A data resource right confirmation method

Abstract

The application provides a data resource right confirmation determination method, comprising the following steps: S1, obtaining to-be-confirmed data, a confirmed data set and a similar data set, and extracting feature vectors in a preset manner, wherein the confirmed data set comprises a plurality of confirmed data, and the similar data set comprises a plurality of pairs of data, each pair of data comprising confirmed data and corresponding similar data; S2, obtaining a similarity threshold, wherein the similarity threshold is calculated by taking the feature vector of the data as a state, the similarity threshold as an action, and whether the similarity between the confirmed data and the corresponding similar data is greater than or equal to the similarity threshold as a value feedback reward, and the similarity threshold is obtained by training a reinforcement learning model with the similar data set until convergence; S3, calculating the similarity between the feature vector of the to-be-confirmed data and the feature vector of each confirmed data, and confirming based on the similarity threshold. By using the similarity threshold determined by the reinforcement learning method, the problem of confirmation accuracy is solved.

Description

Technical Field

[0001] This invention relates to the field of data transactions, specifically to the field of data resource ownership confirmation, and more specifically, to a method for data resource ownership confirmation based on a similarity threshold obtained through reinforcement learning. Background Technology

[0002] Various data resources are a crucial foundation of the digital age, and establishing standardized data element trading venues and a multi-tiered data market trading system are essential means to leverage data resources' support for the digital age. However, data resources generally have relatively complex ownership relationships, often involving privacy, ownership protection, and profit distribution during use. Therefore, data transactions require clarifying data ownership relationships, i.e., data rights confirmation. In other words, data rights confirmation is a crucial foundation for achieving the legal use of data and the effective protection of its value. Through data rights confirmation, the rights and interests of the data owners with confirmed ownership can be protected, and the theft and misuse of data can be prevented.

[0003] In data ownership confirmation practice, a crucial step is to monitor the similarity between the data to be confirmed and the data already confirmed. If the data to be confirmed and the data already confirmed are determined to be similar, it indicates that the data to be confirmed has been confirmed and belongs to the same owner; otherwise, the data to be confirmed has not yet been confirmed. Existing technical solutions mainly use methods based on Hamming distance and methods based on hashing for similarity determination.

[0004] In methods for similarity determination based on Hamming distance, Hamming distance is an effective similarity metric used to measure the number of different characters between two strings of equal length. By representing the data to be assigned rights and the data already assigned rights as binary strings and calculating the number of different bits between the binary strings, the Hamming distance can be obtained, and then the degree of similarity between the data can be determined based on the Hamming distance threshold.

[0005] Hash-based similarity determination methods map data to a hash space to obtain hash values, thereby efficiently retrieving data while judging the similarity between data to be determined and data already determined based on hash value thresholds. For example, techniques such as Locality-Sensitive Hashing (LSH) can map similar data to similar hash values, thus determining data similarity based on hash value thresholds.

[0006] In practical applications, data privacy protection in setting similarity thresholds and the judgment process is a critical issue. Since there is no unified standard for judging data similarity, thresholds are often set manually or selected from several sets of data, depending on the specific scenario. This manual selection of similarity thresholds can affect the accuracy of similarity judgments, especially when the data is highly heterogeneous.

[0007] For methods that determine similarity based on Hamming distance, the similarity threshold setting affects when two data points are classified as similar or dissimilar. If the threshold is set too low, it may misclassify data that should be dissimilar as similar, increasing the error rate. Conversely, if the similarity threshold is set too high, it may classify actually similar data as dissimilar, reducing the sensitivity of the judgment.

[0008] Similarity determination methods based on hashing also face the challenge of determining the appropriate similarity threshold. Different hash functions and algorithms can lead to different hash value distributions, so setting a similarity threshold requires careful consideration of how to define a suitable boundary within the hash space. An excessively high threshold might cause similar data to be classified as dissimilar, while an excessively low threshold could result in incorrect similarity determinations.

[0009] Therefore, an effective method for determining the similarity threshold and a corresponding method for confirming data resource ownership are needed. Summary of the Invention

[0010] Therefore, the purpose of this invention is to overcome the shortcomings of the prior art and provide a data resource ownership confirmation method based on a similarity threshold determined by reinforcement learning.

[0011] The objective of this invention is achieved through the following technical solution:

[0012] According to a first aspect of the present invention, a method for confirming the ownership of data resources is provided, for confirming the ownership of data to be confirmed based on a similarity threshold, characterized in that the method includes: S1, acquiring the data to be confirmed, a dataset of already confirmed data, and a similar dataset, and extracting feature vectors of the data to be confirmed, the dataset of already confirmed data, and the similar dataset according to a preset method, wherein the dataset of already confirmed data includes multiple data sets of already confirmed data, and the similar dataset includes multiple pairs of data, each pair of data including data set of already confirmed data and its corresponding similar data; S2, acquiring a similarity threshold, wherein the similarity threshold is obtained by calculating a reward based on the feature vector of the data as the state, the similarity threshold as the action, and whether the similarity between the data set of already confirmed data and its corresponding similar data is greater than or equal to the similarity threshold, and training a reinforcement learning model with the similar dataset until convergence; S3, calculating the similarity between the feature vector of the data to be confirmed and the feature vector of each data set of already confirmed data, and confirming the ownership based on the similarity threshold, wherein if the similarity between the data to be confirmed and at least one of all data set of already confirmed data is greater than or equal to the similarity threshold, then the data to be confirmed has been confirmed.

[0013] In some embodiments of the present invention, in step S1, the feature vectors of the data to be confirmed, the data set with confirmed rights, and the similar dataset are extracted in a preset manner as follows: S11, the data to be confirmed, the data set with confirmed rights, and the similar dataset are preprocessed according to preset rules; S12, the preprocessed data is hashed to obtain the feature vectors.

[0014] In some embodiments of the present invention, in step S11, the preset rule is as follows: when the data is text data, the text in the text dataset is segmented and stop words are removed to obtain feature words. The word frequency of each feature word is counted, and the feature word weight is obtained by combining the feature word itself and the multidimensional features in the context. When the data is image data, the images in the image dataset are uniformly scaled to the same size, and the images are subjected to Discrete Cosine Transform (DCT) to convert the images from the pixel domain to the frequency domain. Based on the image converted to the frequency domain, the low-frequency information part of the image is extracted and combined with the color information of the image to obtain the color-spectrum matrix of the low-frequency part of the image.

[0015] In some embodiments of the present invention, in step S12, the preprocessed data is hashed in the following manner to obtain a feature vector: when the data is text data, for the feature words of the text dataset, a preset hash algorithm is selected to generate hash values ​​according to the data features, the generated hash values ​​are converted into weighted hash values ​​according to the weight of the feature words, and the weighted hash values ​​of all feature words are superimposed and merged, and a fixed-length text feature vector is obtained after dimensionality reduction; when the dataset is an image dataset, the color-spectrum matrix is ​​subjected to binary hashing and merged into a fixed-length image feature vector.

[0016] In some embodiments of the present invention, the reinforcement learning model includes a policy network and a value function network, wherein the policy network and the value function network are configured as Siamese networks, wherein the Siamese network includes a Siamese CNN network.

[0017] In some embodiments of the present invention, the reward function of the reinforcement learning model includes a data similarity accuracy reward:

[0018] Reward_m(i,j)=k1*(T-Diff(i,j))

[0019] in,

[0020]

[0021] Where Reward_m(i,j) represents the data similarity accuracy reward, k1 is the weight value of the positive reward for data similarity accuracy, T is the similarity threshold, n represents the nth pair of similar data in the similar dataset, and J (n) (i,j) represents the similarity of the nth pair of similar data in the similar dataset, N represents the number of similar data pairs in the similar dataset, and Diff(i,j) represents the average difference between the similarity of N pairs of similar data in the similar dataset and the similarity threshold.

[0022] In some embodiments of the present invention, in step S2, the reward is calculated by using the feature vector of the data as the state, the similarity threshold and the privacy loss threshold as the action, whether the similarity between the data with confirmed ownership and its corresponding similar data is greater than or equal to the similarity threshold, and whether the privacy loss of the data with confirmed ownership and its corresponding similar data is greater than or equal to the privacy loss threshold, and the reinforcement learning model is trained with similar datasets until convergence to obtain the final similarity threshold.

[0023] In some embodiments of the present invention, the reward function of the reinforcement learning model is:

[0024] Reward_total(i,j)=Reward_m(i,j)+Reward_p(i,j)

[0025] in,

[0026]

[0027] Where Reward_total(i,j) represents the reward function of the reinforcement learning model, Reward_m(i,j) represents the data similarity accuracy reward, Reward_p(i,j) represents the data privacy protection reward, k2 is the weight value of the positive reward for data privacy protection, L represents the privacy loss threshold, and PrivacyLoss... (n) (i,j) represents the privacy loss of the Nth pair of similar data in the similar dataset, where N represents the number of similar data pairs in the similar dataset.

[0028] In some embodiments of the present invention, the privacy loss is differential privacy loss, mutual information loss, or KL divergence loss.

[0029] In some embodiments of the present invention, the similarity is Hamming distance, hash value, or Jaccard similarity.

[0030] According to a second aspect of the present invention, a computer-readable storage medium is provided, characterized in that a computer program is stored thereon, the computer program being executable by a processor to implement the steps of the method as described in the first aspect of the present invention.

[0031] According to a third aspect of the present invention, an electronic device is provided, characterized in that it comprises: one or more processors; and a memory, wherein the memory is configured to store executable instructions; the one or more processors are configured to perform the steps of the method as described in the first aspect of the present invention by executing the executable instructions.

[0032] Compared with the prior art, the advantages of the present invention are as follows:

[0033] 1. This method proposes a data resource ownership confirmation method based on a similarity threshold determined by reinforcement learning. By eliminating the influence of human factors in the process of determining the similarity threshold, the generated similarity threshold is more generalizable, thereby ensuring the accuracy of data resource ownership confirmation.

[0034] 2. This method also uses hash processing technology to convert the feature information of the data into feature vectors, thereby hiding the original data information and ensuring data privacy during the data ownership confirmation process. Attached Figure Description

[0035] The embodiments of the present invention will be further described below with reference to the accompanying drawings, wherein:

[0036] Figure 1 This is a flowchart illustrating the data resource ownership determination process according to an embodiment of the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the invention.

[0038] As mentioned in the background section, the similarity threshold used in existing data resource ownership confirmation methods is manually set. Because there is no unified standard, inaccuracies may occur, potentially affecting the accuracy of similarity determination and consequently the accuracy of data resource ownership confirmation. To address the shortcomings of existing methods, this invention proposes a resource ownership confirmation method based on a similarity threshold determined through reinforcement learning, thereby ensuring the accuracy of resource ownership confirmation. In summary, as... Figure 1 As shown, a data resource ownership confirmation method of the present invention includes the following steps: S1, acquiring data to be confirmed, a confirmed dataset, and a similar dataset, and extracting feature vectors of the data to be confirmed, the confirmed dataset, and the similar dataset according to a preset method, wherein the confirmed dataset includes multiple confirmed data, and the similar dataset includes multiple pairs of data, each pair including confirmed data and its corresponding similar data; S2, acquiring a similarity threshold, wherein the similarity threshold is calculated by using the feature vector of the data as the state, the similarity threshold as the action, and whether the similarity between the confirmed data and its corresponding similar data is greater than or equal to the similarity threshold as the value feedback reward, and using a reinforcement learning model trained on the similar dataset until convergence; S3, calculating the similarity between the feature vector of the data to be confirmed and the feature vector of each confirmed data, and making an ownership confirmation judgment based on the similarity threshold, wherein if the similarity between the data to be confirmed and at least one of all confirmed data is greater than or equal to the similarity threshold, then the data to be confirmed has been confirmed. The data resource ownership confirmation method of the present invention adopts a similarity threshold determined by reinforcement learning, which avoids the accidental factors that may be brought about by manually setting the similarity threshold, and makes the data resource ownership confirmation results more accurate.

[0039] To better understand the present invention, the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0040] In step S1, the data to be confirmed, the confirmed data set, and the similar data set are obtained, and feature vectors are extracted from these datasets according to a preset method. The confirmed data set includes multiple confirmed data sets, and the similar data set includes multiple pairs of data, each pair including confirmed data and its corresponding similar data. To protect data privacy, the feature vectors are obtained using the following preset method: S11, preprocessing the data to be confirmed, the confirmed data set, and the similar data set according to preset rules; S12, hashing the preprocessed data to obtain the feature vectors.

[0041] According to an embodiment of the present invention, in step S11, the preset rule is as follows: when the data is text data, the text in the text dataset is segmented and stop words are removed to obtain feature words. The word frequency of each feature word is counted, and the feature word weight is obtained by combining the feature word itself and the multidimensional features in the context. When the data is image data, the images in the image dataset are uniformly scaled to the same size, and the images are subjected to Discrete Cosine Transform (DCT) to convert the images from the pixel domain to the frequency domain. Based on the image converted to the frequency domain, the low-frequency information part of the image is extracted and combined with the color information of the image to obtain the color-spectrum matrix of the low-frequency part of the image.

[0042] According to an embodiment of the present invention, in step S12, the preprocessed data is hashed in the following manner to obtain a feature vector: when the data is text data, for the feature words of the text dataset, a preset hash algorithm is selected to generate hash values ​​according to the data features, the generated hash values ​​are converted into weighted hash values ​​according to the weight of the feature words, and the weighted hash values ​​of all feature words are superimposed and merged to obtain a text feature vector of fixed length after dimensionality reduction; when the dataset is an image dataset, the color-spectrum matrix is ​​subjected to binary hashing and merged into an image feature vector of fixed length.

[0043] Furthermore, in step S2, a similarity threshold is obtained. This similarity threshold is calculated by using the feature vector of the data as the state, the similarity threshold as the action, and whether the similarity between the weighted data and its corresponding similar data is greater than or equal to the similarity threshold as the value feedback for reward calculation. This is obtained by training a reinforcement learning model to convergence using a similar dataset. As is well known, reinforcement learning models generally include a policy network and a value function network. The policy network learns a policy function that takes the state as input and outputs an action. The value function network evaluates the value feedback of the state to calculate the reward and update the network. In this invention, the reinforcement learning model uses the feature vector of the data as the state, the similarity threshold as the action, and whether the similarity between the weighted data and its corresponding similar data is greater than or equal to the similarity threshold as the value feedback for reward calculation, thus learning the similarity threshold. The purpose of the reinforcement learning model is to learn the similarity threshold; its essence is to obtain a reasonable threshold by learning the similarity between samples.

[0044] According to one embodiment of the present invention, a Siamese network is configured for learning the policy network and the value function network. While many networks exist for similarity learning, the Siamese network structure minimizes the distance between similar samples and maximizes the distance between dissimilar samples during training, thereby learning a good similarity metric. Therefore, in this invention, the Siamese network is considered the optimal candidate for constructing a reinforcement learning model. Furthermore, the Siamese network includes a Siamese CNN network.

[0045] According to one embodiment of the present invention, the reward function of the reinforcement learning model includes a data similarity accuracy reward:

[0046] Reward_m(i,j)=k1*(T-Diff(i,j))

[0047] in,

[0048]

[0049] Where Reward_m(i,j) represents the data similarity accuracy reward, k1 is the weight value of the positive reward for data similarity accuracy, T is the similarity threshold, N represents the nth pair of similar data in the similar dataset, and J (n) (i,j) represents the similarity of the nth pair of similar data in the similar dataset, N represents the number of similar data pairs in the similar dataset, and Diff(i,j) represents the average difference between the similarity of N pairs of similar data in the similar dataset and the similarity threshold.

[0050] Furthermore, according to an embodiment of the present invention, considering that data may be biased after privacy processing (hashing), and that reinforcement learning uses privacy-processed feature vectors for learning, the bias may affect the accuracy of the similarity threshold. Therefore, the present invention introduces privacy loss in the reinforcement learning process to avoid such bias affecting the accuracy of the similarity threshold. Specifically, after introducing privacy loss, in step S2, the reward is calculated using the feature vector of the data as the state, the similarity threshold and the privacy loss threshold as the action, and whether the similarity between the confirmed data and its corresponding similar data is greater than or equal to the similarity threshold and whether the privacy loss between the confirmed data and its corresponding similar data is greater than or equal to the privacy loss threshold as value feedback. The reinforcement learning model is trained on a similar dataset until convergence to obtain the final similarity threshold. The reward function of the reinforcement learning model is:

[0051] Reward_total(i,j)=Reward_m(i,j)+Reward_p(i,j)

[0052] in,

[0053]

[0054] Where Reward_total(i,j) represents the reward function of the reinforcement learning model, Reward_m(i,j) represents the data similarity accuracy reward, Reward_p(i,j) represents the data privacy protection reward, k2 is the weight value of the positive reward for data privacy protection, L represents the privacy loss threshold, and PrivacyLoss... (n) (i,j) represents the privacy loss for the nth pair of similar data in the similar dataset, where N represents the number of similar data pairs in the similar dataset. The privacy loss can be differential privacy loss, mutual information loss, or KL divergence loss; the similarity can be Hamming distance, hash value, or Jaccard similarity.

[0055] According to an embodiment of the present invention, the training of the above-mentioned reinforcement learning model specifically includes the following steps:

[0056] 1. Initialize the policy network and value function network.

[0057] 2. Define hyperparameters, such as the learning rate, the weight parameters k1 and k2 in the reward function, and the total number of training steps.

[0058] 3. Enter the training loop. Each step of the loop is as follows:

[0059] a. Obtain the current state s (data feature vector) from the environment.

[0060] b. Use a policy network to select action a (similarity threshold and privacy loss threshold), use a softmax function to generate a probability distribution, and then randomly sample an action from the distribution.

[0061] c. Execute action a to obtain the next state s′.

[0062] d. Calculate the reward function by considering both the accuracy of data matching and the degree of data privacy protection:

[0063] - Calculate the data matching accuracy reward: Based on the previously defined reward function, use similar datasets (to calculate similarity) and a similarity threshold to calculate the reward Reward_m.

[0064] - Calculate the reward for the degree of data privacy protection: Similarly, based on the previously defined reward function, use similar datasets (to calculate privacy loss) and a privacy loss threshold to calculate the reward Reward_p.

[0065] e. Calculate cumulative rewards

[0066] Reward_total=Reward_m+Reward_p

[0067] f. Use a value function network to evaluate the value function value V(s) of the current state.

[0068] g. Calculate the Advantage function A = Reward_total - V(s), which represents the advantage of action a relative to the current state.

[0069] h. Calculate the policy gradient of the policy network: in This represents the gradient of the policy network with respect to action a.

[0070] i. Update the parameters of the policy network using the policy gradient of the policy network: Where α is the learning rate.

[0071] j. Update the parameters of the value function network using the value function gradient of the value function network: Where β is the learning rate of the value function network.

[0072] k. Update the current state to the next state s′.

[0073] 4. Repeat step 3 until the predetermined number of training steps is reached or the convergence condition is met.

[0074] It should be noted that although the steps are described in a specific order above, it does not mean that the steps must be executed in the above specific order. In fact, some of these steps can be executed concurrently, or even in a different order, as long as the required function can be achieved.

[0075] This invention can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the invention.

[0076] Computer-readable storage media can be tangible devices that hold and store instructions for use by an instruction execution device. Computer-readable storage media can be, for example, including but not limited to, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof.

[0077] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for confirming the ownership of data resources, used to confirm the ownership of data to be confirmed based on a similarity threshold, characterized in that, The method includes: S1. Obtain the data to be confirmed, the confirmed data set, and the similar data set, and extract the feature vectors of the data to be confirmed, the confirmed data set, and the similar data set according to a preset method. The confirmed data set includes multiple confirmed data sets, and the similar data set includes multiple pairs of data, each pair including confirmed data and its corresponding similar data. The feature vectors extracted from the data to be confirmed, the confirmed data set, and the similar data set according to the preset method are as follows: S11. Preprocess the data to be confirmed, the confirmed data set, and the similar data set according to preset rules. The preset rules are as follows: when the data is text data, segment the text in the text dataset and remove stop words to obtain feature words, count the word frequency of each feature word, and combine the feature word itself and the multi-dimensional features in the context to obtain the weight of the feature word; when the data is image data, scale the images in the image dataset to the same size and perform Discrete Cosine Transform (DCT) on the images to convert the images from the pixel domain to the frequency domain. Based on the converted image, extract the low-frequency information part of the image and combine it with the color information of the image to obtain the color-spectrum matrix of the low-frequency part of the image. S12. Perform hash processing on the preprocessed data to obtain the feature vector; S2. Obtain the similarity threshold, wherein the similarity threshold is calculated by using the feature vector of the data as the state, the similarity threshold as the action, and whether the similarity between the weighted data and its corresponding similar data is greater than or equal to the similarity threshold as the value feedback to calculate the reward. The similarity threshold is obtained by training a reinforcement learning model with a similar dataset until convergence. S3. Calculate the similarity between the feature vector of the data to be assigned rights and the feature vector of each data already assigned rights, and assign rights based on the similarity threshold. If the similarity between the data to be assigned rights and at least one of all data already assigned rights is greater than or equal to the similarity threshold, then the data to be assigned rights has been assigned rights.

2. The method according to claim 1, characterized in that, In step S12, the preprocessed data is hashed as follows to obtain the feature vector: When the data is text data, for the feature words of the text dataset, a preset hash algorithm is selected to generate hash values ​​according to the data features. The generated hash values ​​are converted into weighted hash values ​​according to the weight of the feature words. The weighted hash values ​​of all feature words are superimposed and merged. After dimensionality reduction, a text feature vector of fixed length is obtained. When the dataset is an image dataset, the color-spectrum matrix is ​​subjected to binary hashing and merged into a fixed-length image feature vector.

3. The method according to claim 1, characterized in that, The reinforcement learning model includes a policy network and a value function network, wherein the policy network and the value function network are configured as Siamese networks, and the Siamese network includes a Siamese CNN network.

4. The method according to claim 1, characterized in that, The reward function of the reinforcement learning model includes a data similarity accuracy reward: in, in, Rewards are given for data similarity accuracy. It is a weighted value for positive rewards based on the accuracy of data similarity. For similarity threshold, Indicates the first in a similar dataset For similar data, Indicates the first in a similar dataset Similarity of similar data This indicates the number of similar data pairs in a similar dataset. Indicating similar datasets The average difference between the similarity score and the similarity threshold for similar data.

5. The method according to claim 1, characterized in that, In step S2, the reward is calculated by taking the feature vector of the data as the state, the similarity threshold and the privacy loss threshold as the action, and whether the similarity between the data with confirmed ownership and its corresponding similar data is greater than or equal to the similarity threshold and whether the privacy loss of the data with confirmed ownership and its corresponding similar data is greater than or equal to the privacy loss threshold as the value feedback. The reinforcement learning model is trained with similar datasets until convergence to obtain the final similarity threshold.

6. The method according to claim 5, characterized in that, The reward function of the reinforcement learning model is: in, in, This represents the reward function of the reinforcement learning model. Rewards are given for data similarity accuracy. This indicates a reward for data privacy protection. It is a weighted value for positive rewards related to data privacy protection. Indicates the threshold for privacy loss. Indicates the first in a similar dataset Privacy loss due to similar data This indicates the number of similar data pairs in a similar dataset.

7. The method according to claim 6, characterized in that, Privacy loss can be classified as differential privacy loss, mutual information loss, or KL divergence loss.

8. The method according to claim 1, characterized in that, Similarity can be measured by Hamming distance, hash value, or Jaccard similarity.

9. A computer-readable storage medium, characterized in that, It contains a computer program that can be executed by a processor to implement the steps of the method according to any one of claims 1 to 8.

10. An electronic device, characterized in that, include: One or more processors; as well as Memory, wherein the memory is used to store executable instructions; The one or more processors are configured to implement the steps of the method according to any one of claims 1 to 8 by executing the executable instructions.

Images