Advertisement delivery system and delivery method based on user feedback

Abstract

The present application discloses an advertisement delivery system and delivery method based on user feedback, wherein the system includes: an interactive data collection module for collecting user feedback data; an interactive data processing module for retrieving, combining or generating advertisement content elements, combinations of advertisement content elements or advertisement push data packets from the feedback data; and an advertisement push data playback module for playing advertisement content elements, combinations of advertisement content elements or advertisement push data packets from the interactive data processing module. The application is beneficial in that it provides an advertisement delivery system and delivery method based on user feedback, which can perform secondary processing on the content of delivered advertisements according to user feedback, thereby improving the pertinence of delivery.

Description

[0001] The implementation of the technical solution of this application is legal use. Technical Field

[0002] The present application relates to an advertisement delivery system and delivery method based on user feedback. Background Art

[0003] Existing multimedia advertising delivery systems are mostly based on existing strategies and algorithms. On the one hand, they cannot realize audience feedback and thus cannot achieve intelligent delivery; on the other hand, the delivered advertisements cannot be played in a targeted manner according to the actual audience, resulting in limited actual communication effects.

[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0005] The main purpose of this application is to provide an advertisement delivery system and delivery method based on user feedback, so as to solve the problem in the related art that advertisement delivery cannot match content according to user feedback.

[0006] In order to achieve the above objective, according to one aspect of the present application, a user feedback-based advertising delivery system is provided.

[0007] The user feedback-based advertising delivery system includes:

[0008] Interactive data collection module, used to collect user feedback data;

[0009] An interactive data processing module is used to retrieve, combine or generate advertising content elements, combinations of advertising content elements or advertising push data packets according to feedback data; the advertising content elements include: advertising subject, decorative elements, and personality parameters;

[0010] An advertisement push data playback module, used to play the advertisement content element, a combination of advertisement content elements or an advertisement push data packet of the interactive data processing module;

[0011] The interactive data acquisition module includes: a touch screen;

[0012] The feedback data includes the user's operation on the touch screen;

[0013] The interactive data acquisition module includes: an image acquisition camera;

[0014] The feedback data includes a face image captured by the image acquisition camera;

[0015] The interactive data acquisition module includes: a microphone;

[0016] The feedback data includes voice data collected by the microphone;

[0017] The user feedback-based advertising delivery system further includes:

[0018] A character portrait data recognition module, used to recognize the user's identity data based on the data collected by the character portrait data collection module;

[0019] An advertisement push data matching module is used to match a preset number of advertisement push data packets according to the user's identity and the character portrait data corresponding to the identity;

[0020] Wherein, the interactive data processing module replaces the advertisement content elements of the advertisement push data packet pushed by the advertisement push data matching module according to the character portrait data recognition module and the feedback data;

[0021] The user feedback-based advertising delivery system is also used to respond to the voice data fed back by the user;

[0022] Wherein, the character portrait data recognition module is used to:

[0023] Identify the user;

[0024] Matching a preset number of advertising push data packets according to the user's identity and the corresponding character portrait data;

[0025] Playing the media content of the matched advertisement push data packet;

[0026] The character portrait data includes: biological attribute data of the user;

[0027] The biometric attribute data includes: user's face image data;

[0028] Identify the identity of the user based on the collected facial image data of the user;

[0029] The biological attribute data includes: gender attribute data or age attribute data of the user;

[0030] The character portrait data includes: social attribute data of the user;

[0031] The social attribute data includes: one or more of the user's vehicle information data, clothing information data, and consumption capacity data;

[0032] Matching the advertisement push data packet at least according to the social attribute data;

[0033] Make combination decisions based on user feedback and user personas;

[0034] The advertising delivery method based on character portrait data executed by the advertising delivery system based on user feedback specifically includes the following steps:

[0035] 1: Create an advertisement to be played, and create the conditions for playing the advertisement: the start time of playing, the number of times of playing, the location, and the content of the advertisement; the location includes a garage, and the content of the advertisement includes: pictures, videos, and audio;

[0036] 2: Based on the location conditions of the ads to be played, all the ads to be played the next day are generated into the next day's ad package and uploaded to the cloud server on the same day, taking the garage as the unit;

[0037] 3: Compare the garage advertising package of the day with the garage advertising package of the next day to avoid repeated downloads and occupying memory;

[0038] 4: Download the ads after comparing the garage ad package of the day with the garage ad package of the next day. The principle of downloading is to delete more and add less to ensure the uniqueness of the ads and the accuracy of the ads of the next day;

[0039] 5: Generate the next day's advertising package to be downloaded;

[0040] 6: Ordinary video ads are played in order on the media from 9:00 am to 11:00 pm every day and enter the carousel state;

[0041] 7: Identify the user moving to the face recognition area;

[0042] 8: The face recognition system collects pictures of people;

[0043] 9: The face recognition system identifies the user ID;

[0044] 10. The face recognition system identifies the user tag, which specifically includes the following structural aspects:

[0045] 10.1 Static Tags:

[0046] Gender: male, female;

[0047] Age stage: infants, adolescents, young adults, middle-aged and elderly;

[0048] 10.2 Dynamic tag data includes: clothing color;

[0049] 10.3 Vehicle labels include: model, price range, brand, color;

[0050] The above tags are confirmed according to the real-time collection, or judged according to the accumulated data;

[0051] 11: Recognize each person’s face once and recalculate the person portrait of this user ID;

[0052] User portrait: gender, age group, annual income, color preference; the specific calculation method is as follows:

[0053] Gender: The most recognized label;

[0054] Age group: the label with the most recognitions;

[0055] Color preference: car color label and clothing color label, among multiple colors, take the label that appears most often;

[0056] The user portrait of the garage as a collection is used as the garage portrait to support the advertising placement location. The advertising matching degree is compared with the user portrait of the advertising demand and the garage portrait. The garage with the highest matching degree and its media points are selected for advertising placement to increase the target user range and accuracy;

[0057] 12: Retrieve the advertisement package of the user's garage on that day, and match the user's portrait with the portrait requirements of the advertisements in the advertisement package one by one;

[0058] 13: The three advertisements with the highest matching degree form a user advertisement package;

[0059] 14: User voice triggers interaction;

[0060] 15: The advertising media enters the interactive state;

[0061] 16: Regular video ad rotation is paused;

[0062] 17: Determine whether the advertisement with the highest matching degree has been downloaded and cached on the media triggered by the user;

[0063] 18: If not downloaded, download the advertisement with the highest matching degree from the user advertisement package of the edge server;

[0064] 19: The content of the user's advertising package is rotated until the user leaves;

[0065] When performing face recognition, once the user's identity is identified, the last targeted advertisement played to the user is called, and the user's face data continues to be collected. If the user's expression shows boredom or his eyes are shifted, the character portrait calculation is performed in the background, and then an advertisement is selected for delivery according to the calculation result of this character portrait, and the user's face data continues to be collected. If the user's expression shows like or his eyes are focused, the character portrait data of this time is used as the user portrait data for the next time the user is identified. If the user still does not express interest in the current advertisement through face image collection, the character portrait data with the largest number of occurrences is selected from all the character portrait data for the user calculated previously to match the advertisement data;

[0066] A weight value is set for each character portrait data, and the user's facial expression and gaze time when the advertisement matching the character portrait data is played are used as the standard for calculating the weight value;

[0067] The weight value Y=X+Z, where X is the average expression parameter during the gaze period, Z is the gaze time, and the average expression parameter X=0.4A+0.6B, where A is the arc angle value of the user's mouth corner, and B is the angle value of the user's eye corner; when matching advertising data, the character portrait data obtained by sorting according to the weight value will be matched with the advertisements in the current advertising data first;

[0068] Modify each individual item of the user's overall character portrait data according to the weight value, add the weight values ​​of the historical data of color preference and clothing items, and obtain absolute advantage data;

[0069] When the historical data of a certain single item exceeds the preset ratio of the sum of the overall weight values, in the next delivery, an attempt will be made to use the second single item value as the current character portrait data for delivery;

[0070] A convolutional neural network is set up to train and construct the model through the collected person images and the set person portrait standard model; after the person images are collected, the image data is input into the convolutional neural network, and the convolutional neural network performs intelligent judgment instead of the natural person; multiple convolutional neural networks are used for training to identify the individual data of different person portrait data;

[0071] The advertising data is stored in a labeled manner, and the advertising data is stored in a snowflake model or a star model using individual data in the user persona as data labels, and the individual data in the user persona of the advertising data is stored as the primary key;

[0072] When multiple people appear in the captured image, the similarity between the two users is determined. If the similarity is greater than the preset value, the user who came first is used as the main ad matcher. If the similarity is lower, the user who is looking at the screen is determined, and the ad matcher is mainly based on the user who is looking at the screen. If two people are looking at each other, the expression data of the two users is determined, and the user with the higher expression value is used as the matching standard. If it is impossible to determine, the user is selected for matching based on the order of the users.

[0073] When matching ads, weather and time data are combined as one of the input items for ad matching.

[0074] The benefit of the application lies in: providing an advertisement delivery system and delivery method based on user feedback, which can perform secondary processing on the content of delivered advertisements according to user feedback, thereby improving the targeting of delivery. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The drawings constituting a part of this application are used to provide a further understanding of this application, so that other features, purposes and advantages of this application become more obvious. The schematic embodiment drawings and their descriptions of this application are used to explain this application and do not constitute an improper limitation on this application. In the drawings:

[0076] Figure 1 This is a block diagram of the modules of an advertising delivery system based on character portrait data according to an embodiment of the present application;

[0077] Figure 2 It is a schematic diagram of the steps of an advertising delivery method based on character portrait data according to an embodiment of the present application;

[0078] Figure 3 is a schematic diagram of a precise advertising delivery process based on character portrait data according to an embodiment of the present application;

[0079] Figure 4 It is a schematic diagram of the module composition of an advertisement delivery system based on user feedback according to an embodiment of the present application;

[0080] Figure 5 It is a schematic diagram of the module composition of an advertisement delivery method based on user feedback according to an embodiment of the present application; DETAILED DESCRIPTION

[0081] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0082] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0083] Reference Figure 1 As shown, the advertising delivery system of the present application includes: a character portrait data collection module, a character portrait data recognition module, an advertising push data matching module and an advertising push data playback module.

[0084] Among them, the character portrait data collection module is used to collect the user's character portrait data; the character portrait data recognition module is used to identify the user's identity based on the data collected by the character portrait data collection module; the advertising push data matching module is used to match a preset number of advertising push data packets based on the user's identity and the character portrait data corresponding to the identity; the advertising push data playback module is used to play the media content of the advertising push data packet matched by the advertising push data matching module.

[0085] Among them, the character portrait data collection module and the advertisement push data playback module can be set in an advertisement delivery interactive terminal device, while the character portrait data recognition module and the advertisement push data matching module can be composed of a server system.

[0086] As a specific solution, the character portrait data acquisition module includes: an image acquisition camera.

[0087] Character portrait data includes two categories: user's biological attribute data and social attribute data.

[0088] Specifically, the biometric attribute data includes: user's facial image data, user's gender attribute data and age attribute data.

[0089] As a specific plan, age attribute data is divided into: infants / teenagers / youth / middle-aged / elderly.

[0090] The image acquisition camera is at least used to collect the user's face image data; the character portrait data recognition module recognizes the user's identity based on the collected user's face image data. Of course, other attributes of the user can also be recognized, such as fingerprints and irises.

[0091] The user's social attribute data includes: one or more of the user's vehicle information data, clothing information data, and consumption capacity data. The advertisement push data matching module matches the advertisement push data packet at least according to the social attribute data.

[0092] As a specific solution, the vehicle information data includes: vehicle model level data, price range data, vehicle brand data and vehicle color data.

[0093] As a specific solution, the clothing information data can adopt the following solution:

[0094] Clothing color.

[0095] As an optional solution, the image acquisition camera can collect license plate information and perform data association through the license plate information.

[0096] like Figure 2 As shown, as a specific solution, the advertising delivery method based on character portrait data according to the present application includes the following steps:

[0097] S1 collects user's character portrait data.

[0098] S2 identifies the user.

[0099] S3 matches a preset number of advertising push data packets according to the user's identity and its corresponding character portrait data.

[0100] S4 plays the media content of the matched advertisement push data packet.

[0101] Specifically, the character portrait data includes: the user's biological attribute data; the biological attribute data includes: the user's facial image data; the user's identity is identified based on the collected user's facial image data. The user's biological attribute data includes: the user's gender attribute data or age attribute data. The character portrait data includes: the user's social attribute data; the social attribute data includes: one or more of the user's vehicle information data, clothing information data and consumption capacity data. At least the advertising push data packet is matched based on the social attribute data.

[0102] like Figure 3 As shown in the figure, as a specific system configuration and delivery method, it includes the following processes:

[0103] 1: The EBC system creates the advertisement to be played and creates the conditions for playing the advertisement: the start time of the advertisement, the number of times the advertisement is played, the location (garage), and the content of the advertisement (picture, video, audio);

[0104] 2: Based on the location conditions of the ads to be played, all the ads to be played the next day are generated into the next day's ad package and uploaded to the cloud server on the same day, taking the garage as the unit;

[0105] 3: Edge server: compares the garage advertising package of the day with the garage advertising package of the next day to avoid repeated downloads and occupying memory;

[0106] 4: Edge server: Download the advertisements after comparing the garage advertisement package of the current day with the garage advertisement package of the next day. The principle of downloading is to delete more and add less to ensure the uniqueness of the advertisements and the accuracy of the advertisements of the next day.

[0107] 5: Edge server: forms the next day's advertising package to be downloaded;

[0108] 6: Advertisement playback media: Generally, ordinary video ads are played in order on the media from 9:00 am to 11:00 pm every day, and enter the rotation state;

[0109] 7: The user moves to the face recognition area;

[0110] 8: The face recognition system collects pictures of people;

[0111] 9: The face recognition system identifies the user ID;

[0112] 10. The face recognition system identifies the user tag, which specifically includes the following structural aspects:

[0113] 10.1 Static Tags:

[0114] Gender: male, female;

[0115] Age stage: infants, adolescents, young adults, middle-aged and elderly;

[0116] 10.2 Dynamic tag data includes: clothing color;

[0117] 10.3 Vehicle labels include: model, price range, brand, color;

[0118] The above tags can be expanded with the business. These tags can be confirmed according to the real-time collection or judged according to the accumulated data.

[0119] 11: Recognize each person’s face once and recalculate the person portrait of this user ID;

[0120] User portrait: gender, age group, annual income, color preference. The specific calculation method is as follows:

[0121] Gender: The most recognized label;

[0122] Age group: the label with the most recognitions;

[0123] Annual income: sorted by vehicle price range labels as less than 100,000, 100,000 to less than 350,000, 350,000 to less than 500,000, and more than 500,000;

[0124] Color preference: one car color label (select the label with the most occurrences among multiple colors), one clothing color label (select the label with the most occurrences among multiple colors).

[0125] In addition, the user portrait based on garages is used as a garage portrait to provide support for advertising placement locations. The advertising matching degree is compared between the user portrait of advertising demand and the garage portrait, and the garages and media locations with the highest matching degree are screened out for advertising placement, thereby increasing the target user range and accuracy.

[0126] 12: Retrieve the advertisement package of the user's garage on that day, and match the user's portrait with the portrait requirements of the advertisements in the advertisement package one by one;

[0127] 13: The three advertisements with the highest matching degree form a user advertisement package;

[0128] 14: User voice triggers interaction;

[0129] 15: The advertising media enters the interactive state;

[0130] 16: Regular video ad rotation is paused;

[0131] 17: Determine whether the advertisement with the highest matching degree has been downloaded and cached on the media triggered by the user;

[0132] 18: If not downloaded, download the advertisement with the highest matching degree from the user advertisement package of the edge server;

[0133] 19: The content of the user's advertising package will be rotated until the user leaves.

[0134] Each time face recognition is performed, the character portrait is recalculated, which is likely to increase the uncertainty of the character portrait and easily cause computational redundancy or low efficiency. It may happen that the calculation is not completed yet and the user has left the advertising location, resulting in the failure of precise delivery.

[0135] As an extended solution, when performing face recognition, once the user's identity is identified, the last precisely delivered advertisement played to the user is called, and the user's facial data continues to be collected. If the user's expression shows boredom or his gaze is deflected, the character portrait calculation is performed in the background, and then according to the calculation result of this character portrait, an advertisement is selected for delivery, and the user's facial data continues to be collected. If the user's expression shows like or his gaze is focused, the character portrait data is used as the user portrait data for the next time the user is identified. If the user still does not express interest in the current advertisement through facial image collection, the character portrait data with the largest number of appearances among all the previously calculated character portrait data for the user is selected to match the advertisement data.

[0136] As a preferred solution, a weight value is set for each character portrait data, and the user's facial expression and gaze time when the advertisement matched with the character portrait data is played are used as the standard for calculating the weight value.

[0137] Specifically, the weight value Y=X+Z, where X is the average expression parameter during the gaze period, Z is the gaze time, and the average expression parameter X=0.4A+0.6B, where A is the arc angle value of the user's mouth corner, and B is the angle value of the user's eye corner. When matching advertising data, the character portrait data obtained by installing the weight value sorting will preferentially match the advertisements in the current advertising data.

[0138] As another specific scheme, the individual data of the user's overall character portrait data are corrected according to the weight values. For example, in some variable data items, such as color preference and clothing items, the weight values ​​of the historical data that have appeared are added to obtain absolute advantage data. For example, in the historical data, the color preferences have appeared in yellow and red, but the weight of yellow in the historical data is greater than the weight of red, then the user's color preference is considered to be yellow in the overall character portrait data.

[0139] As a further solution, when the historical data of a certain single data exceeds the preset ratio of the sum of the overall weight values, the next time the single data value is placed, an attempt will be made to place it with the second single data value as the current character portrait data. This can reduce the data bias caused by machine calculation.

[0140] As an extension solution, a convolutional neural network is set up in the system, and the model is trained and constructed through the collected person images and the set person portrait standard model; after the person image is collected, the image data is input into the convolutional neural network, and the convolutional neural network replaces the natural person to make intelligent judgments. Multiple convolutional neural networks can be used for training to identify individual data of different person portrait data.

[0141] As a further expansion plan, this application can adopt the method of building a data middle platform and a business middle platform to communicate between the underlying data and the front end, avoiding duplication of construction for different advertising products (such as garage advertising and elevator advertising).

[0142] The advertisement data is stored in a tagged manner, and the advertisement data is stored in a snowflake model or a star model with the single data in the user's persona as the data tag for reuse. As a further preferred solution, the single data in the user's persona of the advertisement data is stored as the primary key. In this way, many-to-many data matching can be achieved during matching.

[0143] As a further expansion scheme, when the present application collects multiple people in the image, it determines the similarity of two users. If the similarity is greater than a preset value, the user who came first is used as the main advertising matcher. If the similarity is lower, it determines which user is looking at the screen, and the user who is looking at the screen is used as the main advertising matcher. If two people are looking at each other, it determines the expression data of the two users, and the user with the higher expression value is used as the matching standard. If it is impossible to determine, the user is selected for matching based on the order of the users.

[0144] As a further extension, the present application combines weather and time data as one of the input items for advertisement matching when matching advertisements.

[0145] As another aspect of this application, reference is made to Figures 4 to 5As shown, the advertisement delivery system based on user feedback includes: an interactive data collection module, an interactive data processing module and an advertisement push data playback module.

[0146] Among them, the interactive data collection module is used to collect user feedback data; the interactive data processing module is used to retrieve, combine or generate advertising content elements, combinations of advertising content elements or advertising push data packets for feedback data; the advertising push data playback module is used to play the advertising content elements, combinations of advertising content elements or advertising push data packets of the interactive data processing module.

[0147] Specifically, the interactive data acquisition module includes: a touch screen, an image acquisition camera and a microphone. The feedback data includes the user's operation on the touch screen, the face image acquired by the image acquisition camera or the voice data acquired by the microphone.

[0148] As a preferred solution, the touch screen can provide menus or buttons for the user to operate, thereby realizing user interaction. In addition, the interactive data processing module can convert voice data into field data, thereby facilitating the retrieval of advertising content elements.

[0149] As a preferred solution, the interactive data collection module and the character portrait data collection module can be the same module.

[0150] There are three types of advertising content elements: advertising body, decorative elements, and personality parameters.

[0151] The advertisement body includes: picture files, video file segments, audio file segments or a combination thereof.

[0152] Decorative elements include: image front frame, image background, and floating icon.

[0153] Personality parameters include data such as color, position, brightness, etc. of decorative elements.

[0154] A combination of advertising content elements may be output according to user feedback, or an already packaged advertising file may be modified and replaced.

[0155] like Figure 5 As shown, the advertising delivery method based on user feedback according to the present application includes the following steps: collecting user feedback data; retrieving, combining or generating advertising content elements, a combination of advertising content elements or an advertising push data packet based on the feedback data; playing the advertising content elements, a combination of advertising content elements or an advertising push data packet of the interactive data processing module.

[0156] Specifically, the feedback data includes: user operation data, face data and user voice data.

[0157] As a more specific scheme, the user feedback-based advertising delivery method of the present application includes the following steps: collecting user character portrait data; identifying the user's identity; matching a preset number of advertising push data packets according to the user's identity and its corresponding character portrait data; replacing the advertising content elements in the advertising push data packets according to the character portrait data identification module and feedback data.

[0158] As an extension of the previous solution, combined decisions can be made based on user feedback and user personas.

[0159] like Figure 5 As shown, the advertisement delivery method based on user feedback of the present application includes the following steps:

[0160] After the user triggers an interaction, the system enters the interactive state.

[0161] The advertisement delivery interactive terminal device sends a voice message to the user, asking the user questions related to the advertisement.

[0162] After the user responds, the system identifies the user's response, retrieves relevant advertising data, and intelligently responds to the user's question.

[0163] As a more specific solution, the system of the present application has a built-in AI system for providing question-and-answer feedback on user questions.

[0164] The user can trigger the interaction mode with the advertising delivery interactive terminal device of the present application by the voice pronunciation of a specific field. For example, the user can trigger the interaction mode of the advertising delivery interactive terminal device of the present application by calling a specific nickname.

[0165] As an optional solution, the advertising playback of the advertising delivery interactive terminal device of the present application can be independent of the interaction with the AI ​​interactive system, that is, the advertising can be played according to the character portrait when the AI ​​system interacts. As a preferred solution, the advertising playback of the advertising delivery interactive terminal device of the present application can be associated with the AI ​​system, that is, the advertising played by the advertising delivery interactive terminal device is based on the question and answer results of the AI ​​interactive system.

[0166] Specifically, after triggering the interactive mode, the user asks "What delicious food is nearby?" The AI ​​system obtains field data based on the voice, and then searches for the corresponding advertising content elements or combinations in the current advertising package or database based on the keywords "nearby" and "delicious food" for display. At the same time, the AI ​​system responds based on the user's voice data, such as "The nearby delicious food is recommended to you, please click to learn more."

[0167] If the user clicks on the corresponding area, the advertising delivery interactive terminal device will continue to display further advertising data, such as menu and price, based on the user's click feedback.

[0168] As a preferred solution, when providing feedback based on the user's voice, the system of the present application still analyzes the user's preferences through the character portrait data and pushes the preferred data based on the user's preferences.

[0169] For example, if the same user asks "What delicious food is nearby?", the interactive advertising terminal will collect and analyze images.

[0170] As a further expansion plan, even if the system pushes hot pot advertisements at the same time, it will adjust the content elements of the advertisements based on the user's character portrait tag data.

[0171] As a further preferred solution, when the question raised by the user is not about advertising or is a question that cannot be answered by the current advertisement, the AI ​​system first answers the question and then obtains new corresponding advertising data from the cloud server or other data sources.

[0172] For example, when a user asks "What's the weather like today?", the AI ​​system obtains weather data from the server or accesses Internet data to respond to the user's inquiry.

[0173] At this time, the advertising interactive terminal has three modes: 1. Maintain carousel, 2. Push advertisements based on user portrait data, and 3. Make intelligent associations based on the results generated by user inquiries. For example, when a user asks "What's the weather like today?", the AI ​​system answers "It's cloudy with light rain today", then advertising data about rain gear will be pushed.

[0174] In addition, the advertising delivery interactive terminal of the present application can also realize general AI voice question and answer functions, such as assisting users in finding vehicles based on voice questions and answers.

[0175] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An advertisement delivery system based on user feedback, characterized in that: The user feedback-based advertising delivery system includes: Interactive data collection module, used to collect user feedback data; An interactive data processing module is used to retrieve, combine or generate advertising content elements, combinations of advertising content elements or advertising push data packets according to feedback data; the advertising content elements include: advertising subject, decorative elements, and personality parameters; An advertisement push data playback module, used to play the advertisement content element, a combination of advertisement content elements or an advertisement push data packet of the interactive data processing module; The interactive data acquisition module includes: a touch screen; The feedback data includes the user's operation on the touch screen; The interactive data acquisition module includes: an image acquisition camera; The feedback data includes a face image captured by the image acquisition camera; The interactive data acquisition module includes: a microphone; The feedback data includes voice data collected by the microphone; The user feedback-based advertising delivery system further includes: A character portrait data recognition module is used to recognize the user's identity data based on the data collected by the character portrait data collection module; An advertisement push data matching module is used to match a preset number of advertisement push data packets according to the user's identity and the character portrait data corresponding to the identity; Wherein, the interactive data processing module replaces the advertisement content elements of the advertisement push data packet pushed by the advertisement push data matching module according to the character portrait data recognition module and the feedback data; The user feedback-based advertising delivery system is also used to respond to the voice data fed back by the user; Wherein, the character portrait data recognition module is used to: Identify the user; Matching a preset number of advertising push data packets according to the user's identity and the corresponding character portrait data; Playing the media content of the matched advertisement push data packet; The character portrait data includes: biological attribute data of the user; The biometric attribute data includes: user's face image data; Identify the identity of the user based on the collected facial image data of the user; The biological attribute data includes: gender attribute data or age attribute data of the user; The character portrait data includes: social attribute data of the user; The social attribute data includes: one or more of the user's vehicle information data, clothing information data, and consumption capacity data; Matching the advertisement push data packet at least according to the social attribute data; Make combination decisions based on user feedback and user personas; The advertising delivery method based on character portrait data executed by the advertising delivery system based on user feedback specifically includes the following steps: 1: Create an advertisement to be played, and create the conditions for playing the advertisement: the start time of playing, the number of times of playing, the location, and the content of the advertisement; the location includes a garage, and the content of the advertisement includes: pictures, videos, and audio; 2: Based on the location conditions of the ads to be played, all the ads to be played the next day are generated into the next day's ad package and uploaded to the cloud server on the same day, taking the garage as the unit; 3: Compare the garage advertising package of the day with the garage advertising package of the next day to avoid repeated downloads and occupying memory; 4: Download the ads after comparing the garage ad package of the day with the garage ad package of the next day. The principle of downloading is to delete more and add less to ensure the uniqueness of the ads and the accuracy of the ads of the next day; 5: Generate the next day's advertising package to be downloaded; 6: Ordinary video ads are played in order on the media from 9:00 am to 11:00 pm every day and enter the carousel state; 7: Identify the user moving to the face recognition area; 8: The face recognition system collects pictures of people; 9: The face recognition system identifies the user ID; 10. The face recognition system identifies the user tag, which specifically includes the following structural aspects: 10.1 Static Tags: Gender: male, female; Age stage: infants, adolescents, young adults, middle-aged and elderly; 10.2 Dynamic tag data includes: clothing color; 10.3 Vehicle labels include: model, price range, brand, color; The above tags are confirmed according to the real-time collection, or judged according to the accumulated data; 11: Recognize each person’s face once and recalculate the person portrait of this user ID; User portrait: gender, age group, annual income, color preference; the specific calculation method is as follows: Gender: The most recognized label; Age group: the label with the most recognitions; Color preference: car color label and clothing color label, among multiple colors, take the label that appears most often; The user portrait of the garage as a collection is used as the garage portrait to support the advertising placement location. The advertising matching degree is compared between the user portrait of the advertising demand and the garage portrait. The garage with the highest matching degree and its media points are selected for advertising placement to increase the target user range and accuracy; 12: Retrieve the advertisement package of the user's garage on that day, and match the user's portrait with the portrait requirements of the advertisements in the advertisement package one by one; 13: The three advertisements with the highest matching degree form a user advertisement package; 14: User voice triggers interaction; 15: The advertising media enters the interactive state; 16: Regular video ad rotation is paused; 17: Determine whether the advertisement with the highest matching degree has been downloaded and cached on the media triggered by the user; 18: If not downloaded, download the advertisement with the highest matching degree from the user advertisement package of the edge server; 19: The content of the user's advertising package is rotated until the user leaves; When performing face recognition, once the user's identity is identified, the last targeted advertisement played to the user is called, and the user's face data continues to be collected. If the user's expression shows boredom or his eyes are shifted, the character portrait calculation is performed in the background, and then an advertisement is selected for delivery according to the calculation result of this character portrait, and the user's face data continues to be collected. If the user's expression shows like or his eyes are focused, the character portrait data of this time is used as the user portrait data for the next time the user is identified. If the user still does not express interest in the current advertisement through face image collection, the character portrait data with the largest number of occurrences is selected from all the character portrait data for the user calculated previously to match the advertisement data; A weight value is set for each character portrait data, and the user's facial expression and gaze time when the advertisement matching the character portrait data is played are used as the standard for calculating the weight value; The weight value Y=X+Z, where X is the average expression parameter during the gaze period, Z is the gaze time, and the average expression parameter X=0.4A+0.6B, where A is the arc angle value of the user's mouth corner, and B is the angle value of the user's eye corner; when matching advertising data, the character portrait data obtained by sorting according to the weight value will be matched with the advertisements in the current advertising data first; Modify each individual item of the user's overall character portrait data according to the weight value, add the weight values ​​of the historical data of color preference and clothing items, and obtain absolute advantage data; When the historical data of a certain single item exceeds the preset ratio of the sum of the overall weight values, in the next delivery, an attempt will be made to use the second single item value as the current character portrait data for delivery; A convolutional neural network is set up to train and construct the model through the collected person images and the set person portrait standard model; after the person images are collected, the image data is input into the convolutional neural network, and the convolutional neural network performs intelligent judgment instead of the natural person; multiple convolutional neural networks are used for training to identify the individual data of different person portrait data; The advertising data is stored in a labeled manner, and the advertising data is stored in a snowflake model or a star model using individual data in the user persona as data labels, and the individual data in the user persona of the advertising data is stored as the primary key; When multiple people appear in the captured image, the similarity between the two users is determined. If the similarity is greater than the preset value, the user who came first is used as the main ad matcher. If the similarity is lower, the user who is looking at the screen is determined, and the ad matcher is mainly based on the user who is looking at the screen. If two people are looking at each other, the expression data of the two users is determined, and the user with the higher expression value is used as the matching standard. If it is impossible to determine, the user is selected for matching based on the order of the users. When matching ads, weather and time data are combined as one of the input items for ad matching.

Images