Advertisement push method and system based on age and gender recognition

Abstract

The present invention discloses an advertisement push method and system based on age and gender recognition, and relates to the field of intelligent advertisement push technology, including: using a multimodal fusion network to perform polymorphic fusion on the extracted user features, and comprehensively analyzing the user's age and gender information; identifying the advertisement material that best matches the user's age and gender from the advertisement library according to the user's age and gender information; based on the selected advertisement material, using NLP to generate personalized advertisement copy, and using GANs to generate personalized advertisement pictures and videos; combining personalized advertisement copy with personalized advertisement pictures and videos to generate personalized advertisement content. Through multimodal data collection and fusion, the present invention improves the accuracy of age and gender recognition, accurately matches advertisement content, and generates personalized copy and vision, thereby enhancing advertisement appeal and user experience.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent advertisement push, and in particular to an advertisement push method and system based on age and gender recognition. Background Art

[0002] In recent years, with the rapid development of big data and artificial intelligence technology, personalized advertising has become an important means to improve advertising effectiveness and user experience. Traditional advertising systems mainly rely on users' basic information and historical behavior data, and use simple statistical analysis to achieve targeted advertising. However, with the diversification of user needs and the enrichment of advertising content, traditional single-modal data processing methods can no longer meet the needs of accurate advertising. In this context, multimodal information fusion technology has gradually become a research hotspot. By integrating multiple data sources such as users' facial images, voice signals, and gait information, it can more comprehensively analyze users' characteristics, thereby improving the accuracy and personalization of advertising.

[0003] Although there are some systems on the market that recommend ads based on user behavior and preferences, these systems generally have the following shortcomings: First, most of them rely on single-modal data (such as users' historical browsing records or shopping behaviors), ignoring other potentially useful sources of information, resulting in limited accuracy of ad push; second, existing age and gender recognition technologies are mostly based on facial recognition, which is easily affected by lighting conditions, obstructions, and individual differences, and the recognition accuracy is not high, making it difficult to meet the needs of accurate ad push. In addition, the existing ad push system lacks an effective feedback mechanism and cannot adjust the ad content and push strategy in a timely manner based on real-time user feedback, resulting in poor user experience and low ad conversion rate. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an advertisement push method and system based on age and gender recognition to solve the problems of insufficient multimodal information fusion, inaccurate age and gender recognition, and low advertisement push accuracy.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides an advertising push method based on age and gender recognition, which includes collecting user information and performing feature extraction on the collected user information; using a multimodal fusion network to perform polymorphic fusion on the extracted user features and comprehensively analyze the user's age and gender information; identifying, from an advertising library, advertising materials that best match the user's age and gender based on the user's age and gender information; based on the selected advertising materials, using NLP to generate personalized advertising copy, and using GANs to generate personalized advertising images and videos; combining the personalized advertising copy with personalized advertising images and videos to generate personalized advertising content; pushing the personalized advertising content to users, and collecting user feedback in real time, and optimizing the advertising content and push strategy based on the user's real-time feedback.

[0008] As a preferred solution of the method for pushing advertisements based on age and gender recognition of the present invention, the steps of collecting user information and extracting features from the collected user information are as follows:

[0009] The user's facial image is captured by a high-resolution camera installed in a public place, the user's voice is recorded by an integrated microphone array, and the user's gait information is detected using a thermal imager and an infrared sensor;

[0010] Use facial detection algorithms to detect facial regions from images and extract facial features, including eyes, nose, and mouth features;

[0011] Perform noise reduction on the audio signal, use FFT to perform spectrum analysis, and extract speech features, including fundamental frequency and formant frequency;

[0012] The gait information is filtered and smoothed, and the Kalman filter is used to remove noise and extract gait features, including step length, stride length, and speed.

[0013] As a preferred solution of the method for advertising push based on age and gender recognition of the present invention, the following specific steps are used to perform multimodal fusion of the extracted user features and comprehensively analyze the user's age and gender information:

[0014] The extracted facial features, language features and gait features are fused using a multimodal fusion network. The expression is:

[0015] F=[F f ; F v ; F g ];

[0016] Among them, F is the comprehensive feature vector, F f is the facial feature vector, F v is the speech feature vector, F gis the gait feature vector;

[0017] Based on the fused comprehensive feature vector F, the comprehensive perception model MLP is used to comprehensively analyze the user's age and gender information;

[0018] User age analysis, the expression is:

[0019]

[0020] Among them, A is the predicted value of user age, F i is the i-th feature in the comprehensive feature vector F, α i is the amplitude coefficient of the ith feature, d j is the bias term of the jth layer, m is the number of hidden layers, n is the dimension of the comprehensive feature vector F, σ is the activation function, e is the global bias term, tanh is the nonlinear activation function, and Y i is the phase shift of the i-th eigenvector;

[0021] User gender analysis, the expression is:

[0022]

[0023] Among them, β i is the frequency coefficient of the ith feature, and G represents the gender bias rate;

[0024] Based on the comprehensive feature vector F, define a gender discrimination threshold T G , determine the user's gender;

[0025] When G>T G , then it is judged as female;

[0026] When G<T G , it is determined to be male.

[0027] As a preferred solution of the method for pushing advertisements based on age and gender identification of the present invention, the steps of identifying the advertisement material that best matches the age and gender of the user from the advertisement library according to the age and gender information of the user are as follows:

[0028] Based on the product information and creative content, define an ad library containing multiple creatives, each with a target age range and target gender skew associated with it;

[0029] Based on the user's age and gender, a matching function M is defined to identify the matching degree between each ad creative and the user's age and gender information. The expression is:

[0030]

[0031] Among them, q k represents the kth creative, μ k Indicates the creative q k The average of the target age range, & k Indicates the creative q k Standard deviation of the target age range, ρ k Indicates the creative q k The central value of the target gender bias rate, T k Indicates the creative q k The standard deviation of the target gender bias rate, M(q k , A, G) is a measure of the advertising creative q k The matching degree with the user's age prediction value A and gender bias rate G;

[0032] Based on business needs and historical data, define a matching standard threshold T M , identify highly matching advertising creatives;

[0033] When M(q k , A, G) ≥ T M When the current ad creative is considered to be a high-matching creative;

[0034] When M(q k , A, G)<T M , the current creative is considered not to be a high-matching creative.

[0035] As a preferred solution of the advertising push method based on age and gender recognition described in the present invention, wherein: based on the selected advertising material, NLP is used to generate personalized advertising copy, and GANs is used to generate personalized advertising pictures and videos. The specific steps are as follows:

[0036] Extract key information from the selected advertising materials, including product name, functional features and target audience, to form a set of keywords K;

[0037] Generate a copy template according to the type of advertising material and the target audience, fill the relevant information in the keyword set K into the placeholder position in the copy template, and generate preliminary personalized advertising copy;

[0038] Use NLP to polish the generated preliminary copy, including synonym replacement and sentence structure adjustment, to generate personalized advertising copy;

[0039] According to the content and style of the selected creatives, collect relevant reference images and video clips to train the GANs model;

[0040] Use the trained GANs model to generate personalized images and videos that are consistent with the style of the ad creatives.

[0041] As a preferred solution of the method for pushing advertisements based on age and gender recognition described in the present invention, the steps of combining personalized advertisement text with personalized advertisement pictures and videos to generate personalized advertisement content are as follows:

[0042] By coordinating the consistency between the copy and the image content, the generated personalized advertising copy is matched and combined with the personalized advertising image to enhance the visual appeal;

[0043] By embedding the generated personalized advertising copy into personalized videos and presenting the copy content in the form of subtitles and narration, the video's appeal can be improved;

[0044] Once the composition is complete, use Adobe Premiere Pro to export the final personalized advertising content.

[0045] As a preferred solution of the method for pushing advertisements based on age and gender recognition described in the present invention, the personalized advertisement content is pushed to the user, and user feedback is collected in real time. According to the real-time feedback of the user, the advertisement content and push strategy are optimized. The specific steps are as follows:

[0046] Based on the user's active time period, the generated personalized advertising content will be pushed to the target users through social media;

[0047] Through the built-in feedback collection mechanism, users' responses to ads are collected in real time, including click-through rate, dwell time and number of shares, and subjective feedback from users is directly obtained through questionnaires and online evaluations;

[0048] Optimize push timing, frequency and advertising content based on user feedback to increase the effective reach of advertising;

[0049] For users with positive feedback, provide more personalized advertising content; for users with negative feedback, optimize push strategies to reduce interference with users' lives.

[0050] In a second aspect, the present invention provides an advertising push system based on age and gender recognition, including a data collection and feature extraction module, a multimodal fusion and analysis module, an advertising material matching module, an advertising content generation module based on selected advertising materials, and an advertising push and feedback module; the data collection and feature extraction module is used to collect user information and perform feature extraction on the collected user information; the multimodal fusion and analysis module is used to use a multimodal fusion network to perform multimodal fusion on the extracted user features, and comprehensively analyze the user's age and gender information; the advertising material matching module is used to identify the advertising material that best matches the user's age and gender from the advertising library based on the user's age and gender information; based on the selected advertising material: used to use NLP to generate personalized advertising copy, and use GANs to generate personalized advertising pictures and videos; the advertising content generation module is used to combine personalized advertising copy with personalized advertising pictures and videos to generate personalized advertising content; the advertising push and feedback module is used to push personalized advertising content to users, and collect user feedback in real time, and optimize advertising content and push strategies based on the user's real-time feedback.

[0051] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the advertising push method based on age and gender recognition as described in the first aspect of the present invention is implemented.

[0052] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for advertising push based on age and gender recognition as described in the first aspect of the present invention is implemented.

[0053] The beneficial effects of the present invention are as follows: through multimodal data collection and feature extraction, the present invention realizes comprehensive acquisition of user features from facial images, voice signals and gait information, thereby enhancing the diversity and accuracy of data; through multimodal fusion and comprehensive analysis, multi-source heterogeneous data are effectively integrated, thereby improving the accuracy of age and gender recognition; through the selection of highly matching advertising materials, advertising content that best matches user characteristics is accurately screened out, thereby improving the relevance and effectiveness of advertising; through personalized advertising content generation, NLP and GANs technologies are used to generate highly customized copy and visual content, thereby enhancing the attractiveness and interactivity of advertising, ultimately achieving the precision and personalization of advertising push, and greatly improving user experience and advertising conversion rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0055] Figure 1 This is a flow chart of the method for pushing advertisements based on age and gender recognition in Example 1.

[0056] Figure 2 This is a flow chart of the advertisement push system based on age and gender recognition in Example 1. DETAILED DESCRIPTION

[0057] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0058] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0059] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0060] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides an advertisement push method based on age and gender recognition, comprising the following steps:

[0061] S1. Collect user information and perform feature extraction on the collected user information.

[0062] Going a step further, high-resolution cameras installed in public places are used to capture users’ facial images, integrated microphone arrays are used to record users’ voices, and thermal imagers and infrared sensors are used to detect users’ gait information;

[0063] Use facial detection algorithms to detect facial regions from images and extract facial features, including eyes, nose, and mouth features; for example, when a high-resolution camera captures a user's facial image, the system uses facial detection algorithms (such as MTCNN based on deep learning) to locate the facial region and further identify the specific locations of the eyes, nose, and mouth. By analyzing the morphology and position of these key parts, the system can extract features such as the size and spacing of the eyes, the shape and position of the nose, and the outline of the mouth. These facial feature data will be used for subsequent age and gender recognition analysis.

[0064] The audio signal is processed for noise reduction, and the spectrum analysis is performed using FFT to extract speech features, including the fundamental frequency and formant frequency; for example, when the integrated microphone array records the user's voice, the system first performs noise reduction on the audio signal to remove environmental noise. Then, the time domain signal is converted into a frequency domain signal using Fast Fourier Transform (FFT), from which the fundamental frequency of the voice (i.e., the basic frequency of the vocal cord vibration) and the formant frequency (i.e., the resonant frequency of different vowels in the voice) are extracted. These speech features will help the system more accurately determine the user's age and gender.

[0065] The gait information is filtered and smoothed, and the Kalman filter is used to remove noise and extract gait features, including step length, stride length, and speed. For example, after the thermal imager and infrared sensor detect the user's gait information, the system first filters and smoothes the raw data to eliminate noise introduced by environmental interference and sensor errors. Then, the Kalman filter is used to further optimize the data to improve its stability and accuracy. By analyzing the processed data, the system is able to extract the user's step length (the length of each step), stride length (the total length of two consecutive steps), and speed (walking speed). These gait features will provide additional supporting information for age and gender recognition.

[0066] S2. Use the multimodal fusion network to perform multimodal fusion on the extracted user features and comprehensively analyze the user's age and gender information.

[0067] Furthermore, a multimodal fusion network is used to fuse the extracted facial features, language features, and gait features. The expression is:

[0068] F=[F f ; F v ; F g ];

[0069] Among them, F is the comprehensive feature vector, F f is the facial feature vector, F v is the speech feature vector, F g is the gait feature vector;

[0070] It should be noted that the system first expresses the extracted facial features (such as the shape of eyes, nose and mouth), language features (such as fundamental frequency and formant frequency) and gait features (such as step length, stride and speed) as feature vectors Ff, F v and F g Then, a multimodal fusion network is used to fuse these three feature vectors into a comprehensive feature vector F = [F f ; F v ; F g ], the fused feature vector is processed by the multi-layer perceptron (MLP) of the neural network, and finally a comprehensive user feature representation is generated for subsequent age and gender recognition analysis. This process effectively integrates multi-source heterogeneous data through a deep learning model, improving the accuracy and robustness of age and gender recognition.

[0071] Based on the fused comprehensive feature vector F, the comprehensive perception model MLP is used to comprehensively analyze the user's age and gender information;

[0072] User age analysis, the expression is:

[0073]

[0074] Among them, A is the predicted value of user age, F i is the i-th feature in the comprehensive feature vector F, α i is the amplitude coefficient of the ith feature, d j is the bias term of the jth layer, m is the number of hidden layers, n is the dimension of the comprehensive feature vector F, σ is the activation function (Sigmoid function), e is the global bias term, tanh is the nonlinear activation function, and Y i is the phase shift of the i-th eigenvector;

[0075] User gender analysis, the expression is:

[0076]

[0077] Among them, β i is the frequency coefficient of the ith feature, and G represents the gender bias rate;

[0078] Based on the comprehensive feature vector F, define a gender discrimination threshold T G , determine the user's gender;

[0079] When G>T G , then it is judged as female;

[0080] When G<T G , it is determined to be male.

[0081] It should also be noted that the activation function σ maps the weighted input value of each node in the neural network to a nonlinear interval, allowing the model to capture complex patterns in the data. In this process, the activation function plays a role in converting the internal calculation results into a probability distribution that is more consistent with the actual probability distribution, thereby improving the accuracy and robustness of age and gender recognition. For example, in age analysis, the activation function converts the internal calculation results into a value between 0 and 1, which can be interpreted as the probability of the user's age prediction; in gender analysis, the activation function also converts the internal calculation results into a probability value, indicating the probability of the user's gender bias. In this way, the activation function helps the system better understand and identify the user's age and gender characteristics.

[0082] S3. According to the user's age and gender information, identify the advertising creative that best matches the user's age and gender from the advertising library.

[0083] Furthermore, based on product information and creative content, an advertising library containing multiple advertising creatives is defined, and each advertising creative has an associated target age range and target gender bias rate; the specific process is: by analyzing the basic information of the product (such as product name, functional features, brand image, etc.) and creative content (such as copy, pictures and videos), the target age range and target gender bias rate of each advertising creative are determined, thereby building a systematic advertising library to ensure that each advertising creative can accurately match a specific audience group.

[0084] Based on the user's age and gender, a matching function M is defined to identify the matching degree between each ad creative and the user's age and gender information. The expression is:

[0085]

[0086] Among them, q k represents the kth creative, μ k Indicates the creative q k The average of the target age range, & k Indicates the creative q k Standard deviation of the target age range, ρ k Indicates the creative q k The central value of the target gender bias rate, T k Indicates the creative q k The standard deviation of the target gender bias rate, M(q k , A, G) is a measure of the advertising creative q k The matching degree with the user's age prediction value A and gender bias rate G;

[0087] Based on business needs and historical data, define a matching standard threshold T M, identify highly matching advertising creatives;

[0088] When M(q k , A, G) ≥ T M When the current ad creative is considered to be a high-matching creative;

[0089] When M(q k , A, G)<T M , the current creative is considered not to be a high-matching creative.

[0090] Based on the comparison results, the system can screen out highly matched advertising materials that are most suitable for the target users, thereby improving the accuracy and effectiveness of advertising push.

[0091] S4. Based on the selected advertising materials, use NLP to generate personalized advertising copy, and use GANs to generate personalized advertising images and videos;

[0092] Furthermore, key information is extracted from the selected advertising material, including product name, functional features and target audience, to form a set of keyword sets K; for example, from a selected advertising material, the system will extract the product name "smart watch", functional features "health monitoring, sports tracking, smart notifications" and target audience "young professionals". These key information are summarized into a set of keyword sets K, including "smart watch", "health monitoring", "sports tracking", "smart notifications" and "young professionals". This process ensures that the extracted key information can accurately reflect the core content of the advertising material and the characteristics of the target audience.

[0093] Generate a copy template based on the type of advertising material and the target audience, fill the relevant information in the keyword set K into the placeholder position in the copy template, and generate preliminary personalized advertising copy; for example, generate a copy template based on the type of advertising material and the target audience, and the template contains placeholders. Assume that the advertising material is about a smart watch, the target audience is young professionals, and the keyword set K includes "smart watch", "health monitoring", "sports tracking", "smart notification" and "young professionals".

[0094] Use NLP to polish the generated preliminary copy, including synonym replacement and sentence structure adjustment, to generate personalized advertising copy; for example, "Smart watches designed for young professionals, with health monitoring, exercise tracking and smart notifications, making your life more convenient and healthy" is polished to "Smart watches designed for busy young professionals, with health monitoring, exercise tracking and smart notification functions, helping you to easily manage your life and enjoy a healthier every day." By replacing synonyms and adjusting sentence structure, the copy is made more vivid and attractive.

[0095] According to the content and style of the selected creative, relevant reference pictures and video clips are collected to train the GANs model; the specific process is as follows: According to the content and style of the selected creative, the system collects a series of relevant reference pictures and video clips, and then uses these reference materials to train the generative adversarial network (GANs) model to ensure that the generated personalized pictures and videos are consistent with the style of the creative.

[0096] Use the trained GANs model to generate personalized images and videos that are consistent with the style of the ad creatives.

[0097] It should also be noted that through the trained GANs model, corresponding personalized pictures and videos are generated according to the content and style of the advertising creatives, ensuring that the generated visual content is consistent with the style of the advertising creatives, thereby enhancing the attractiveness and relevance of the advertisements.

[0098] S5. combining personalized advertising copy with personalized advertising pictures and videos to generate personalized advertising content;

[0099] Furthermore, by coordinating the consistency between the copy and the image content, the generated personalized advertising copy is matched and combined with the personalized advertising image to enhance the visual appeal;

[0100] By embedding the generated personalized advertising copy into personalized videos and presenting the copy content in the form of subtitles and narration, the video's appeal can be increased. For example, the generated personalized advertising copy "A smart watch designed for busy young professionals, with health monitoring, exercise tracking and smart notification functions, helps you easily manage your life and enjoy a healthier every day" is displayed in the form of subtitles at key moments in the video, accompanied by corresponding narration to ensure that the copy content is coordinated with the video screen, thereby increasing the video's appeal.

[0101] Once the composition is complete, use Adobe Premiere Pro to export the final personalized advertising content.

[0102] It should also be noted that by coordinating the consistency of the copy and the image content, and embedding the personalized advertising copy into the video, presenting it in the form of subtitles and narration, ensuring that the copy is coordinated with the video screen, the video's appeal is improved. Finally, Adobe Premiere Pro was used to export high-quality personalized advertising content, which enhanced the overall consistency and visual appeal of the advertisement, and improved the user experience and advertising effect.

[0103] S6. Push personalized advertising content to users and collect user feedback in real time. Optimize advertising content and push strategies based on real-time user feedback.

[0104] Furthermore, the generated personalized advertising content will be pushed to the target users through social media according to the user's active time period;

[0105] Through the built-in feedback collection mechanism, users' responses to ads are collected in real time, including click-through rate, dwell time and number of shares, and subjective feedback from users is directly obtained through questionnaires and online evaluations;

[0106] Optimize push timing, frequency and advertising content based on user feedback to increase the effective reach of advertising;

[0107] For users with positive feedback, provide more personalized advertising content; for users with negative feedback, optimize push strategies to reduce interference with users' lives.

[0108] It should also be noted that the built-in feedback collection mechanism automatically tracks and records users' immediate reactions to push ads, such as objective data such as click-through rate, page dwell time, and sharing behavior, and collects users' subjective feedback through questionnaires and online evaluations. Subsequently, based on the collected feedback information, the system will dynamically adjust the push time, push frequency, and content itself of the ads to improve the effective reach of the ads. For those users with positive feedback, the system will further provide personalized advertising content that is more in line with their interests, while for users with negative feedback, the push strategy will be adjusted to reduce the push frequency or change the push method, thereby reducing the disturbance to users and improving the user experience. This mechanism not only improves the user experience, but also enhances the system's adaptive ability, enabling it to continuously improve and optimize the ad push strategy, thereby improving the overall advertising effect and user satisfaction.

[0109] The present embodiment also provides an advertisement push system based on age and gender recognition, including: a data collection and feature extraction module, a multimodal fusion and analysis module, an advertisement material matching module, an advertisement content generation module based on selected advertisement materials, and an advertisement push and feedback module; the data collection and feature extraction module is used to collect user information and perform feature extraction on the collected user information; the multimodal fusion and analysis module is used to use a multimodal fusion network to perform multimodal fusion on the extracted user features and comprehensively analyze the user's age and gender information; the advertisement material matching module is used to identify the advertisement material that best matches the user's age and gender from the advertisement library based on the user's age and gender information; based on the selected advertisement material: it is used to use NLP to generate personalized advertisement copy, and use GANs to generate personalized advertisement pictures and videos; the advertisement content generation module is used to combine personalized advertisement copy with personalized advertisement pictures and videos to generate personalized advertisement content; the advertisement push and feedback module is used to push personalized advertisement content to users, collect user feedback in real time, and optimize advertisement content and push strategy based on the user's real-time feedback.

[0110] This embodiment also provides a computer device, which is suitable for the case of an advertisement push method based on age and gender recognition, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the advertisement push method based on age and gender recognition proposed in the above embodiment.

[0111] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0112] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for implementing the advertisement push based on age and gender recognition proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0113] In summary, the present invention achieves comprehensive acquisition of user features from facial images, voice signals and gait information through: multimodal data collection and feature extraction, thereby enhancing data diversity and accuracy; through multimodal fusion and comprehensive analysis, multi-source heterogeneous data are effectively integrated to improve the accuracy of age and gender recognition; through the selection of highly matching advertising materials, advertising content that best matches user features is accurately screened out, thereby improving the relevance and effectiveness of advertising; through personalized advertising content generation, NLP and GANs technology are used to generate highly customized copy and visual content, thereby enhancing the attractiveness and interactivity of advertising, and ultimately achieving precision and personalization of advertising push, greatly improving user experience and advertising conversion rate.

[0114] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of an advertisement push method based on age and gender recognition is provided.

[0115] In order to verify the effectiveness of the present invention, this embodiment sets up several information collection points in a large shopping mall. The collection points are equipped with high-resolution cameras, microphone arrays, thermal imagers and infrared sensors. These devices are used to capture the user's facial image, record the user's voice and detect the user's gait information.

[0116] First, the experiment set up information collection points in the shopping mall, using high-resolution cameras, microphone arrays, thermal imagers and infrared sensors to collect users' facial images, sounds and gait information, and extracted facial, voice and gait features through algorithms.

[0117] Secondly, the extracted multimodal features are comprehensively processed through a multimodal fusion network to generate a comprehensive feature vector for accurate identification of the user's age and gender.

[0118] Next, the system selects highly matching advertising materials from the advertising library based on the user’s age and gender information, and uses NLP and GANs technology to generate personalized advertising copy and pictures and videos.

[0119] Finally, personalized advertising content is pushed to users through social media platforms, and user feedback is collected in real time. The advertising push strategy is optimized based on the feedback to improve user experience and advertising effectiveness.

[0120] The details are shown in Table 1 below:

[0121] Table 1 User information and advertising effect statistics

[0122]

[0123] It can be seen from the above experimental data that the multimodal user information collection and analysis system of the present invention has shown remarkable accuracy in age prediction and gender recognition. Taking user 1 as an example, the age prediction value is 28 years old, and the gender bias rate is 0.54 (close to 0.5 means that the gender bias is not obvious), which shows that the system can accurately identify the user's age and gender characteristics. Further observing the advertising material matching degree, the matching degree of user 1 is 0.82, which means that the system can effectively match the advertising material with the user's age and gender information.

[0124] The age- and gender-based advertising push method of the present invention can not only significantly improve the accuracy and effectiveness of advertising, but also significantly improve user experience and satisfaction. In particular, through multimodal feature fusion and personalized content generation, this reflects the significant advantages of the present invention in improving advertising matching and user interaction.

[0125] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. 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, which should all be included in the scope of the claims of the present invention.

Claims

1. An advertisement push method based on age and gender recognition, characterized in that: include, Collect user information and perform feature extraction on the collected user information; Use a multimodal fusion network to perform multimodal fusion on the extracted user features and comprehensively analyze the user's age and gender information; According to the user's age and gender information, identify the advertising creative that best matches the user's age and gender from the advertising library; Based on the selected ad creatives, NLP is used to generate personalized ad copy, and GANs is used to generate personalized ad images and videos. Combine personalized advertising copy with personalized advertising pictures and videos to generate personalized advertising content; Push personalized advertising content to users and collect user feedback in real time. Optimize advertising content and push strategies based on real-time user feedback. The multimodal fusion network is used to perform multimodal fusion on the extracted user features and comprehensively analyze the user's age and gender information. The specific steps are as follows: The extracted facial features, language features and gait features are fused using a multimodal fusion network. The expression is: in, is the comprehensive feature vector, is the facial feature vector, is the speech feature vector, is the gait feature vector; Based on the fused comprehensive feature vector F, the comprehensive perception model MLP is used to comprehensively analyze the user's age and gender information; User age analysis, the expression is: Among them, A is the predicted value of user age, is the i-th feature in the comprehensive feature vector F, is the amplitude coefficient of the ith feature, is the bias term of the jth layer, is the number of hidden layers, is the dimension of the comprehensive feature vector F, is the activation function, is the global bias term, is a nonlinear activation function, is the phase shift of the i-th eigenvector; User gender analysis, the expression is: in, is the frequency coefficient of the ith feature, and G represents the gender bias rate; The specific steps of identifying the advertising material that best matches the user's age and gender from the advertising library according to the user's age and gender information are as follows: Based on the product information and creative content, define an ad library containing multiple creatives, each with a target age range and target gender skew associated with it; Based on the user's age and gender, define a matching function to identify the matching degree between each ad creative and the user's age and gender information. The expression is: in, represents the kth creative, Represents a creative The average of the target age range, Represents a creative The standard deviation of the target age range, Represents a creative The central value of the target gender bias ratio, Represents a creative The standard deviation of the target gender bias rate, is to measure the creative The degree of match with the user's predicted age value A and gender bias rate G.

2. The method for advertising based on age and gender recognition as claimed in claim 1, characterized in that: The user information is collected and features are extracted from the collected user information. The specific steps are as follows: The user's facial image is captured by a high-resolution camera installed in a public place, the user's voice is recorded by an integrated microphone array, and the user's gait information is detected using a thermal imager and an infrared sensor; Use facial detection algorithms to detect facial regions from images and extract facial features, including eyes, nose, and mouth features; Perform noise reduction on the audio signal, use FFT to perform spectrum analysis, and extract speech features, including fundamental frequency and formant frequency; The gait information is filtered and smoothed, and the Kalman filter is used to remove noise and extract gait features, including step length, stride length, and speed.

3. The method for advertising based on age and gender recognition as claimed in claim 2, characterized in that: Based on the comprehensive feature vector F, define a gender discrimination threshold T G , determine the user's gender; When G>T G , then it is judged as female; When G<T G , it is determined to be male.

4. The method for advertising based on age and gender recognition as claimed in claim 3, characterized in that: Based on business needs and historical data, define a matching standard threshold T M , identify highly matching advertising creatives; when ≥T M When the current ad creative is considered to be a high-matching creative; when <T M , the current creative is considered not to be a high-matching creative.

5. The method for advertising based on age and gender recognition as claimed in claim 4, characterized in that: Based on the selected advertising material, NLP is used to generate personalized advertising copy, and GANs is used to generate personalized advertising pictures and videos. The specific steps are as follows: Extract key information from the selected advertising materials, including product name, functional features and target audience, to form a set of keywords K; Generate a copy template according to the type of advertising material and the target audience, fill the relevant information in the keyword set K into the placeholder position in the copy template, and generate preliminary personalized advertising copy; Use NLP to polish the generated preliminary copy, including synonym replacement and sentence structure adjustment, to generate personalized advertising copy; According to the content and style of the selected creatives, collect relevant reference images and video clips to train the GANs model; Use the trained GANs model to generate personalized images and videos that are consistent with the style of the ad creatives.

6. The method for advertising based on age and gender recognition as claimed in claim 5, characterized in that: The personalized advertising copy is combined with personalized advertising pictures and videos to generate personalized advertising content. The specific steps are as follows: By coordinating the consistency between the copy and the image content, the generated personalized advertising copy is matched and combined with the personalized advertising image to enhance the visual appeal; By embedding the generated personalized advertising copy into personalized videos and presenting the copy content in the form of subtitles and narration, the video's appeal can be improved; Once the composition is complete, use Adobe Premiere Pro to export the final personalized advertising content.

7. The method for advertising based on age and gender recognition as claimed in claim 6, characterized in that: The personalized advertising content is pushed to the user, and the user feedback is collected in real time. The advertising content and push strategy are optimized according to the user's real-time feedback. The specific steps are as follows: Based on the user's active time period, the generated personalized advertising content will be pushed to the target users through social media; Through the built-in feedback collection mechanism, users' responses to ads are collected in real time, including click-through rate, dwell time and number of shares, and subjective feedback from users is directly obtained through questionnaires and online evaluations; Optimize push timing, frequency and advertising content based on user feedback to increase the effective reach of advertising; For users with positive feedback, provide more personalized advertising content; for users with negative feedback, optimize push strategies to reduce interference with users' lives.

8. An advertisement push system based on age and gender recognition, based on the advertisement push method based on age and gender recognition according to any one of claims 1 to 7, characterized in that: Including data collection and feature extraction module, multimodal fusion and analysis module, advertising material matching module, advertising content generation module based on selected advertising materials, and advertising push and feedback module; The data collection and feature extraction module is used to collect user information and perform feature extraction on the collected user information; The multimodal fusion and analysis module is used to use the multimodal fusion network to perform multimodal fusion on the extracted user features and comprehensively analyze the user's age and gender information; An advertisement creative matching module, used for identifying, from an advertisement library, advertisement creatives that best match the age and gender of the user based on the age and gender information of the user; Based on the selected ad creatives: used to generate personalized ad copy using NLP, and to generate personalized ad images and videos using GANs; An advertising content generation module, used to combine personalized advertising copy with personalized advertising pictures and videos to generate personalized advertising content; The advertising push and feedback module is used to push personalized advertising content to users, collect user feedback in real time, and optimize advertising content and push strategies based on users' real-time feedback.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the advertisement push method based on age and gender recognition described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the advertisement push method based on age and gender recognition described in any one of claims 1 to 7 are implemented.

Images