Multimedia information fusion push method and system based on 5G messages

Abstract

The present invention relates to the field of information push technology, and discloses a method and system for multimedia information fusion push based on 5G messages. The method comprises: performing sequence expansion on a user behavior sequence of a target user to obtain a target behavior sequence; extracting local behavior features and global behavior features of the target behavior sequence, fusing the local behavior features and the global behavior features to obtain user behavior features; calculating a multimedia interest resource set based on the user behavior features, filtering the interest resource set to obtain a multimedia resource set; extracting hierarchical features of each resource in the multimedia resource set, and constructing target features based on the hierarchical features; compiling a 5G message signal based on the target features, and pushing multimedia information based on the 5G message signal. The present invention can improve the accuracy of multimedia information push.

Description

Technical Field

[0001] The present invention relates to the field of information push technology, and in particular to a method and system for multimedia information fusion push based on 5G messages. Background Art

[0002] With the rapid development of the communications industry, the integration of big data and 5G communication technologies in multimedia messaging is becoming increasingly widespread and in-depth. Compared to traditional SMS, 5G messaging supports multiple media formats, including text, images, audio, video, location, contacts, and documents. It can precisely push multimedia messages based on multi-dimensional data such as user characteristics and behavioral habits, ensuring that users receive personalized and tailored information in different scenarios. This also meets the diverse needs of businesses, such as precision marketing and service delivery, and expands the reach of push notifications.

[0003] However, multimedia information from different sources differs in format, encoding, resolution, etc., which requires a lot of format conversion and adaptation work when merging data, increasing technical difficulty and complexity. In addition, information loss or quality degradation may occur during the conversion process. At the same time, it is difficult to fully tap into users' personalized preferences and needs when pushing messages, and the push content is not targeted and accurate enough, resulting in users receiving a large amount of multimedia information they are not interested in, which reduces information utilization efficiency and user satisfaction.

[0004] Therefore, how to improve the accuracy of multimedia information push becomes an urgent problem to be solved. Summary of the Invention

[0005] The present invention provides a method and system for multimedia information fusion push based on 5G messages, the main purpose of which is to solve the problem of poor accuracy in multimedia information push.

[0006] Obtaining a user behavior sequence of a target user, performing sequence expansion on the user behavior sequence to obtain a target behavior sequence;

[0007] Extracting local behavior features and global behavior features of the target behavior sequence, and fusing the local behavior features with the global behavior features to obtain user behavior features;

[0008] Calculating a multimedia resource set of interest to the target user based on the user behavior characteristics, and filtering the resource set of interest to obtain a multimedia resource set;

[0009] Extracting hierarchical features of each resource in the multimedia resource set, and constructing target features of the multimedia resource set based on the hierarchical features;

[0010] A 5G message signal is compiled according to the target characteristics, and multimedia information is pushed according to the 5G message signal.

[0011] Optionally, the performing sequence expansion on the user behavior sequence to obtain a target behavior sequence includes:

[0012] Reversely sorting the user behavior sequence to obtain a reverse behavior sequence;

[0013] Constructing a sequence embedding matrix of the reverse behavior sequence, and calculating a behavior preference score for each user behavior in the user behavior sequence according to the sequence embedding matrix;

[0014] Select the user behavior with the largest behavior preference score and add it to the user behavior sequence to obtain an updated behavior sequence;

[0015] The updated behavior sequence is used as a user behavior sequence for iteration until the number of iterations is greater than a preset threshold, thereby obtaining a target behavior sequence.

[0016] Optionally, extracting local behavior features and global behavior features of the target behavior sequence includes:

[0017] Constructing a behavior node graph according to the target behavior sequence, performing self-attention calculation on the behavior node graph to obtain local behavior features;

[0018] Identify neighbor nodes of each behavior node in the behavior node graph and calculate the attention coefficient of the neighbor nodes;

[0019] Perform feature aggregation on the neighbor nodes according to the attention coefficient to obtain node aggregation features;

[0020] The behavior nodes are updated according to the node aggregation features to obtain global behavior features.

[0021] Optionally, fusing the local behavior features and the global behavior features to obtain user behavior features includes:

[0022] Performing feature splicing on the local behavior features to obtain local fusion features;

[0023] Performing feature weighted averaging on the global behavior features to obtain global fusion features;

[0024] The local fusion features and the global fusion features are summed and pooled to obtain user behavior features.

[0025] Optionally, the calculating the multimedia interest resource set of the target user according to the user behavior characteristics includes:

[0026] Calculating the user similarity between the target user and users in a preset user set based on the user behavior characteristics;

[0027] Constructing a similar user set based on the user similarity, and obtaining the interest resource scores corresponding to the similar user set;

[0028] Calculating a user resource score of the target user based on the user similarity and the interest resource score;

[0029] A multimedia resource set of interest to the target user is determined according to the user resource score.

[0030] Optionally, extracting the hierarchical features of each resource in the multimedia resource set includes:

[0031] Performing vector conversion and type classification on the multimedia resource set to obtain text resource vectors and other resource vectors;

[0032] Performing multi-layer encoding on the text resource vector to obtain multi-layer encoding features;

[0033] Performing mean pooling on the multi-layer coding features to obtain hierarchical pooling features, and performing feature splicing on the hierarchical pooling features to obtain text hierarchical features;

[0034] Performing multi-layer convolution pooling on the other resource vectors to obtain pooling features, and performing full connection processing on the pooling features to obtain other hierarchical features;

[0035] The text hierarchical features and the other hierarchical features are aggregated to obtain hierarchical features of the multimedia resource set.

[0036] Optionally, constructing the target feature of the multimedia resource set according to the hierarchical feature includes:

[0037] Constructing a feature vector sequence according to the hierarchical features, performing attention calculation on the feature vector sequence to obtain an attention feature;

[0038] Performing a residual connection between the attention feature and the target feature to obtain a connection feature;

[0039] Regularizing the connection features to obtain regularized features, and performing full-connection feedforward calculation on the regularized features to obtain feedforward features;

[0040] Performing a residual connection on the feedforward feature and the regularization feature to obtain an initial target feature;

[0041] Regularization is performed on the initial target features to obtain target features of the multimedia resource set.

[0042] Optionally, the compiling of a 5G message signal according to the target feature:

[0043] performing channel coding on the target feature to obtain coded information;

[0044] Performing signal modulation on the coded information to obtain a communication signal of the coded information;

[0045] performing wavelet transform on the communication signal to obtain wavelet coefficients;

[0046] Calculating a coefficient threshold of the communication signal according to the wavelet coefficient, and extracting a target coefficient of the wavelet coefficient according to the coefficient threshold;

[0047] The target coefficients are reconstructed using wavelet transform to obtain a 5G message signal.

[0048] Optionally, performing signal modulation on the coded information to obtain a communication signal of the coded information includes:

[0049] Determining a transmission distance and a channel quality indicator corresponding to the coded information;

[0050] When the transmission distance is greater than a preset communication threshold, performing constant envelope data modulation processing on the coded information to obtain a constant envelope modulated signal;

[0051] Delay compensation processing is performed on the constant envelope modulation signal according to the channel quality indicator to obtain a communication signal.

[0052] In order to solve the above problems, the present invention also provides a multimedia information fusion push system based on 5G messages, the system comprising:

[0053] A sequence expansion module is used to obtain a user behavior sequence of a target user, perform sequence expansion on the user behavior sequence, and obtain a target behavior sequence;

[0054] A feature information fusion module is used to extract local behavior features and global behavior features of the target behavior sequence, and fuse the local behavior features and the global behavior features to obtain user behavior features;

[0055] A multimedia resource set calculation module is used to calculate the multimedia resource set of the target user according to the user behavior characteristics, and filter the resource set of interest to obtain a multimedia resource set;

[0056] a target feature construction module, configured to extract hierarchical features of each resource in the multimedia resource set and construct target features of the multimedia resource set based on the hierarchical features;

[0057] The information push module is used to compile a 5G message signal according to the target characteristics and push multimedia information according to the 5G message signal.

[0058] The embodiment of the present invention expands the user behavior sequence of the target user, learns the dependency relationship between different user behaviors, and obtains a more comprehensive and accurate target behavior sequence of user behavior; extracts the local behavior features and global behavior features of the target behavior sequence, and performs information fusion to obtain user behavior features, which can obtain richer feature information and further understand the user's behavioral motivation and intention more comprehensively; calculates a multimedia interest resource set based on the user behavior features, filters the resources, obtains a multimedia resource set, and removes bad information to ensure the security and accuracy of information push; extracts the hierarchical features of each resource, constructs the target features based on the hierarchical features, and can fuse the feature information in the multimedia resource set so that the target features contain the feature information of each resource in the multimedia resource set, providing a basis for subsequent multimedia information push; compiles a 5G message signal based on the target features, and pushes multimedia information based on the 5G message signal, which can accurately push the multimedia resource set to the target user, effectively improve the quality of information push, and further improve the accuracy of multimedia information push. Therefore, the multimedia information fusion push method and system based on 5G messages proposed in the present invention can solve the problem of poor accuracy of multimedia information push. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A flowchart of a method for fusion pushing multimedia information based on 5G messages provided in one embodiment of the present invention;

[0060] Figure 2 A schematic diagram of a process for extracting local behavior features and global behavior features of a target behavior sequence according to an embodiment of the present invention;

[0061] Figure 3 A schematic diagram of a process for calculating a target user's multimedia interest resource set according to an embodiment of the present invention;

[0062] Figure 4 This is a functional module diagram of a multimedia information fusion push system based on 5G messages provided in one embodiment of the present invention.

[0063] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0064] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0065] The embodiment of the present application provides a multimedia information fusion push method based on 5G messages. The execution subject of the multimedia information fusion push method based on 5G messages includes but is not limited to at least one of the electronic devices such as the server, the terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the multimedia information fusion push method based on 5G messages can be executed by software or hardware installed on the terminal device or the server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and basic cloud computing services such as big data and artificial intelligence platforms.

[0066] Reference Figure 1 FIG2 is a flow chart of a method for merging and pushing multimedia information based on 5G messages according to an embodiment of the present invention. In this embodiment, the method for merging and pushing multimedia information based on 5G messages includes:

[0067] S1. Obtain a user behavior sequence of a target user, perform sequence expansion on the user behavior sequence, and obtain a target behavior sequence.

[0068] In an embodiment of the present invention, a user behavior sequence is a collection of various types of information generated by a target user while using various platform functions and interacting with the platform, sorted by generation time. This may include the target user's dwell time when browsing products, reflecting the user's level of interest in a particular product; the target user's search keywords for products, reflecting the target user's demand direction; the target user's frequency of music playback, reflecting the target user's musical preferences; and may also include the frequency and time pattern of the target user's location sharing. The user behavior sequence contains rich information about user preferences. Based on the user behavior sequence, a user's potential preferences and interests can be analyzed in multiple dimensions. Furthermore, from a vast amount of project resources, projects that may be of interest to the target user can be accurately recommended.

[0069] Specifically, sequence expansion is to supplement the information in the user behavior sequence to prevent poor accuracy of information push caused by too little behavior information in the user behavior sequence of the target user.

[0070] In the embodiment of the present invention, the step of performing sequence expansion on the user behavior sequence to obtain a target behavior sequence includes:

[0071] Reversely sorting the user behavior sequence to obtain a reverse behavior sequence;

[0072] Constructing a sequence embedding matrix of the reverse behavior sequence, and calculating a behavior preference score for each user behavior in the user behavior sequence according to the sequence embedding matrix;

[0073] Select the user behavior with the largest behavior preference score and add it to the user behavior sequence to obtain an updated behavior sequence;

[0074] The updated behavior sequence is used as a user behavior sequence for iteration until the number of iterations is greater than a preset threshold, thereby obtaining a target behavior sequence.

[0075] In this embodiment of the present invention, the reversed order of the user behavior sequence is reversed to obtain a reversed behavior sequence. A sequence embedding matrix is a representation that maps the reversed behavior sequence to a low-dimensional vector space. Pre-built recurrent neural networks (RNNs) or Transformer architectures can be used to convert the elements in the reversed behavior sequence into corresponding sequence embedding matrices using the corresponding initialization embedding matrices.

[0076] Furthermore, the behavior preference score of each user behavior is calculated by multiplying the sequence embedding matrix with the learnable weight matrix of each user behavior using a bidirectional self-attention architecture to obtain the behavior preference score of each user behavior. The user phase with the largest behavior preference score is added to the user behavior sequence to obtain an updated behavior sequence. The updated behavior sequence is iterated to obtain the target behavior sequence.

[0077] Among them, the bidirectional self-attention architecture includes a multi-head self-attention mechanism (Multi-HeadAttention) with residual connections and a feed-forward neural network (Feed-Forward Neural Network). Through the bidirectional self-attention architecture, the dependencies between different user behaviors can be learned, and then the user behavior sequence can be predicted to obtain a more comprehensive and accurate target behavior sequence of user behaviors.

[0078] S2. Extract local behavior features and global behavior features of the target behavior sequence, and fuse the local behavior features and the global behavior features to obtain user behavior features.

[0079] In an embodiment of the present invention, the local behavior feature is the behavior transition feature information contained in each individual target behavior in the target behavior sequence, while the global behavior feature is the cross-behavior feature information between target behaviors, thereby achieving more fine-grained feature extraction.

[0080] Specifically, see Figure 2 As shown, the extracting of local behavior features and global behavior features of the target behavior sequence includes:

[0081] S21, constructing a behavior node graph according to the target behavior sequence, performing self-attention calculation on the behavior node graph to obtain local behavior features;

[0082] S22. Identify neighbor nodes of each behavior node in the behavior node graph and calculate the attention coefficient of the neighbor nodes;

[0083] S23, performing feature aggregation on the neighbor nodes according to the attention coefficient to obtain node aggregation features;

[0084] S24: Update the characteristics of the behavior nodes according to the node aggregation characteristics to obtain global behavior characteristics.

[0085] In an embodiment of the present invention, a behavior node graph is a node graph constructed with each target behavior in the target behavior sequence as a node, wherein the edges between nodes include unidirectional and bidirectional edges. A unidirectional edge indicates that there is a unidirectional transition between nodes in the target behavior sequence, for example, node 1 and node 2 are adjacent in the target behavior sequence. A bidirectional edge indicates that there is a bidirectional transition between nodes in the target behavior sequence, for example, node 1 and node 2 are adjacent in both directions in the target behavior sequence, for example, the positions of node 1 and node 2 in the target behavior sequence are (…1,2…2,1…).

[0086] Furthermore, self-attention calculation calculates the attention coefficient of each behavior node based on the behavior node graph using the self-attention mechanism, and multiplies the attention coefficient with the node feature element-wise to obtain the local behavior feature.

[0087] In detail, the neighbor nodes are the user behaviors of each behavior node in the target behavior sequence that are separated by a preset sequence interval. The ground features of each neighbor node are feature spliced, as well as the feature point multiplication results between each behavior node and the neighbor node. The feature splicing result and the feature point multiplication result are multiplied with the preset first feature weight, and activated using the Relu activation function to obtain the attention coefficient of each neighbor node. The attention coefficient of the neighbor node is feature multiplied with the node feature of each behavior node respectively and then added to obtain the final node aggregation feature.

[0088] In detail, a preset graph neural network (GNN) can be used to update the node features of the behavior node and the node aggregation features of the neighbor nodes to obtain the global behavior features that include the node itself and its neighbor nodes.

[0089] In the embodiment of the present invention, information fusion is to fuse feature information representing transition feature information of a single target behavior with cross-behavior feature information between target behaviors, thereby learning comprehensive user behavior features.

[0090] Specifically, fusing the local behavior features and the global behavior features to obtain user behavior features includes:

[0091] Performing feature splicing on the local behavior features to obtain local fusion features;

[0092] Performing feature weighted averaging on the global behavior features to obtain global fusion features;

[0093] The local fusion features and the global fusion features are summed and pooled to obtain user behavior features.

[0094] In an embodiment of the present invention, feature splicing is to connect the local behavior features of non-overnight behavior nodes on the same dimension to form a higher-dimensional feature representation, which can preserve all feature information intact and learn the complex associations between different local behavior features.

[0095] The weighted summation of features is to multiply multiple global behavioral features by the corresponding weight coefficients and then add them together to obtain a comprehensive feature representation, which can highlight more important features and suppress less important features, and help improve the accuracy of subsequent calculation of multimedia interest resource sets.

[0096] Furthermore, the summation pooling process is to perform pooling processing using the SumPooling function, which divides the local fusion features and the global fusion features into several regions, then sums the features in each region, and finally concatenates the summation results of all regions as user behavior features.

[0097] In an embodiment of the present invention, by fusing local behavior features and global behavior features, richer feature information can be obtained, thereby more comprehensively understanding the user's behavioral motivations and intentions, avoiding the one-sidedness caused by focusing only on local behavior, and effectively improving the accuracy of subsequent interest resource calculations.

[0098] S3. Calculate a multimedia resource set of interest to the target user based on the user behavior characteristics, and filter the resource set of interest to obtain a multimedia resource set.

[0099] In the embodiment of the present invention, multimedia interest resources are push resources that may be of interest to the user and are presented in various forms (such as video, audio, image, text, etc.).

[0100] Specifically, see Figure 3 As shown, the calculation of the multimedia interest resource set of the target user based on the user behavior characteristics includes:

[0101] S31, calculating the user similarity between the target user and users in a preset user set based on the user behavior characteristics;

[0102] S32: construct a similar user set based on the user similarity, and obtain the interest resource scores corresponding to the similar user set;

[0103] S33, calculating the user resource score of the target user according to the user similarity and the interest resource score;

[0104] S34: Determine a multimedia resource set of interest to the target user based on the user resource score.

[0105] In an embodiment of the present invention, the preset user set can be a set of users who have interacted with the target user, for example, users communicating on a social network platform, members of an interest group, netizens on a forum, etc. The user similarity between users is calculated using user behavior characteristics, wherein the feature similarity calculation can be performed based on the user behavior characteristics of the users in the user set and the user behavior characteristics of the target user to obtain the user similarity.

[0106] Furthermore, a preset number of users with the greatest user similarity are selected to construct a similar user set, and the historical ratings of multimedia resources by each similar user in the similar user set are obtained to obtain the interest resource rating, wherein the interest resource rating is a quantitative rating given by the user to the project (such as products, services, applications, courses, etc.) based on his or her own usage experience, satisfaction and other relevant factors, and is usually presented in numerical form, such as 1 to 5 stars, 1 to 10 points, etc.

[0107] Specifically, the user resource score of the target user can be calculated using the following formula: in, Indicates target users Resources User resource ratings, Indicates the target user Users in the similar user set The similarity between users Represents a user Resources Interest resource ratings, Represents the total number of users in the similar user set.

[0108] Specifically, multimedia resources corresponding to user resource scores are sorted from high to low to obtain a multimedia resource sequence, and a preset number of multimedia resources with the highest scores are selected from the multimedia resource sequence to construct a multimedia interest resource set that requires information push.

[0109] Specifically, resource filtering is to filter out bad information in the multimedia interest resource set to ensure the security and accuracy of multimedia information push.

[0110] In detail, based on the bad vocabulary corpus, a homophone corpus and a variant word corpus corresponding to the bad vocabulary can be established, and the corpus can be matched with keywords in the multimedia interest resource set. Multimedia interest resources with a preset threshold of matching quantity are extracted from the multimedia interest resource set to obtain a multimedia resource set.

[0111] Specifically, the multimedia resource collection can be processed, the text of each resource in the multimedia resource collection can be extracted, and function words and punctuation marks can be removed according to a common function word corpus. The frequency of the phrases formed after the processing can be counted. According to a set number of keywords, the top keywords are sequentially extracted to obtain the keywords of each multimedia resource in the multimedia resource collection.

[0112] S4. Extracting hierarchical features of each resource in the multimedia resource set, and constructing target features of the multimedia resource set according to the hierarchical features.

[0113] In an embodiment of the present invention, hierarchical features are hierarchical feature information of each resource, including low-level and high-level features. Low-level features can intuitively display the information of the resource, and high-level features can more directly reflect the semantic information and intrinsic meaning of each resource, thereby comprehensively performing feature analysis through low-level features and high-level features.

[0114] Specifically, extracting the hierarchical features of each resource in the multimedia resource set includes:

[0115] Performing vector conversion and type classification on the multimedia resource set to obtain text resource vectors and other resource vectors;

[0116] Performing multi-layer encoding on the text resource vector to obtain multi-layer encoding features;

[0117] Performing mean pooling on the multi-layer coding features to obtain hierarchical pooling features, and performing feature splicing on the hierarchical pooling features to obtain text hierarchical features;

[0118] Performing multi-layer convolution pooling on the other resource vectors to obtain pooling features, and performing full connection processing on the pooling features to obtain other hierarchical features;

[0119] The text hierarchical features and the other hierarchical features are aggregated to obtain hierarchical features of the multimedia resource set.

[0120] In an embodiment of the present invention, vector conversion can be performed on each resource in a multimedia resource collection through a pre-built encoding module. For example, different types of vector conversions can be performed on resources of different modalities. For example, text resources can be converted into vectors using Word2Vec, images and videos can be converted into vectors through convolutional neural networks, and audio can be converted into vectors through waveform coding.

[0121] Furthermore, the vector converted from the multimedia resource set vector is divided to obtain a text resource vector corresponding to the text resource and other resource vectors corresponding to other resources.

[0122] In detail, the text resource vector is encoded using a multi-layer Transformer encoder to obtain multi-layer encoding features of each layer of encoding, and then the encoding features of each layer in the multi-layer encoding features are mean pooled to obtain hierarchical pooling features, and the hierarchical pooling features are feature spliced to obtain text hierarchical features.

[0123] Specifically, multi-layer convolutional pooling involves sequentially performing Dense Block multi-layer convolution processing, convolution layer, and pooling layer processing on other resource vectors to obtain pooled features at each layer. The pooled features at each layer are then fully connected to obtain other hierarchical features with consistent feature dimensions. Each Dense Block contains multiple convolutional layers, which are connected together in a specific manner, typically using a 3x3 convolution kernel. Each layer in a Dense Block connects its input to the output of all previous layers along the channel dimension. This Dense Block strengthens the flow and interaction of features between different layers, allowing each layer to obtain richer feature information, thereby improving the feature content of other hierarchical features and obtaining more accurate other hierarchical features.

[0124] In the embodiment of the present invention, by classifying multimedia resource collections into types, feature extraction can be performed for different types of resources, and hierarchical features of different feature information can be extracted. Feature information of multimedia resources can be comprehensively acquired through the hierarchical features.

[0125] In the embodiment of the present invention, constructing the target feature of the multimedia resource set according to the hierarchical feature includes:

[0126] Constructing a feature vector sequence according to the hierarchical features, performing attention calculation on the feature vector sequence to obtain an attention feature;

[0127] Performing a residual connection between the attention feature and the target feature to obtain a connection feature;

[0128] Regularizing the connection features to obtain regularized features, and performing full-connection feedforward calculation on the regularized features to obtain feedforward features;

[0129] Performing a residual connection on the feedforward feature and the regularization feature to obtain an initial target feature;

[0130] Regularization is performed on the initial target features to obtain target features of the multimedia resource set.

[0131] In an embodiment of the present invention, the feature vector sequence is obtained by vector splicing the hierarchical features of each resource to obtain a feature vector sequence composed of multiple hierarchical features, and the feature vector sequence is multiplied by three learnable projection matrices to obtain Q, K, and V vectors. Then, the attention weight of each feature of the feature vector sequence to other features is calculated using Q and K, and the product of the attention weight and V is used as the attention feature.

[0132] Specifically, the feature vector sequence is added to the attention feature through residual connection to obtain the connection feature. Regularization processing refers to imposing certain constraints on the connection feature so that the connection feature can better perform subsequent feature fusion while meeting specific requirements. For example, regularization processing can be performed through orthogonal constraints on the feature vector.

[0133] Furthermore, fully connected feedforward calculations are performed using a pre-built fully connected feedforward network (Fully-ConnectedFeedforward Network). The fully connected feedforward network includes an input layer, a hidden layer, and an output layer. Regularized features enter the network from the input layer, undergo initial processing by the input layer neurons, and are then passed to the hidden layer neurons. In the hidden layer neurons, the regularized features are multiplied by the corresponding weights, a bias is added, and then a nonlinear transformation is performed through the activation function to obtain the output of the neuron. These outputs are then used as inputs to the next layer and continue to propagate forward until the output layer obtains the final prediction results, namely the feedforward features.

[0134] Among them, the neurons in the fully connected feedforward network are connected to each other in a specific way, that is, all neurons in the previous layer are connected to all neurons in the next layer, forming a fully connected relationship, and information is transmitted one-way between neurons, starting from the input layer and passed forward layer by layer to the output layer, without any feedback connection.

[0135] In the embodiment of the present invention, feature information in a multimedia resource set can be integrated by constructing a target feature, so that the target feature includes feature information of each resource in the multimedia resource set, providing a basis for subsequent multimedia information push.

[0136] S5. Compile a 5G message signal according to the target characteristics, and push multimedia information according to the 5G message signal.

[0137] In an embodiment of the present invention, the 5G message signal is a 5G communication signal that can be used for data transmission. Multimedia resources can be transmitted through the 5G message signal, thereby realizing multimedia information push.

[0138] Specifically, the 5G message signal is compiled according to the target characteristics:

[0139] performing channel coding on the target feature to obtain coded information;

[0140] Performing signal modulation on the coded information to obtain a communication signal of the coded information;

[0141] performing wavelet transform on the communication signal to obtain wavelet coefficients;

[0142] Calculating a coefficient threshold of the communication signal according to the wavelet coefficient, and extracting a target coefficient of the wavelet coefficient according to the coefficient threshold;

[0143] The target coefficients are reconstructed using wavelet transform to obtain a 5G message signal.

[0144] In the embodiment of the present invention, channel coding is to encode target characteristics during transmission, thereby improving transmission reliability and accuracy. For example, channel coding can be performed through low-density parity check (LDPC) coding or polarization coding to obtain coded information.

[0145] Furthermore, signal modulation is used to convert coded information into a form suitable for transmission, improving the signal's anti-interference ability and transmission efficiency. By varying the amplitude, frequency, or phase of the coded information, the information is encoded to produce a communication signal suitable for information push. The receiving client can decode the information using the information encoding method to obtain the multimedia resources contained in the target feature.

[0146] Specifically, the signal modulating the coded information to obtain a communication signal of the coded information includes:

[0147] Determining a transmission distance and a channel quality indicator corresponding to the coded information;

[0148] When the transmission distance is greater than a preset communication threshold, performing constant envelope data modulation processing on the coded information to obtain a constant envelope modulated signal;

[0149] Delay compensation processing is performed on the constant envelope modulation signal according to the channel quality indicator to obtain a communication signal.

[0150] Specifically, the transmission distance is the distance between the terminal where the target user is located and the sending base station, and the channel quality index is an indicator of the communication channel quality reflected by the terminal where the target user is located, for example, channel quality indicators such as the current channel signal-to-noise ratio and the current channel capacity.

[0151] Specifically, constant envelope data modulation (CEDDM) is a modulation method in which the amplitude of the modulated signal remains constant, with only the phase or frequency changing. Gaussian filtering of the baseband signal before modulation can improve the signal's spectral characteristics.

[0152] Specifically, delay compensation involves appropriately delaying or advancing the signal based on the channel quality indicator, aligning the signal timing with that under normal delay conditions. For example, when channel quality is poor, such as in environments with strong interference or areas with severe signal fading, the received signal delay may fluctuate significantly. This requires a larger delay compensation range. The equalizer or delay adjustment circuit at the target user's receiving end can appropriately delay or advance the signal to align the signal timing with that under normal delay conditions. Pre-set rules can be used to identify channel quality based on the channel quality indicator.

[0153] Specifically, a request signaling can be sent to obtain the signaling sending time. The signal transmission delay can be calculated based on the preset expected transmission time and the signaling sending time, and then delay compensation processing can be performed based on the signal transmission delay, so that the communication signal can reduce the interference of delay and improve the time synchronization accuracy of multimedia information push.

[0154] Furthermore, the wavelet transform uses the wavelet function to perform wavelet decomposition of the communication signal layer by layer to obtain wavelet decomposition coefficients at different levels. The high-frequency information and low-frequency information in the transformed signal can be obtained through wavelet decomposition. The wavelet coefficients are screened by calculating the appropriate coefficient threshold, and the wavelet decomposition coefficients less than the noise threshold are set to 0. The other signal values remain unchanged, and the communication signal can be denoised.

[0155] Furthermore, the coefficient threshold of the communication signal is calculated using the following formula: in, represents the coefficient threshold, Indicates the signal length of the communication signal, Indicates the maximum number of decomposition levels of wavelet transform, represents the preset heuristic parameters, represents the median of the absolute values of the wavelet coefficients, represents the wavelet coefficients.

[0156] In detail, wavelet reconstruction is the inverse process of wavelet transform. In wavelet transform, the communication signal is decomposed into approximate and detail coefficients of different scales. Wavelet reconstruction can reconstruct the signal through the target coefficients for signal recovery, and recombine the decomposed signal into the original signal to obtain the 5G message signal, and send the multimedia resource collection to the target user through the 5G message signal.

[0157] In an embodiment of the present invention, by compiling a 5G message signal of a multimedia resource set, the signal anti-interference capability and transmission efficiency of the 5G message signal can be improved. By utilizing the 5G message signal multimedia information push, the multimedia resource set is sent to the client corresponding to the target user, thereby accurately pushing the multimedia resource set to the target user, effectively improving the quality of information push, and further improving the accuracy of multimedia information push.

[0158] like Figure 4 As shown in FIG, it is a functional module diagram of a multimedia information fusion push system based on 5G messages provided by an embodiment of the present invention.

[0159] The multimedia information fusion push system 400 based on 5G messages described in the present invention can be installed in an electronic device. Depending on the functions implemented, the multimedia information fusion push system 400 based on 5G messages can include a sequence expansion module 401, a feature information fusion module 402, a multimedia resource collection calculation module 403, a target feature construction module 404, and an information push module 405. The module described in the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can perform fixed functions, which are stored in the memory of the electronic device.

[0160] In this embodiment, the functions of each module / unit are as follows:

[0161] The sequence expansion module 401 is used to obtain a user behavior sequence of a target user, and perform sequence expansion on the user behavior sequence to obtain a target behavior sequence;

[0162] The feature information fusion module 402 is used to extract local behavior features and global behavior features of the target behavior sequence, and fuse the local behavior features and the global behavior features to obtain user behavior features;

[0163] The multimedia resource set calculation module 403 is used to calculate the multimedia resource set of the target user according to the user behavior characteristics, and filter the resource set of interest to obtain a multimedia resource set;

[0164] The target feature construction module 404 is configured to extract hierarchical features of each resource in the multimedia resource set and construct target features of the multimedia resource set based on the hierarchical features;

[0165] The information push module 405 is used to compile a 5G message signal according to the target characteristics and push multimedia information according to the 5G message signal.

[0166] In detail, each module described in the multimedia information fusion push system 400 based on 5G messages in the embodiment of the present invention adopts the same method as above when used. Figures 1 to 3 The technical means are the same as the multimedia information fusion push method based on 5G messages described in, and can produce the same technical effects, so I will not go into details here.

[0167] The present invention also provides an electronic device, which may include a processor, a memory, a communication bus and a communication interface, and may also include a computer program stored in the memory and run on the processor, such as a multimedia information fusion push method program based on 5G messages.

[0168] In some embodiments, the processor may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips.

[0169] The memory includes at least one type of readable storage medium, including a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory may be an internal storage unit of the electronic device, such as a mobile hard disk of the electronic device.

[0170] The communication bus may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory and at least one processor.

[0171] The communication interface is used for communication between the electronic device and other devices, and includes a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a display or an input unit (such as a keyboard). Optionally, the user interface may also be a standard wired interface or a wireless interface.

[0172] The figure only shows an electronic device with components. Those skilled in the art will understand that the structure shown in the figure does not constitute a limitation on the electronic device, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0173] Specifically, the specific implementation method of the processor for the above instructions can refer to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, which will not be repeated here.

[0174] In the several embodiments provided herein, it should be understood that the disclosed devices, systems, and methods may be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the module division is merely a logical functional division, and actual implementation may employ other division methods.

[0175] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0176] Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the claims are intended to be embraced therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0177] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results.

[0178] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or systems recited in a system claim may also be implemented by a single unit or system through software or hardware. Terms such as "first" and "second" are used to indicate names and do not imply any particular order.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multimedia information fusion push method based on 5G messages, characterized in that: The method comprises: Obtain a user behavior sequence of a target user, perform sequence expansion on the user behavior sequence, and obtain a target behavior sequence; wherein, performing sequence expansion on the user behavior sequence to obtain a target behavior sequence includes: reverse sorting the user behavior sequence to obtain a reverse behavior sequence; constructing a sequence embedding matrix of the reverse behavior sequence, and calculating a behavior preference score for each user behavior in the user behavior sequence according to the sequence embedding matrix; selecting the user behavior with the largest behavior preference score and adding it to the user behavior sequence to obtain an updated behavior sequence; iterating the updated behavior sequence as the user behavior sequence until the number of iterations exceeds a preset threshold, thereby obtaining a target behavior sequence; Extracting local behavior features and global behavior features of the target behavior sequence, and fusing the local behavior features and the global behavior features to obtain user behavior features; wherein, extracting the local behavior features and the global behavior features of the target behavior sequence includes: constructing a behavior node graph according to the target behavior sequence, performing self-attention calculation on the behavior node graph to obtain local behavior features; identifying neighbor nodes of each behavior node in the behavior node graph, and calculating the attention coefficient of the neighbor node; performing feature aggregation on the neighbor nodes according to the attention coefficient to obtain node aggregation features; and performing feature update on the behavior node according to the node aggregation features to obtain global behavior features; Calculating a multimedia resource set of interest to the target user based on the user behavior characteristics, and filtering the resource set of interest to obtain a multimedia resource set; Extracting hierarchical features of each resource in the multimedia resource set, and constructing target features of the multimedia resource set based on the hierarchical features; wherein, extracting hierarchical features of each resource in the multimedia resource set includes: performing vector conversion and type classification on the multimedia resource set to obtain text resource vectors and other resource vectors; performing multi-layer encoding on the text resource vectors to obtain multi-layer encoding features; performing mean pooling on the multi-layer encoding features to obtain hierarchical pooling features, performing feature splicing on the hierarchical pooling features to obtain text hierarchical features; performing multi-layer convolution pooling on the other resource vectors to obtain pooling features, performing full connection processing on the pooling features to obtain other hierarchical features; aggregating the text hierarchical features and the other hierarchical features to obtain hierarchical features of the multimedia resource set; A 5G message signal is compiled according to the target characteristics, and multimedia information is pushed according to the 5G message signal.

2. The multimedia information fusion push method based on 5G messages according to claim 1, characterized in that: The fusing the local behavior features and the global behavior features to obtain user behavior features includes: Performing feature splicing on the local behavior features to obtain local fusion features; Performing feature weighted averaging on the global behavior features to obtain global fusion features; The local fusion features and the global fusion features are summed and pooled to obtain user behavior features.

3. The multimedia information fusion push method based on 5G messages according to claim 1, characterized in that: The calculating the multimedia interest resource set of the target user according to the user behavior characteristics includes: Calculating the user similarity between the target user and users in a preset user set based on the user behavior characteristics; Constructing a similar user set based on the user similarity, and obtaining the interest resource scores corresponding to the similar user set; Calculating a user resource score of the target user based on the user similarity and the interest resource score; A multimedia resource set of interest to the target user is determined according to the user resource score.

4. The multimedia information fusion push method based on 5G messages according to claim 1, characterized in that: The constructing the target feature of the multimedia resource set according to the hierarchical feature includes: Constructing a feature vector sequence according to the hierarchical features, performing attention calculation on the feature vector sequence to obtain an attention feature; Performing a residual connection between the attention feature and the target feature to obtain a connection feature; Regularizing the connection features to obtain regularized features, and performing full-connection feedforward calculation on the regularized features to obtain feedforward features; Performing a residual connection on the feedforward feature and the regularization feature to obtain an initial target feature; Regularization is performed on the initial target features to obtain target features of the multimedia resource set.

5. The multimedia information fusion push method based on 5G messages according to claim 1, characterized in that: The 5G message signal is compiled according to the target characteristics: performing channel coding on the target feature to obtain coded information; Performing signal modulation on the coded information to obtain a communication signal of the coded information; performing wavelet transform on the communication signal to obtain wavelet coefficients; Calculating a coefficient threshold of the communication signal according to the wavelet coefficient, and extracting a target coefficient of the wavelet coefficient according to the coefficient threshold; The target coefficients are reconstructed using wavelet transform to obtain a 5G message signal.

6. The multimedia information fusion push method based on 5G messages according to claim 5, characterized in that: The signal modulation of the coded information to obtain a communication signal of the coded information includes: Determining a transmission distance and a channel quality indicator corresponding to the coded information; When the transmission distance is greater than a preset communication threshold, performing constant envelope data modulation processing on the coded information to obtain a constant envelope modulated signal; Delay compensation processing is performed on the constant envelope modulation signal according to the channel quality indicator to obtain a communication signal.

7. A multimedia information fusion push system based on 5G messages, characterized in that: The system comprises: A sequence expansion module is used to obtain a user behavior sequence of a target user, perform sequence expansion on the user behavior sequence, and obtain a target behavior sequence; wherein, when the sequence expansion module performs sequence expansion on the user behavior sequence to obtain a target behavior sequence, it includes: reverse sorting the user behavior sequence to obtain a reverse behavior sequence; constructing a sequence embedding matrix of the reverse behavior sequence, and calculating a behavior preference score for each user behavior in the user behavior sequence based on the sequence embedding matrix; selecting the user behavior with the largest behavior preference score and adding it to the user behavior sequence to obtain an updated behavior sequence; iterating the updated behavior sequence as the user behavior sequence until the number of iterations exceeds a preset threshold number of times to obtain a target behavior sequence; A feature information fusion module is used to extract local behavior features and global behavior features of the target behavior sequence, and fuse the local behavior features and the global behavior features to obtain user behavior features. When extracting the local behavior features and the global behavior features of the target behavior sequence, the feature information fusion module includes: constructing a behavior node graph according to the target behavior sequence, performing self-attention calculation on the behavior node graph to obtain local behavior features; identifying neighbor nodes of each behavior node in the behavior node graph, and calculating the attention coefficient of the neighbor node; performing feature aggregation on the neighbor nodes according to the attention coefficient to obtain node aggregation features; and performing feature update on the behavior node according to the node aggregation features to obtain global behavior features. A multimedia resource set calculation module is used to calculate the multimedia resource set of the target user according to the user behavior characteristics, and filter the resource set of interest to obtain a multimedia resource set; A target feature construction module is used to extract the hierarchical features of each resource in the multimedia resource set, and construct the target features of the multimedia resource set based on the hierarchical features; wherein, when extracting the hierarchical features of each resource in the multimedia resource set, the target feature construction module includes: performing vector conversion and type classification on the multimedia resource set to obtain text resource vectors and other resource vectors; performing multi-layer encoding on the text resource vectors to obtain multi-layer encoding features; performing mean pooling on the multi-layer encoding features to obtain hierarchical pooling features, performing feature splicing on the hierarchical pooling features to obtain text hierarchical features; performing multi-layer convolution pooling on the other resource vectors to obtain pooling features, performing full connection processing on the pooling features to obtain other hierarchical features; and aggregating the text hierarchical features and the other hierarchical features to obtain hierarchical features of the multimedia resource set; The information push module is used to compile a 5G message signal according to the target characteristics and push multimedia information according to the 5G message signal.

Images