Network class copyright authentication method based on time domain hidden watermark

Abstract

The invention provides an online class copyright authentication method based on a time domain hidden watermark, and relates to the technical field of electrical digital data process.The online class copyright authentication method comprises the steps that a microscopic noise signal in a display screen drive circuit is collected from a client side, time domain sampling, filtering normalization and principal component analysis dimension reduction processing are conducted on the noise signal, and a unique noise feature vector NS is extracted; further combining with local video statistical parameters of a client video frame, and realizing dynamic adjustment of watermark embedding amplitude and position by adopting an adaptive modulation model; the randomness and the non-replicability of noise features are utilized, unique generation of a watermark key is achieved through coupling of local video parameters and a nonlinear mapping function, and the safety and the tamper-resistant capacity of copyright authentication are remarkably improved. And a feasible technical path is provided for realizing unique determination of the watermark key.

Description

Technical Field

[0001] The present invention relates to the technical field of digital data processing, and specifically to an online course copyright authentication method based on time-domain hidden watermarking. Background Art

[0002] Traditional digital watermarking schemes mostly rely on methods such as frequency domain, spatial domain or hybrid domain. Through processes such as transform domain decomposition, feature extraction and watermark modulation, the hiding and transmission of copyright information are realized. Among them, frequency domain watermarking technology uses methods such as discrete cosine transform (DCT) and discrete wavelet transform (DWT) to embed watermarks into mid-low frequency coefficients in order to maintain the robustness of watermark information during compression and transmission; while spatial domain watermarking technology directly operates on pixel values and has advantages such as simple implementation and high real-time performance. However, traditional technologies generally face limitations in practical applications, such as insufficient anti-interference ability of the watermark after embedding, easy tampering of the embedding position, and affecting the stability of the watermark signal in a high-noise environment.

[0003] In the prior art, the publication number is CN110543749A, and the name is a method for adding an invisible watermark to a client. By obtaining an invisible watermark processing request sent by a user, user information, and a solid-color picture selected by the user, the user information is transformed according to the invisible watermark processing request to generate a unique first watermark ID, and the first watermark ID is stored in a watermark database. Then, the first watermark ID and the solid-color picture are synthesized to generate an invisible watermark picture with the first watermark ID, and finally the invisible watermark picture is returned to the user for the user to use the invisible watermark picture as the background of the client.

[0004] However, there are still many deficiencies in practical applications. Especially in multi-scenario applications such as online courses, higher requirements are placed on the uniqueness, robustness and stability of copyright information. Existing technologies mostly rely on traditional frequency domain or spatial domain watermarking methods. When facing complex processing methods such as intelligent video compression, editing, and malicious tampering, the watermark signal is often lost or distorted due to local suppression or reconstruction error. At the same time, traditional solutions do not fully utilize the microscopic noise information in the display driving circuit and the local video statistical parameters of the client, and these information have potential advantages in ensuring the unpredictability and uniqueness of embedding.

[0005] The above information disclosed in the above background art section is only used to enhance the understanding of the background of the present disclosure, and therefore it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide an online course copyright authentication method based on time-domain hidden watermarking to solve the problems raised in the above background art.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for online course copyright authentication based on time-domain hidden watermark, the specific steps include:

[0009] Step S1: Collect the microscopic noise signal in the display screen drive circuit at the client, and the microscopic noise signal is a digital noise data sequence obtained by continuous sampling within a time-domain sampling window;

[0010] Step S2: Filter and normalize the collected digital noise data sequence according to a unified time-domain block, and then extract the preliminary noise statistical feature sequence of the digital noise data sequence;

[0011] Step S3: Use the principal component analysis method to reduce the dimension of the extracted preliminary noise statistical feature sequence to obtain the noise feature vector NS;

[0012] Step S4: Based on the local video statistical parameters of the video frames in the client audio-visual data, dynamically modulate the amplitude and position of the time-domain watermark embedding through an adaptive modulation model, and the modulation process uses a non-linear function to couple the local video statistical parameters with the watermark embedding amplitude;

[0013] Step S5: Combine the noise feature vector NS with the adaptive modulation model, and generate a unique watermark key through a determined non-linear mapping function;

[0014] Step S6: During the time-domain hidden watermark embedding process, couple the watermark key with the watermark modulation parameters, and the coupling process involves adjusting the watermark embedding position and amplitude within the time-domain data window of the audio-visual data.

[0015] Compared with the prior art, the beneficial effects of the present invention are: by collecting the microscopic noise signal in the display screen drive circuit at the client, performing time-domain sampling, filtering normalization and principal component analysis dimensionality reduction processing on the noise signal, extracting a unique noise feature vector NS, and further combining it with the local video statistical parameters of the client video frames, an adaptive modulation model is used to realize the dynamic adjustment of the watermark embedding amplitude and position; not only making use of the randomness and non-replicability of the noise features, but also realizing the unique generation of the watermark key through the coupling of local video parameters and non-linear mapping functions, significantly improving the security and anti-tampering ability of copyright authentication; providing a practical technical path for realizing the unique determination of the watermark key. Description of the Drawings

[0016] Figure 1 It is a schematic diagram of the overall method flow of the present invention. Detailed Embodiments

[0017] To make the objectives, technical solutions, and advantages of the present invention more clearly understood, the present invention will be further described in detail below in conjunction with specific embodiments.

[0018] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meanings understood by those of ordinary skill in the field to which the present invention pertains. The "first", "second", and similar terms used in the present invention do not denote any order, quantity, or importance, but are only used to distinguish different components. The terms such as "comprising" or "including" mean that the elements or objects appearing before this term cover the elements or objects listed after this term and their equivalents, without excluding other elements or objects. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left", "right", etc. are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0019] Embodiment 1:

[0020] Please refer to Figure 1 , the present invention provides a technical solution:

[0021] An online course copyright authentication method based on time-domain hidden watermark, the specific steps include:

[0022] Step S1: Collect the microscopic noise signal in the display driver circuit at the client side, and the microscopic noise signal is a digital noise data sequence obtained by continuous sampling within a time-domain sampling window;

[0023] Further explanation: The specific operation steps are as follows:

[0024] The collection of the microscopic noise signal is realized by the cooperation of the high-precision analog-to-digital converter (ADC) and the low-noise amplifier of the collection module;

[0025] In the client device, a set of integrated high-precision ADC circuits and low-noise amplifiers are set, connected to the test points of the display driver circuit to collect the microscopic noise signals existing in the circuit in real time, and the microscopic noise signals include thermal noise signals and random current noise signals;

[0026] Thermal noise signal: Originating from the thermal motion of electrons in the conductor to which the client belongs, it will cause random fluctuations in voltage and current. This kind of noise changes with temperature, is calculated by the root mean square value formula, and has broadband characteristics.

[0027] For the characterization of the thermal noise signal:

[0028] Root mean square voltage (RMS Voltage):

[0029] Thermal noise is usually characterized by voltage, and its root mean square voltage can be calculated by the formula:

[0030] ;

[0031] where, is the root mean square value of the thermal noise voltage; is the Boltzmann constant, with a value of ; T is the absolute temperature (unit: Kelvin); R is the resistance value (unit: Ohm); B is the bandwidth (unit: Hertz).

[0032] Spectral characteristics: Thermal noise has white noise characteristics, that is, it is evenly distributed across the entire spectrum, which means its power spectral density is equal at all frequencies.

[0033] Random current noise signal: It is obvious at low frequencies and is generated due to the random variation of current and material inhomogeneity. This noise is also called flicker noise or "1 / f noise", and its intensity is related to frequency;

[0034] Characterization of the random current noise signal:

[0035] Power Spectral Density (PSD):

[0036] Random current noise is usually significant in the low-frequency region, and its power spectral density can be defined as:

[0037] <{ ;

[0038] where, is the power spectral density of the noise current; is a constant related to the material and process; is the frequency; is the spectral index, with a value between 0.5 and 1.5.

[0039] 1 / f noise characteristics:

[0040] Random current noise is "flicker noise" or "1 / f noise", which is characterized by a higher power spectrum in the low-frequency band and decreases with increasing frequency.

[0041] Record the original analog signal of the collected data and transfer it to the subsequent digital conversion unit;

[0042] Digitally convert the collected microscopic noise signal according to a fixed time-domain sampling window to obtain a continuous digital noise data sequence; specifically: the fixed time-domain sampling window in the embodiment is a sampling period of 50 milliseconds; within each 50-millisecond sampling period, the analog microscopic noise signal output from the acquisition module is input to the ADC module;

[0043] The ADC module completes the analog conversion of data within each time-domain sampling window according to the set sampling rate, and outputs a continuous digital noise data sequence.

[0044] Cache the digital noise data sequence within each time-domain sampling window obtained by continuous sampling; for subsequent preprocessing; specifically:

[0045] Store the digital noise data sequence within each time-domain sampling window output from the ADC module into the local cache system;

[0046] The cache design adopts a circular storage structure to continuously store the continuous sampling data sequence;

[0047] After the data caching is completed, it provides a data source for subsequent preprocessing and hidden watermark embedding.

[0048] Step S2: Filter and normalize the collected digital noise data sequence according to a unified time-domain block, and then extract the preliminary noise statistical feature sequence of the digital noise data sequence;

[0049] Further explanation: For the digital noise data sequence within each time-domain sampling window in the cache, a digital low-pass filter is used for filtering to eliminate high-frequency interference clutter;

[0050] Specifically: Perform digital low-pass filtering on the digital noise data sequence within each time-domain sampling window stored in the client cache;

[0051] Configure the parameters of the digital low-pass filter, and set the cut-off frequency to ensure that high-frequency interference clutter is effectively suppressed from the data sequence;

[0052] Output the filtered data sequence to the next processing unit;

[0053] Normalize the filtered digital noise data sequence, and the processing result falls within the preset numerical interval [0, 1];

[0054] Within each time-domain sampling window, calculate the digital noise data sequence by combining the edge detection algorithm and the moving average method to generate a preliminary noise statistical feature sequence; this statistical feature includes but is not limited to the mean, variance, and extreme values; specifically:

[0055] Using the normalized digital noise data sequence, first perform the first-order differential detection of the edge detection algorithm to identify the edge information of the numerical fluctuations in the sequence;

[0056] Based on the edge detection result, adopt the moving average method, set the fixed window size to 10 sampling points, and smooth the data sequence;

[0057] Calculate the statistical parameters of the smoothed data sequence respectively, including the mean, variance, and extreme values represented by the maximum or minimum value in the sequence, and form a preliminary noise statistical feature sequence;

[0058] Transfer the generated preliminary noise statistical feature sequence to the subsequent step S3.

[0059] Step S3: Use the principal component analysis method to reduce the dimension of the extracted preliminary noise statistical feature sequence to obtain the noise feature vector NS;

[0060] Further explanation: Convert the preliminary noise statistical feature sequence into a noise feature vector with a fixed dimension; the specific operation steps are as follows:

[0061] Take the preliminary noise statistical feature sequence normalized within each time-domain sampling window as the input data matrix. Each row of the data matrix corresponds to a different time-domain sampling window, and each column is a normalized statistical index;

[0062] Use the standard PCA algorithm to perform covariance matrix calculation on the input data matrix to obtain eigenvalues and eigenvectors; specifically:

[0063] Perform covariance matrix calculation, and its formula is , where X is the input data matrix and μ is the mean of each column;

[0064] Perform the operation of solving eigenvalues and eigenvectors of the covariance matrix to obtain all eigenvalues and the corresponding eigenvectors;

[0065] Arrange the obtained eigenvalues in descending order and arrange the eigenvectors correspondingly;

[0066] And select the first k principal components to make their cumulative contribution rate reach the preset contribution rate threshold; in this embodiment, the preset contribution rate threshold is set to 90%;

[0067] Specifically, construct the selected principal components into a noise feature vector NS with a fixed length, that is ; represents the noise feature component of the kth principal component; specifically:

[0068] For the obtained eigenvalues arranged in descending order, calculate the contribution rate of each principal component, and its single contribution rate is: ;

[0069] The cumulative contribution rate is accumulated in turn , until the cumulative value is greater than or equal to the preset contribution rate threshold of 90%, and determine to select the first k principal components, where k is the smallest integer that satisfies Σ i =1 k CR i ≥90%;

[0070] Σ i = 1 k CR i ≥ 90% means that the accumulation starts from the first item \(i = 1\) of the principal component contribution rate corresponding to the eigenvalue, and accumulates to the \(k\)th item in sequence until the cumulative contribution rate reaches or exceeds the preset contribution rate threshold of 90%;

[0071] Construct the noise feature vector NS as , where represents the eigen-component corresponding to the \(i\)th principal component.

[0072] Serialize and store the constructed noise feature vector NS in the order of each time-domain sampling window; ensure that the noise feature vector forms a stable input in time-domain order. The specific implementation steps are as follows:

[0073] Number the constructed noise feature vector NS according to the arrangement order of the corresponding time-domain sampling windows in the original data matrix;

[0074] Use the serialization module to store each NS in a fixed format to ensure the unity of the storage format;

[0075] Ensure that the NS sequence after serialization storage forms a continuous and stable input in time-domain order, providing a data basis for subsequent watermark embedding and copyright authentication.

[0076] Step S4: Based on the local video statistical parameters of the video frames in the client audio-visual data, dynamically modulate the amplitude and position of the time-domain watermark embedding through an adaptive modulation model. The modulation process uses a non-linear function to couple the local video statistical parameters with the watermark embedding amplitude;

[0077] Further explanation: In the audio-visual data, divide the continuous video frame sequence according to fixed time-domain sampling windows, and extract the video frame images corresponding to each time-domain sampling window;

[0078] Divide the video frame image into multiple regions, denoted as \(\{1, 2, \ldots, r, \ldots, R\}\), where \(r\) represents the \(r\)th divided region in the video frame image, and \(R\) is the total number of divided regions;

[0079] Perform local light intensity analysis on each divided region of the video frame image within each time-domain sampling window, and calculate the local video statistical parameters of each divided region. The local video statistical parameters include the local light intensity mean and variance and the magnitude of the motion vector ;

[0080] It should be noted that use the image processing algorithm to calculate the average light intensity of all pixel points in region \(r\), and record it as ; Calculate the variance of the light intensity values of all pixel points within the calculation region r, and record it as ;

[0081] Perform optical flow calculation on the pixel motion between adjacent video frame images to obtain the motion vectors of each region r;

[0082] Calculate and record the magnitude of the motion vector of region r as ;

[0083] Use the optical flow method to process the pixel motion between consecutive video frame images, and extract the magnitude of the motion vector of each region within each time-domain sampling window;

[0084] Detect the light intensity change regions in the video frame images. By setting the light intensity gradient edge detection threshold, divide the light intensity change regions into significantly light intensity changing regions and smoothly light intensity changing regions;

[0085] It should be noted that: for each divided region r, use the Sobel operator of the edge detection algorithm to calculate its light intensity gradient; set a preset light intensity gradient threshold ;

[0086] Regard the regions where the light intensity change is greater than or equal to the light intensity gradient threshold as significantly light intensity changing regions, and the rest as smoothly light intensity changing regions;

[0087] Record the light intensity change classification results of each region, and mark them as the significantly changing region set and the smoothly changing region set respectively;

[0088] Define the adaptive modulation model to include:

[0089] Mark the regions where the magnitude of the motion vector of the pixel regions between video frame images is lower than the preset threshold, and make a comprehensive comparison with the light intensity change regions. Prioritize and select the regions where the light intensity change is smooth and the magnitude of the motion vector is lower than the preset threshold as the target positions for watermark embedding, and record these regions as the target position region set, denoted as {1, 2,..., r1,..., R1}, where r1 represents the index of the target position for watermark embedding, R1 is the total number of target positions for watermark embedding, and r1 ∈ {1, 2,..., R}, R1 ≤ R;

[0090] In this embodiment, the preset threshold of the magnitude of the motion vector is set as ;

[0091] Dynamically exclude the regions where the magnitude of the motion vector of the pixel regions between video frame images is higher than the preset threshold; to ensure that there will be no loss or artifacts during the compression or editing process after watermark embedding.

[0092] Taking the local light intensity mean, variance, and motion vector amplitude calculated for each region in the target position area as inputs, a coupling function between the watermark embedding amplitude and these parameters is constructed using the non - linear least - squares regression method, forming the following adaptive modulation model of the target position r1 with respect to the watermark embedding amplitude:

[0093] ;

[0094] Wherein, is the preliminary watermark embedding amplitude modulated by the target position r1 under the time - domain variable t, is the basic watermark amplitude of the target position r1; is a non - linear function related to the local light intensity mean and variance at the target position r1; In this embodiment is a quadratic polynomial or exponential model, is a non - linear function related to the motion vector amplitude at the target position r1, and α, β are coupling coefficients; Obtained by fitting through on - site experiments; In this embodiment the non - linear function uses the log or sigmoid function.

[0095] It should be noted that:

[0096] Define Adopting the quadratic polynomial form, the specific formula is as follows:

[0097] ;

[0098] Wherein: is the local light intensity mean of the target position r1;

[0099] is the local light intensity variance of the target position r1;

[0100] 、 and c1 are regression coefficients obtained by fitting according to on - site experimental data through the non - linear least - squares regression method;

[0101] Define Adopting the Sigmoid function form, the specific formula is as follows:

[0102] ;

[0103] Wherein, is the motion vector amplitude of the target position r1; is the regression coefficient obtained by fitting according to on - site experimental data through the non - linear least - squares regression method.

[0104] Determine the regression coefficients 、 、c1 and Specific steps:

[0105] Collect on-site experimental data, including , and values and the corresponding ideal watermark embedding amplitudes ;

[0106] Construct a least squares regression model, and estimate the regression coefficients , , c1 and such that the fitting functions f and g can best match the experimental data and minimize the prediction error.

[0107] This embodiment gives the following test contents:

[0108] The experiment takes the audio-visual data of multiple clients as the main test object, selects video materials under different dynamic scenarios for analysis and testing to verify the innovation and advantages of the time-domain watermark embedding mechanism based on local video statistical parameters; the selected video materials include "city street night scene video", "indoor static interview video", "sports event video", etc., covering high-dynamic and low-dynamic scenarios. The experiment uses an optical flow detection, edge detection and adaptive modulation model coupling mechanism to analyze the watermark embedding behavior and record relevant data. The specific steps and implementation process are as follows:

[0109] Video sampling and division:

[0110] Divide the continuous video frame sequence of the experimental video sample according to a 50-millisecond time-domain sampling window to ensure that the sampling time-domain window has sufficient accuracy and stability.

[0111] Each video frame image is evenly divided into a 5×5 regional grid, obtaining 25 regions, and each region is labeled from 1 to 25. These regions provide a basis for subsequent local statistical parameter and target location area selection.

[0112] Extraction of local statistical parameters:

[0113] Light intensity analysis: Calculate the local light intensity values of the pixel points in each region, and extract the light intensity mean and variance. The light intensity mean describes the overall brightness level of the region, and the variance reflects the fluctuation range of the pixel values.

[0114] Motion vector calculation: Calculate the motion vectors between consecutive frames by the optical flow method, and extract the motion amplitude of each region as a dynamic change index.

[0115] Effectively record the light intensity mean, variance and motion vector data in the experimental design, and normalize them to the numerical interval [0,1].

[0116] Apply gradient edge detection and analyze the changes in local motion vectors in different video scenes by combining with the optical flow method.

[0117] Screen the areas where the light intensity change is stable and the amplitude of the motion vector is lower than the preset threshold (0.4) as the target position areas for watermark embedding. The significant areas are excluded to ensure storage stability.

[0118] Adaptive modulation model construction and watermark embedding:

[0119] Based on the mean value, variance of the light intensity and the amplitude of the motion vector in the target position area, establish a dynamic watermark modulation model through a non - linear coupling function.

[0120] Real - time modulate the embedding amplitude and position of the watermark in the target position area, and adopt serialized storage to form a stable and verifiable watermark.

[0121] The data includes the amplitude modulation range of the watermark after embedding, the integrity of the video image after watermark generation, and the retention of the watermark after compression and clipping.

[0122] Call and test with different dynamic scenes, and record the embedding efficiency, error rate and watermark robustness.

[0123] The following table records the local statistical parameters of the experimental video and the verification results after watermark embedding:

[0124] Table 1 Innovation and advantages of the time - domain watermark embedding mechanism for local video statistical parameters:

[0125]

[0126] The experimental results show that:

[0127] The dynamic modulation mechanism of the watermark embedding amplitude is based on the local statistical parameters of the video frame, which can adaptively adjust the embedding range for different dynamic scenes and effectively improve the robustness.

[0128] The embedded watermark maintains a high integrity after compression and clipping, verifying the stability of the adaptive modulation model in the present invention.

[0129] Compared with the traditional watermark embedding method based on the overall frame parameters, this method selects the target position area through local light intensity and motion vector, effectively reducing the risk of embedding position deviation and information loss, and reflecting high innovation.

[0130] Step S5: Combine the noise feature vector NS with the adaptive modulation model and generate a unique watermark key through a determined non - linear mapping function;

[0131] Further explanation: The non-linear mapping function non-linearly couples the noise characteristics and local video statistical parameters based on the time-domain information to generate a watermark key in a fixed format; denote the watermark key as ; The specific implementation steps include:

[0132] Predetermine the following non-linear mapping function:

[0133]

[0134] where are the components in the noise feature vector NS, and δ are constant parameters obtained by fitting offline experimental data; is the fusion coefficient;

[0135] Substitute the generated noise feature vector NS and the calculated illumination and motion parameters into the non-linear mapping function in sequence to calculate a unique and fixed-length watermark key ;

[0136] Combine the watermark key with the time-stamp data related to the time domain to ensure the stability and uniqueness of the generated key in the entire time domain;

[0137] Store the watermark key into the watermark key module; and input the watermark key module as a key parameter for the subsequent time-domain watermark embedding algorithm;

[0138] In this embodiment, the determination method of the non-linear mapping function is as follows:

[0139] Determine the constant parameters , δ and the fusion coefficient in the non-linear mapping function by fitting offline experimental data.

[0140] Collect an experimental data set containing multiple groups of noise feature vectors NS and corresponding local video statistical parameters, including , , and values of multiple target position regions;

[0141] Use the non-linear least squares regression method to fit and estimate the constant parameters and δ in the non-linear mapping function based on the collected experimental data. The specific steps are as follows:

[0142] Construct an objective function, defined as the sum of the squares of the fitting errors;

[0143] Adopt the gradient descent method or other optimization algorithms to iteratively adjust the parameters , δ and the fusion coefficient until the objective function reaches the minimum value;

[0144] Verify the fitting result to ensure that the parameters are consistent in different target position regions, and guarantee the unity and stability of the non-linear mapping function.

[0145] Step S6: In the process of time-domain hidden watermark embedding, couple the watermark key with the watermark modulation parameters. This coupling process involves adjusting the watermark embedding position and amplitude within the time-domain data window of the audio-visual data.

[0146] Further explanation: The audio-visual data includes online course videos or audio data. In the online course videos or audio data, continuous time-domain data windows are divided; in this embodiment, each time-domain data window corresponds to the audio-visual data within 10 milliseconds.

[0147] The watermark modulation parameters include the preliminary watermark embedding amplitude and the time-domain variable t;

[0148] Use the generated watermark key as a dynamic seed, and perform time-domain modulation calculations with the preliminary watermark embedding amplitude and the local video statistical parameters respectively to determine the final watermark embedding amplitude and the final embedding position within each time-domain data window.

[0149] Explanation of the final watermark embedding amplitude:

[0150] Within each time-domain data window, perform dynamic calculations on the preliminary watermark embedding amplitude according to the modulation function where is a non-linear modulation function based on the watermark key and the time-domain variable t; is the final watermark embedding amplitude in the target position r1 region;

[0151] Explanation of the final embedding position:

[0152] Within each time-domain data window, combine and and perform the following calculations according to the modulation function:

[0153] ;

[0154] where is the position screening factor in the target position r1 region; select the region with the smallest value from the target position region set {1, 2,..., r1,..., R1} as the final embedding position;

[0155] Specifically, obtain the position screening factor of all regions r1 from the target position region set {1, 2, …, r1, …, R1} ;

[0156] Compare the values of each region, and select the region with the smallest numerical value as the final watermark embedding position;

[0157] If there are multiple regions with the same and smallest value, select the region with the smallest index in the matrix as the final embedding position;

[0158] Ensure that only one final embedding position is selected within each time-domain data window to ensure the consistency and uniqueness of watermark embedding.

[0159] It should be noted that the smallest numerical value indicates that it is necessary to and / or be smaller; furthermore, it indicates that the corresponding local light intensity mean and variance are smaller, and the pixel brightness change in the corresponding region is smoother, which is conducive to the selection of the watermark embedding position;

[0160] Determine the mathematical expression of the non-linear modulation function , and the specific formula is as follows:

[0161] ;

[0162] Among them, is the unique watermark key generated under the time-domain variable t; t is the time variable of the current time-domain data window, in milliseconds;

[0163] For each time-domain data window t, perform the following operations:

[0164] Obtain the final watermark embedding amplitude ; obtain the final embedding position r2, r2 ∈ {1, 2, …, R1};

[0165] Within the time-domain data window t of the original audio-visual data, for region r2, embed the watermark signal S(t) according to the following formula:

[0166] ;

[0167] Among them, is the signal of the original audio-visual data within the time-domain data window t;

[0168] S(t) is the pre-defined watermark signal; is the audio-visual data after embedding the watermark.

[0169] Ensure that the embedding process is performed only for the final embedding position r2 within each time-domain data window to avoid signal distortion caused by multiple embeddings.

[0170] Embed the watermark into the signal with the modulated final watermark embedding amplitude and the final embedding position in an invisible manner into the original audio-visual data, and form the final watermark-embedded video with the watermark embedding result;

[0171] And send the watermark-embedded video to the client through the data transmission interface for online course copyright authentication.

[0172] It should be noted that: all calculation formulas in this application document adopt regression analysis including but not limited to machine learning algorithms to deeply analyze the relevant parameters collected, identify their natural trends and interrelationships. Using professional software, such as the Scikit-learn library of Python or the R language, automatically generate a mathematical model that matches the data. Then, objectively evaluate the model performance through methods such as cross-validation, and combine continuous feedback and optimization to ensure that the created formula truly reflects the internal laws of the data, thereby ensuring its effectiveness and accuracy. In all calculation formulas of this application, the parameters in each formula are processed by dimensionless normalization within a consistent range to ensure that different physical quantities are compared on the same scale; the dimensionless technical means include but not limited to Min-Max Normalization and Z-Score standardization;

[0173] The technical solution of the present invention can be embodied in the form of a software product in essence or the part that contributes to the prior art. This computer software product can be stored in a computer-readable storage medium, such as a floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disc of a computer, etc., including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0174] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a defined sequence of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

[0175] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

[0176] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for online course copyright authentication based on time-domain hidden watermark, characterized in that, The specific steps include: Step S1: Collect the microscopic noise signals in the display driving circuit at the client side. The microscopic noise signals are digital noise data sequences obtained by continuous sampling within a time-domain sampling window. Step S2: Filter and normalize the collected digital noise data sequences according to a unified time-domain block, and then extract the preliminary noise statistical feature sequences of the digital noise data sequences. Step S3: Use the principal component analysis method to reduce the dimension of the extracted preliminary noise statistical feature sequences to obtain the noise feature vector NS. Step S4: Based on the local video statistical parameters of the video frames in the client audio-visual data, dynamically modulate the amplitude and position of the time-domain watermark embedding through an adaptive modulation model. The modulation process uses a non-linear function to couple the local video statistical parameters with the watermark embedding amplitude. Step S5: Combine the noise feature vector NS with the adaptive modulation model, and generate a unique watermark key through a determined non-linear mapping function. Step S6: In the process of time-domain hidden watermark embedding, couple the watermark key with the watermark modulation parameters. The coupling process involves adjusting the watermark embedding position and amplitude within the time-domain data window of the audio-visual data.

2. The online course copyright authentication method based on time-domain hidden watermark according to claim 1, wherein: The microscopic noise signals include thermal noise signals and random current noise signals. Perform digital conversion on the collected microscopic noise signals according to a fixed time-domain sampling window to obtain a continuous digital noise data sequence. Cache the digital noise data sequences within each time-domain sampling window obtained by continuous sampling.

3. The method for online course copyright authentication based on time-domain hidden watermark according to claim 2, wherein: For the digital noise data sequences cached in each time-domain sampling window, use a digital low-pass filter for filtering to eliminate high-frequency interference clutter. Perform normalization processing on the filtered digital noise data sequences, and the processing result falls within a preset numerical range [0, 1]. Within each time-domain sampling window, calculate the digital noise data sequences by combining the edge detection algorithm and the moving average method to generate preliminary noise statistical feature sequences. The statistical features include but are not limited to mean, variance, and extreme values.

4. The online course copyright authentication method based on time-domain hidden watermark according to claim 3, characterized in that: Convert the preliminary noise statistical feature sequences into noise feature vectors with a fixed dimension. Specifically, the selected principal components are constructed into a noise feature vector NS with a fixed length, that is ; represents the noise feature component of the k-th principal component; Serialize and store the constructed noise feature vector NS in the order of each time-domain sampling window.

5. The online course copyright authentication method based on time-domain hidden watermark according to claim 4, wherein: In the audio-visual data, divide the continuous video frame sequences according to a fixed time-domain sampling window, and extract the video frame images corresponding to each time-domain sampling window. Divide the video frame images into multiple regions, denoted as {1, 2,..., r,..., R}, where r represents the r-th divided region in the video frame image, and R is the total number of divided regions. Perform local light intensity analysis on each divided region of the video frame image within each time-domain sampling window, and calculate the local video statistical parameters of each divided region. The local video statistical parameters include the local light intensity mean and variance as well as the magnitude of the motion vector ; Use the optical flow method to process the pixel motion between continuous video frame images, and extract the motion vector amplitudes of each region within each time-domain sampling window. Detect the light intensity change regions of the video frame images, and divide the light intensity change regions into light intensity change significant regions and light intensity change stable regions by setting the light intensity gradient edge detection threshold.

6. The method for online course copyright authentication based on time-domain hidden watermark according to claim 5, wherein: Define the adaptive modulation model as including: Mark the regions where the magnitude of the motion vector in the pixel region between video frame images is lower than the preset threshold, and comprehensively compare them with the regions of light intensity change. Prioritize the regions where the light intensity change is stable and the magnitude of the motion vector is lower than the preset threshold as the target positions for watermark embedding, and denote these regions as the target position region set, denoted as {1, 2, …, r1, …, R1}, where r1 represents the index of the target position for watermark embedding, R1 is the total number of target positions for watermark embedding, and r1 ∈ {1, 2, …, R}, R1 ≤ R; Dynamically exclude the regions where the magnitude of the motion vector in the pixel region between video frame images is higher than the preset threshold; Taking the local light intensity mean, variance, and motion vector magnitude calculated for each region in the target position region set as inputs, use the non-linear least squares regression method to construct a coupling function between the watermark embedding amplitude and these parameters, and form the following adaptive modulation model of the target position r1 with respect to the watermark embedding amplitude: ；  Among them, is the preliminary watermark embedding amplitude of the target position r1 modulated under the time domain variable t, is the watermark base amplitude of the target position r1; is a non-linear function related to the local light intensity mean and variance at the target position r1; is a non-linear function related to the motion vector amplitude at the target position r1, and α and β are coupling coefficients.

7. The method for online course copyright authentication based on time-domain hidden watermark according to claim 6, characterized in that: The non-linear mapping function non-linearly couples the noise characteristics and local video statistical parameters based on the time-domain information to generate a watermark key in a fixed format; The generated noise feature vector NS and the calculated and are successively substituted into the non-linear mapping function to calculate a unique and fixed-length watermark key .

8. The method for online course copyright authentication based on time-domain hidden watermark according to claim 7, characterized in that: The audio-visual data includes online course videos or audio data. In the online course videos or audio data, continuous time-domain data windows are divided; The watermark modulation parameters include a preliminary watermark embedding amplitude and a time-domain variable t; The generated watermark key is used as a dynamic seed and subjected to time-domain modulation calculations respectively with the preliminary watermark embedding amplitude and local video statistical parameters to determine the final watermark embedding amplitude and the final embedding position within each time-domain data window; Explanation of the final watermark embedding amplitude: Within each time-domain data window, set the modulation function Perform dynamic calculation on the preliminary watermark embedding amplitude where is a non-linear modulation function based on the watermark key and the time-domain variable t; is the final watermark embedding amplitude in the target position r1 region; Explanation of the final embedding position: Within each time-domain data window, in combination with and perform the following calculations according to the modulation function: ； Among them, is the position screening factor for the target position r1 area; select from the target position area set {1, 2, …, r1, …, R1} The area with the smallest numerical value is used as the final embedding position; Embed the watermark into the signal with the modulated final watermark embedding amplitude and the final embedding position are embedded into the original audio-visual data in an invisible manner, and the watermark embedding result forms the final watermark-embedded video; And send the watermark-embedded video to the client through the data transmission interface for online course copyright authentication.

Images