and embedding the watermark message in the watermark information addition bits to obtain an audio signal carrying the watermark message specifically include:

[0088] Divide the original host audio signal to obtain several audio signal segments;

[0089] Counting the frequency range of each audio signal segment, and determining the non-sensitive audio signal segment based on the frequency range of each audio signal segment;

[0090] The watermark information is embedded in the non-sensitive audio signal segment through the audio watermark model;

[0091] Embedding the watermark message body in the first adding position;

[0092] An audio synchronization code is embedded in the second adding position to generate an audio signal carrying a watermark.

[0093] In this embodiment, the watermark embedding step first divides the original host audio signal into several audio signal segments. This facilitates independent analysis of each audio component, ensuring more accurate watermark embedding. Next, the system calculates the frequency range of each audio signal segment to identify those portions of the audio that are less impactful to the auditory sense—in other words, insensitive audio signal segments. Insensitive segments typically fall within frequency ranges that are difficult for the human ear to perceive. By analyzing the frequency characteristics of the audio, the system automatically selects these segments, ensuring that watermark embedding does not significantly impact sound quality.

[0094] The audio watermarking model then locates the watermarking locations within these non-sensitive segments. This approach avoids the complexity of manually selecting locations and reduces reliance on expert knowledge. The watermarking bits are divided into a primary and secondary bit, used to embed the watermark message and an audio synchronization code, respectively. The watermark message contains core hidden information, such as copyright or verification information, while the audio synchronization code ensures accurate location and recovery of the watermark during extraction. The synchronization code, by marking the time or frequency domain features of the audio, makes watermark extraction more accurate and efficient.

[0095] Through the above steps, an automated watermark embedding process is achieved, reducing dependence on manual labor and expert knowledge. By embedding the watermark in non-sensitive segments, the fidelity of the sound quality is ensured, while the concealment and anti-attack capabilities of the watermark are improved. The introduction of audio synchronization code enhances the robustness of watermark extraction, allowing the watermark information to remain extractable after audio processing.

[0096] Furthermore, before the step of identifying the watermark information addition bit in the original host audio signal by using the pre-trained audio watermark model and embedding the watermark message in the watermark information addition bit to obtain the audio signal carrying the watermark message, the method further includes:

[0097] Obtain historical host voice and perform Fourier transform on the historical host voice to obtain historical host audio signal;

[0098] Divide the historical host audio signal to obtain several historical audio signal segments;

[0099] Counting the frequency range of each historical audio signal segment, and determining the historical non-sensitive audio signal segment based on the frequency range of each historical audio signal segment;

[0100] Embedding a historical watermark information adding bit in a historical non-sensitive audio signal segment and marking the historical watermark information adding bit;

[0101] The marked historical host audio signal is input into a preset initial audio watermark model for model training to obtain a trained audio watermark model.

[0102] In this embodiment, the audio watermark model needs to be pre-trained before watermark embedding. Specifically, the system first obtains historical host speech and performs a Fourier transform on it, converting the audio signal from the time domain to the frequency domain. Next, the system divides the Fourier-transformed historical audio signal into several historical audio segments. The system then calculates the frequency range of each historical audio segment and identifies which frequency ranges are not easily perceived by the auditory sense, determining them as historical non-sensitive audio segments. These segments are considered to have a smaller impact on the embedded watermark and can be automatically identified through statistical analysis.

[0103] The system then embeds historical watermark information into non-sensitive audio segments and labels these locations. This labeling allows the system to better understand the characteristics of different audio segments. The system then feeds these labeled historical audio segments into the initial audio watermark model for training, enabling the model to automatically learn how to embed watermarks at different frequencies and locations, thereby forming a fully trained watermark model. This step enables the model to adapt to different types of audio and automatically identify suitable locations for watermarking, improving the automation and flexibility of watermark embedding.

[0104] Through the above steps, the pre-trained watermark model is trained using historical audio data, enabling it to automatically select the watermark embedding location based on the characteristics of the audio signal, reducing manual intervention and improving the robustness and adaptability of the model. By utilizing frequency analysis of historical non-sensitive fragments, the watermark embedding is more covert and has minimal impact on sound quality. At the same time, through training on large-scale data, the model has stronger generalization capabilities.

[0105] Furthermore, the initial audio watermark model includes a message embedding network, a message extraction network, and an attack simulation layer. The steps of inputting the marked historical host audio signal into the preset initial audio watermark model for model training to obtain the trained audio watermark model specifically include:

[0106] Adding a bit-embedded historical watermark message to the historical watermark information in the historical host audio signal through a message embedding network to obtain an audio signal carrying the historical watermark message, wherein the historical watermark message includes a historical watermark message body and a historical audio synchronization code;

[0107] Use the attack simulation layer to generate a historical interference audio signal, and add the historical interference audio signal to the audio signal carrying the historical watermark message to generate a historical combined audio signal;

[0108] The historical combined audio signal is subjected to audio signal separation and message extraction through a message extraction network to obtain a host audio restoration signal and a restoration watermark message;

[0109] The model prediction error is calculated according to the host audio restoration signal and the restoration watermark message, and the initial audio watermark model is adjusted based on the model prediction error to obtain the trained audio watermark model.

[0110] The audio watermark model in the embodiment of the present application is a dual-embedded digital audio watermark implementation model based on a reversible network. This model addresses the problems of existing digital audio watermark processing methods, which are highly dependent on expert knowledge, significantly affected by expert knowledge reserves, and have slow processing speeds and low efficiency. By designing a two-stage reversible network, the synchronization code and watermark message are separated and embedded twice. During the watermark extraction phase, a lightweight synchronization code extraction network is first used to quickly locate the watermark position, and then a message extractor with more layers is used to extract the embedded watermark message, thereby improving the watermark positioning speed during the extraction phase. To enhance the robustness of the watermark, a watermark removal attack simulation is added between the embedding network and the extraction network during the training process. This is used to train the robustness of the digital audio watermarking method, so that the audio embedded with the digital watermark can still extract a relatively accurate watermark after various watermark removal attacks.

[0111] After the host audio is transformed through the STFT transform, it is input into the embedding network. This embedding network employs a two-layer design. The first layer embeds the watermark message, and the second layer embeds a synchronization code with fewer bits at the same location, resulting in the watermarked audio. Various watermark removal attacks are then applied to the watermarked audio. The attacked audio is then input into the extraction network to extract the synchronization code and message. The extraction network also employs a two-layer design. It first extracts the watermark positioning code, and then feeds the output audio into the second layer to extract the watermark message.

[0112] In this embodiment, first, the historical watermark message is embedded into the historical watermark information adding position of the historical host audio signal through the message embedding network. The historical watermark message includes the watermark message body and the historical audio synchronization code. The watermark message body is the main information that needs to be embedded, such as copyright, identity identification, etc., and the historical audio synchronization code ensures that the watermark can be accurately located during the extraction process.

[0113] The message embedding network analyzes the characteristics of the host audio and covertly embeds the watermark message into the frequency or time domain of the audio to ensure its concealment and anti-interference properties. Next, the system uses the attack simulation layer to generate historical interference audio signals. These interference audio signals simulate various attacks that may occur in real-life scenarios, such as noise, compression, and editing. The interference audio signal is superimposed on the audio signal carrying the historical watermark message to generate a historical combined audio signal. The purpose is to test the robustness and anti-attack capabilities of the watermark. Then, through the message extraction network, the system processes the historical combined audio signal, separates the interference audio signal from the host audio signal, and extracts the embedded watermark message. This step simulates the watermark extraction process in a real environment. By restoring the host audio signal and restoring the watermark message, the anti-attack effect of the watermark is verified.

[0114] Finally, the system calculates the prediction error of the model based on the host audio restoration signal and the restored watermark message. This error reflects the loss or deviation in the embedding and extraction process. By performing error feedback adjustment on the initial watermark model, the model is gradually optimized to have better robustness and restoration capabilities when facing interference.

[0115] Through the above steps, the attack simulation layer and the message extraction network are combined in the watermark embedding process, which enhances the robustness of the model in complex environments. Through feedback adjustment of the prediction error, the model is continuously optimized during the training process. The final generated audio watermark model can effectively resist various interferences and attacks in practical applications and ensure the accurate extraction of the watermark.

[0116] Furthermore, the steps of calculating the model prediction error according to the host audio restoration signal and the restoration watermark message, and adjusting the initial audio watermark model based on the model prediction error to obtain the trained audio watermark model specifically include:

[0117] Calculating the error between the historical host audio signal and the restored host audio signal based on a preset first loss function to obtain a first prediction error;

[0118] Calculate the error between the historical watermark message and the restored watermark message based on a preset second loss function to obtain a second prediction error;

[0119] The initial audio watermark model is iterated based on the first prediction error and the second prediction error until the model is fitted, thereby obtaining a trained audio watermark model.

[0120] In this embodiment, the system first calculates the error between the historical host audio signal and the restored host audio signal based on a preset first loss function, obtaining a first prediction error. The purpose of this step is to measure the degree of restoration of the host audio signal after watermark embedding and attack simulation, namely, the integrity and sound quality fidelity of the audio signal. Secondly, the error between the historical watermark message and the restored watermark message is calculated based on a second loss function, obtaining a second prediction error. This step evaluates the restoration effect of the watermark message and ensures that the watermark information can still be accurately extracted after interference. Subsequently, the system iteratively optimizes the initial audio watermark model based on the first and second prediction errors, gradually reducing the error by continuously adjusting the model parameters until the model reaches the optimal fit state.

[0121] Through the above steps, the audio watermark model is effectively optimized to ensure that watermark embedding has little impact on sound quality and can accurately extract watermark information in complex environments.

[0122] Furthermore, the expression of the first loss function is as follows:

[0123] L integrity =||(X syn ,X msg ),Ext(Emb(X host ,X syn ,X msg ))||2

[0124] Where, L integrity is the first loss function, X host , X syn , X msg They represent the historical host audio signal, historical audio synchronization code, and historical watermark message body respectively. Emb and Ext represent the message embedding operation and message extraction operation respectively.

[0125] In this embodiment, the first loss function L integrity Indicates the watermark integrity loss, requiring the embedded watermark message and the extracted watermark message to be as consistent as possible. Since the output of the extraction network is a number in the range of 0 to 1, the method of setting a threshold and rounding is adopted in practical application judgment. In the training process, the output is directly used as the extracted watermark information to ensure the stability of the training. The first loss function L is used. integrity Constrain the difference between the two.

[0126] Furthermore, the expression of the second loss function is as follows:

[0127] L imperceptibility =||X host ,Emb(X host ,X syn ,X msg )||2

[0128] Where, L imperceptibility is the second loss function, X host , X syn , X msg They represent the historical host audio signal, historical audio synchronization code, and historical watermark message body respectively, and Emb represents the message embedding operation.

[0129] In this embodiment, the second loss function L imperceptibility Indicates the imperceptible loss, requiring the watermarked audio to be as similar as possible to the original host audio, so as to reduce the changes brought by the watermark to the audio. The second loss function L is used. imperceptibility The difference between the two loss constraints.

[0130] Furthermore, the final loss function L of the audio watermark model training is total is the first loss function L integrity and the second loss function L imperceptibility The weighted sum of is expressed as follows:

[0131] L total =a1L integrity +a2L imperceptibility

[0132] Where a1 is the first loss function L integrity The weight of a2 is the second loss function L imperceptibility The weight of .

[0133] In the above embodiment, the present application discloses a watermark processing method for digital audio, which belongs to the field of artificial intelligence technology. The present application automatically identifies the watermark information addition position through the audio watermark model, realizes the addition and extraction of watermark information, significantly reduces the dependence on expert knowledge, and improves the speed and efficiency of watermark processing. Compared with the traditional method of manually selecting watermark embedding positions based on expert experience, the pre-trained model can more efficiently and accurately select the appropriate embedding position, thereby improving the processing speed and efficiency. At the same time, the audio signal is converted into a frequency domain signal through Fourier transform, which is conducive to embedding the watermark in the frequency domain and improving the concealment and robustness of the watermark. In addition, the present application also combines the interference audio signal to enhance the watermark's anti-attack ability, realizes rapid watermark extraction and signal separation, and improves the overall performance of the system.

[0134] In this embodiment, the watermark processing method for digital audio is executed on an electronic device (eg Figure 1The server shown in the figure) can receive instructions or obtain data through a wired connection or a wireless connection. It should be noted that the above-mentioned wireless connection method may include but is not limited to 3G / 4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0135] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned host voice information, the above-mentioned host voice information can also be stored in a node of a blockchain.

[0136] The blockchain referred to in this application is a new application model for computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Blockchain is essentially a decentralized database, a series of data blocks generated using cryptographic methods. Each data block contains information about a batch of network transactions, which is used to verify the validity of this information (to prevent counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, the platform product service layer, and the application service layer.

[0137] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0138] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0139] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware via computer-readable instructions. The computer-readable instructions can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes in the above-described method embodiments. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0140] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0141] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a digital audio watermark processing device. Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0142] like Figure 3 As shown, the digital audio watermark processing device 300 described in this embodiment includes:

[0143] The audio conversion module 301 is used to receive a watermark adding instruction, obtain the original host voice, and perform Fourier transform on the original host voice to obtain the original host audio signal;

[0144] The watermark adding module 302 is used to identify the watermark information adding bit in the original host audio signal through the pre-trained audio watermark model, and embed the watermark message in the watermark information adding bit to obtain the audio signal carrying the watermark message;

[0145] The simulated attack module 303 is used to obtain an interference audio signal, combine the interference audio signal with the audio signal carrying the watermark message to obtain a combined audio signal, and output the combined audio signal;

[0146] The audio separation module 304 is configured to receive the message extraction signal, obtain a combined audio signal, separate the interfering audio signal from the combined audio signal, and obtain an audio signal carrying the watermark message;

[0147] The watermark extraction module 305 is configured to identify watermark information addition bits in the audio signal carrying the watermark message, and extract the watermark message from the watermark information addition bits.

[0148] Furthermore, the watermark information adding bit includes a first adding bit and a second adding bit, and the watermark message includes a watermark message body and an audio synchronization code matching the watermark message body. The watermark adding module 302 is specifically configured to:

[0149] The original host audio signal is divided to obtain several audio signal segments; the frequency range of each audio signal segment is counted, and non-sensitive audio signal segments are determined based on the frequency range of each audio signal segment; watermark information addition positions are embedded in the non-sensitive audio signal segments through the audio watermark model; the watermark message body is embedded in the first addition position; and the audio synchronization code is embedded in the second addition position to generate an audio signal carrying the watermark.

[0150] Furthermore, the digital audio watermark processing apparatus 300 further includes a watermark model training module, which is specifically configured to:

[0151] Acquire historical host voices and perform Fourier transform on the historical host voices to obtain historical host audio signals; divide the historical host audio signals to obtain a number of historical audio signal segments; count the frequency ranges of each historical audio signal segment, and determine historical non-sensitive audio signal segments based on the frequency ranges of each historical audio signal segment; embed historical watermark information addition bits in the historical non-sensitive audio signal segments, and mark the historical watermark information addition bits; input the marked historical host audio signals into a preset initial audio watermark model for model training to obtain a trained audio watermark model.

[0152] Furthermore, the initial audio watermark model includes a message embedding network, a message extraction network, and an attack simulation layer. The training module is further used to:

[0153] A message embedding network is used to add a bit-embedded historical watermark message to the historical watermark information in the historical host audio signal to obtain an audio signal carrying the historical watermark message, wherein the historical watermark message includes the historical watermark message body and the historical audio synchronization code; an attack simulation layer is used to generate a historical interference audio signal, and the historical interference audio signal is added to the audio signal carrying the historical watermark message to generate a historical combined audio signal; audio signal separation and message extraction are performed on the historical combined audio signal through a message extraction network to obtain a host audio restoration signal and a restored watermark message; the model prediction error is calculated according to the host audio restoration signal and the restored watermark message, and the initial audio watermark model is adjusted based on the model prediction error to obtain a trained audio watermark model.

[0154] Furthermore, the training module is further used to specifically include:

[0155] Based on the preset first loss function, the error between the historical host audio signal and the host audio restoration signal is calculated to obtain a first prediction error; based on the preset second loss function, the error between the historical watermark message and the restored watermark message is calculated to obtain a second prediction error; based on the first prediction error and the second prediction error, the initial audio watermark model is iterated until the model is fitted to obtain a trained audio watermark model.

[0156] Furthermore, the expression of the first loss function is as follows:

[0157] L integrity =||(X syn ,X msg ),Ext(Emb(X host ,X syn ,X msg ))||2

[0158] Where, L integrity is the first loss function, X host , X syn , X msg They represent the historical host audio signal, historical audio synchronization code, and historical watermark message body respectively. Emb and Ext represent the message embedding operation and message extraction operation respectively.

[0159] Furthermore, the expression of the second loss function is as follows:

[0160] L imperceptibility =||X host ,Emb(X host ,X syn ,X msg )||2

[0161] Where, L imperceptibility is the second loss function, X host , X syn , X msg They represent the historical host audio signal, historical audio synchronization code, and historical watermark message body respectively, and Emb represents the message embedding operation.

[0162] In the above embodiment, the present application discloses a watermark processing device for digital audio, which belongs to the field of artificial intelligence technology. The present application automatically identifies the watermark information addition position through the audio watermark model, realizes the addition and extraction of watermark information, significantly reduces the dependence on expert knowledge, and improves the speed and efficiency of watermark processing. Compared with the traditional method of manually selecting watermark embedding positions based on expert experience, the pre-trained model can more efficiently and accurately select the appropriate embedding position, thereby improving the processing speed and efficiency. At the same time, the audio signal is converted into a frequency domain signal through Fourier transform, which is conducive to embedding the watermark in the frequency domain and improving the concealment and robustness of the watermark. In addition, the present application also combines the interference audio signal to enhance the watermark's anti-attack ability, realizes rapid watermark extraction and signal separation, and improves the overall performance of the system.

[0163] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4This is a basic structural block diagram of the computer device in this embodiment.

[0164] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with a memory 41, a processor 42, and a network interface 43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, etc.

[0165] The computer device may be a desktop computer, notebook computer, PDA, cloud server, etc. The computer device may interact with the user via a keyboard, mouse, remote control, touchpad, or voice control device.

[0166] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the computer device 4. Of course, the memory 41 may also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for the watermark processing method for digital audio. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0167] In some embodiments, the processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for executing the digital audio watermark processing method.

[0168] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0169] In the above embodiment, the present application discloses a computer device, which belongs to the field of artificial intelligence technology. The present application automatically identifies the watermark information addition position through the audio watermark model, realizes the addition and extraction of watermark information, significantly reduces the dependence on expert knowledge, and improves the speed and efficiency of watermark processing. Compared with the traditional method of manually selecting watermark embedding positions based on expert experience, the pre-trained model can more efficiently and accurately select the appropriate embedding position, thereby improving the processing speed and efficiency. At the same time, the audio signal is converted into a frequency domain signal through Fourier transform, which is conducive to embedding the watermark in the frequency domain and improving the concealment and robustness of the watermark. In addition, the present application also combines the interference audio signal to enhance the watermark's anti-attack ability, realizes rapid watermark extraction and signal separation, and improves the overall performance of the system.

[0170] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the digital audio watermark processing method as described above.

[0171] In the above embodiment, the present application discloses a computer-readable storage medium, which belongs to the field of artificial intelligence technology. The present application automatically identifies the watermark information addition position through the audio watermark model, realizes the addition and extraction of watermark information, significantly reduces the dependence on expert knowledge, and improves the speed and efficiency of watermark processing. Compared with the traditional method of manually selecting watermark embedding positions based on expert experience, the pre-trained model can more efficiently and accurately select the appropriate embedding position, thereby improving the processing speed and efficiency. At the same time, the audio signal is converted into a frequency domain signal through Fourier transform, which is conducive to embedding the watermark in the frequency domain and improving the concealment and robustness of the watermark. In addition, the present application also combines the interference audio signal to enhance the watermark's anti-attack ability, realizes rapid watermark extraction and signal separation, and improves the overall performance of the system.

[0172] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0173] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0174] Obviously, the embodiments described above are only some of the embodiments of the present application, rather than all of the embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions described in the aforementioned specific embodiments, or to make equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the present application specification and the accompanying drawings, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present application.

Claims

1. A digital audio watermark processing method, characterized in that: include: receiving a watermark adding instruction, obtaining an original host voice, and performing Fourier transform on the original host voice to obtain an original host audio signal; Identifying watermark information addition bits in the original host audio signal through a pre-trained audio watermark model, and embedding a watermark message in the watermark information addition bits to obtain an audio signal carrying the watermark message; Acquire an interference audio signal, combine the interference audio signal with the audio signal carrying the watermark message to obtain a combined audio signal, and output the combined audio signal; receiving a message extraction signal, obtaining the combined audio signal, separating the interfering audio signal from the combined audio signal, and obtaining the audio signal carrying the watermark message; Identifying watermark information addition bits in the audio signal carrying the watermark message, and extracting the watermark message from the watermark information addition bits; Before the step of identifying watermark information addition bits in the original host audio signal by using the pre-trained audio watermark model and embedding a watermark message in the watermark information addition bits to obtain an audio signal carrying the watermark message, the method further includes: Acquire historical host voices and perform Fourier transform on the historical host voices to obtain historical host audio signals; Dividing the historical host audio signal to obtain a plurality of historical audio signal segments; Counting the frequency range of each of the historical audio signal segments, and determining a historical non-sensitive audio signal segment based on the frequency range of each of the historical audio signal segments; Embedding a historical watermark information adding bit in the historical non-sensitive audio signal segment and marking the historical watermark information adding bit; Inputting the marked historical host audio signal into a preset initial audio watermark model for model training to obtain a trained audio watermark model; The initial audio watermark model includes a message embedding network, a message extraction network, and an attack simulation layer. The step of inputting the marked historical host audio signal into the preset initial audio watermark model for model training to obtain the trained audio watermark model specifically includes: Adding a bit-embedded historical watermark message to the historical watermark information in the historical host audio signal through the message embedding network to obtain an audio signal carrying the historical watermark message, wherein the historical watermark message includes a historical watermark message body and a historical audio synchronization code; Generate a historical interference audio signal using the attack simulation layer, and add the historical interference audio signal to the audio signal carrying the historical watermark message to generate a historical combined audio signal; Performing audio signal separation and message extraction on the historical combined audio signal through the message extraction network to obtain a host audio restoration signal and a restoration watermark message; A model prediction error is calculated according to the host audio restoration signal and the restoration watermark message, and the initial audio watermark model is adjusted based on the model prediction error to obtain a trained audio watermark model.

2. The digital audio watermark processing method according to claim 1, wherein: The watermark information addition bits include a first addition bit and a second addition bit, the watermark message includes a watermark message body and an audio synchronization code matching the watermark message body, and the steps of identifying the watermark information addition bits in the original host audio signal using a pre-trained audio watermark model and embedding the watermark message in the watermark information addition bits to obtain an audio signal carrying the watermark message specifically include: Dividing the original host audio signal to obtain a plurality of audio signal segments; counting frequency ranges of the respective audio signal segments, and determining non-sensitive audio signal segments based on the frequency ranges of the respective audio signal segments; embedding the watermark information adding bit in the non-sensitive audio signal segment by using the audio watermark model; embedding the watermark message body in the first adding position; The audio synchronization code is embedded in the second adding position to generate an audio signal carrying a watermark.

3. The digital audio watermark processing method according to claim 1, wherein: The step of calculating a model prediction error according to the host audio restoration signal and the restoration watermark message, and adjusting the initial audio watermark model based on the model prediction error to obtain a trained audio watermark model specifically includes: Calculating an error between the historical host audio signal and the restored host audio signal based on a preset first loss function to obtain a first prediction error; Calculating the error between the historical watermark message and the restored watermark message based on a preset second loss function to obtain a second prediction error; The initial audio watermark model is iterated based on the first prediction error and the second prediction error until the model is fitted, thereby obtaining a trained audio watermark model.

4. The digital audio watermark processing method according to claim 3, wherein: The expression of the first loss function is as follows: Where, is the first loss function, , , They represent the historical host audio signal, historical audio synchronization code, and historical watermark message body respectively. Emb and Ext represent the message embedding operation and message extraction operation respectively.

5. The digital audio watermark processing method according to claim 3, wherein: The expression of the second loss function is as follows: Where, is the second loss function, , , They represent the historical host audio signal, historical audio synchronization code, and historical watermark message body respectively, and Emb represents the message embedding operation.

6. A digital audio watermark processing device, characterized in that: include: An audio conversion module is used to receive a watermark adding instruction, obtain an original host voice, and perform Fourier transform on the original host voice to obtain an original host audio signal; a watermark adding module, configured to identify watermark information adding bits in the original host audio signal using a pre-trained audio watermark model, and embed a watermark message in the watermark information adding bits to obtain an audio signal carrying the watermark message; a simulated attack module, configured to obtain an interference audio signal, combine the interference audio signal with the audio signal carrying the watermark message to obtain a combined audio signal, and output the combined audio signal; an audio separation module, configured to receive a message extraction signal, obtain the combined audio signal, separate the interfering audio signal from the combined audio signal, and obtain the audio signal carrying the watermark message; a watermark extraction module, configured to identify watermark information addition bits in the audio signal carrying the watermark message, and extract the watermark message from the watermark information addition bits; Furthermore, the digital audio watermark processing device further includes a watermark model training module, which is specifically used to: Acquire historical host speech and perform Fourier transform on the historical host speech to obtain a historical host audio signal; divide the historical host audio signal to obtain a plurality of historical audio signal segments; calculate the frequency range of each historical audio signal segment, and determine historical non-sensitive audio signal segments based on the frequency range of each historical audio signal segment; embed historical watermark information addition bits in the historical non-sensitive audio signal segments, and mark the historical watermark information addition bits; input the marked historical host audio signal into a preset initial audio watermark model for model training to obtain a trained audio watermark model; Furthermore, the initial audio watermark model includes a message embedding network, a message extraction network, and an attack simulation layer. The training module is further used to: A message embedding network is used to add a bit-embedded historical watermark message to the historical watermark information in the historical host audio signal to obtain an audio signal carrying the historical watermark message, wherein the historical watermark message includes the historical watermark message body and the historical audio synchronization code; an attack simulation layer is used to generate a historical interference audio signal, and the historical interference audio signal is added to the audio signal carrying the historical watermark message to generate a historical combined audio signal; audio signal separation and message extraction are performed on the historical combined audio signal through a message extraction network to obtain a host audio restoration signal and a restored watermark message; the model prediction error is calculated according to the host audio restoration signal and the restored watermark message, and the initial audio watermark model is adjusted based on the model prediction error to obtain a trained audio watermark model.

7. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the digital audio watermark processing method according to any one of claims 1 to 5 when executing the computer-readable instructions.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the digital audio watermark processing method according to any one of claims 1 to 5.

Images