An advertisement replacement method of an advertisement delivery device

Abstract

This invention belongs to the field of big data processing, specifically relating to a method for updating advertisements on an advertising delivery device. It includes the following steps: Step 1: Obtain the feature data of the currently playing advertisement. Step 2: Obtain feedback data from each user regarding the advertisement playback. Step 3: Based on the keyword dataset, audio stream data, video stream data, and SW value obtained in the first two steps, calculate the user's approval rating value E for the current advertisement. n Step 4: Update the ad playback sequence table according to the following rules: Sort the average approval rating results of each ad in all played ad sequences within an update cycle; delete the ads ranked lower in the rating ranking table, and add the ads that need to be added to the ad playback sequence table. This invention solves the problem that existing ad delivery devices cannot obtain user feedback on the delivered ads, and cannot adjust the ad playback order based on feedback from the target group.

Description

[0001] This application is a divisional application of application number CN 202110686166.3, application date 2021 / 06 / 21, and invention title: A VOC vehicle owner cloud big data platform. Technical Field

[0002] This invention belongs to the field of big data processing, and specifically relates to a method for updating advertisements on an advertising delivery device. Background Technology

[0003] Advertising is a means of commercial marketing. When placing ads, clients are usually very concerned about their effectiveness, and maximizing ad performance is a shared goal of both advertisers and clients. Online marketing ads typically use metrics such as click-through rate (CTR) and page dwell time to evaluate ad performance; these metrics are very direct and effective. Offline, for product marketing ads, the purchase rate is the best indicator of ad performance. However, for most advertising scenarios, users do not usually make a purchase immediately after receiving the ad, so this evaluation method has a significant lag. Furthermore, when using product sales as an indicator of ad performance, because advertisers place ads in a wide variety of scenarios, it is often impossible to pinpoint which specific scenario's ad performance was effective when a sale occurs.

[0004] Traditional advertising devices in locations such as elevators and parking lots typically play ads sequentially according to pre-stored content. Alternatively, they may receive ads from a remote cloud server and play them under its control. Regardless of the method, advertising plays according to pre-set rules by the advertiser, making it impossible to differentiate or adjust the targeting based on the intended audience. Therefore, optimal advertising results cannot be achieved. Summary of the Invention

[0005] To address the problem that existing advertising delivery devices cannot obtain user feedback on delivered ads and therefore cannot adjust the ad playback order based on feedback from the target group, this invention provides an ad re-running method for advertising delivery devices.

[0006] This invention is achieved using the following technical solution:

[0007] A method for updating advertisements on an advertising delivery device includes the following steps:

[0008] Step 1: Obtain the feature data of the currently playing advertisement, including: the playback duration T of the currently playing advertisement, and the keyword dataset associated with the advertisement.

[0009] Step 2: Obtain feedback data from each user regarding ad playback, including:

[0010] (1) Acquire the voice stream data generated by the current user group during the advertisement playback, monitor the video stream data of the current user group, and receive instructions from one or more users in the current user group to switch the currently playing advertisement.

[0011] (2) Determine whether an instruction to switch the currently playing advertisement has been received. If yes, assign 1 to the feature quantity SW that reflects the instruction; otherwise, assign 0 to SW.

[0012] Step 3: Based on the keyword dataset and the audio stream data, video stream data, and SW value in the feedback data, calculate the user's approval rating value E for the current advertisement. n .

[0013] Step 4: Update the ad playback sequence table according to the following rules:

[0014] (1) Obtain the average approval rating of each advertisement in the entire playable advertisement sequence list within an update period.

[0015] (2) Based on the E of each advertisement i The scores are sorted from largest to smallest for all played ads within the update period to obtain a ranking table of played ads.

[0016] (3) Obtain the ads that need to be added and their quantity, delete the corresponding number of played ads that rank lower in the rating ranking table from the ad playback sequence table, and add the ads that need to be added to the ad playback sequence table to complete the update process of the ad sequence table.

[0017] As a further improvement of the present invention, in step one, the feature data in the keyword dataset consists of a plurality of pre-defined keywords related to the content of the currently playing advertisement; the feature data in the keyword dataset associated with each advertisement includes at least:

[0018] (1) Keywords that reflect the advertised products.

[0019] (2) Keywords that reflect the target customer group of the advertisement.

[0020] (3) Keywords that reflect the spokesperson or character in the advertisement.

[0021] (4) High-frequency or special keywords in advertising copy.

[0022] (5) Classification of advertising duration: very short, short, medium, long, and extra long.

[0023] (6) Advertising style classification.

[0024] As a further improvement of the present invention, in step two, the sources of the instruction to switch the currently playing advertisement issued by one or more users in the current user group include three:

[0025] (1) A switching instruction issued by a human-computer interaction module to switch the currently playing advertisement.

[0026] (2) Voice interaction commands issued by the user to switch the currently playing advertisement.

[0027] (3) A gesture interaction command issued by a user to switch the currently playing advertisement.

[0028] As a further improvement of the present invention, in step three, each user's approval rating value E for the current advertisement is... n The calculation method is as follows;

[0029] (1) Perform speech recognition on the speech stream data, extract the keywords that match the feature data in the keyword dataset associated with the currently playing advertisement, and count their number N1.

[0030] (2) Perform video action recognition on the video stream data; extract the gestures that represent the user’s response to the currently playing advertisement, and count their number N2.

[0031] (3) Perform video action recognition on the video stream data; extract the feature actions that reflect changes in the current user's eye focus position, and calculate the current user's attention duration t for the currently playing advertisement based on the feature actions. n Where n represents the current user's user ID.

[0032] (4) Perform frame-by-frame sampling on the frame-by-frame images of the video stream data according to the sampling frequency; perform image recognition on the frame-by-frame sampled images; extract the facial expressions of each user and classify the facial expressions as like, ignore or dislike; count the number of the three types of expression classification results for each user and calculate the proportion of the number of the three types of expression classification results for each user in the total sample size of the user.

[0033] (5) Get the value of SW.

[0034] (6) Calculate the approval rating E of each user for the current advertisement using the following formula. n :

[0035]

[0036] In the above formula, n represents the user ID of the current user, E n E represents the rating of user ID n for the currently playing advertisement.n ≥0, and E n The higher the value, the higher the user's approval of the currently playing multimedia. The expression represents the user's attention level to the currently playing advertisement (number n); k1 represents the influence factor of voice feedback on the overall acceptance rating; k2 represents the influence factor of posture feedback on the overall acceptance rating; k3 represents the influence factor of facial expression feedback on the overall acceptance rating; k4 represents the influence factor of attention level on the overall acceptance rating; m1 represents the score of a single keyword in the voice feedback; m2 represents the score of a single posture in the posture feedback; m3 represents the score of attention level; a represents the score of the "like" expression; p 1,n The percentage of images sampled in every frame for user number n who are categorized as having favorite expressions; b represents the score for ignoring expressions, p 2,n The percentage of images sampled every other frame for user number n who are classified as ignoring facial expressions; c represents the score for the disgust expression, p 3,n The percentage of images sampled every frame for user number n who are classified as having expressions of disgust.

[0037] As a further improvement to the present invention, in step four, the average approval rating result E of each advertisement is... i The calculation formula is as follows:

[0038]

[0039] In the above formula, i represents the number of each advertisement in the advertisement playback sequence list.

[0040] As a further improvement of the present invention, the voice stream data is acquired through a voice acquisition module; when each advertisement is played, the voice acquisition module acquires the voice information generated by the user group watching the advertisement; the voice acquisition module includes multiple microphones, which are installed on the advertisement delivery device.

[0041] As a further improvement of the present invention, video stream data is acquired through a video monitoring module; the video monitoring module monitors the user group watching the advertisement from multiple angles during each advertisement; the video monitoring module includes multiple monitoring cameras; the monitoring cameras are installed on the advertisement delivery device.

[0042] The technical solution provided by this invention has the following beneficial effects:

[0043] The method provided by this invention can acquire user feedback information on different advertisements within a certain period and calculate a quantitative value of the popularity of each advertisement based on the feedback information. It also automatically adjusts the type of advertisement placement based on user feedback. This not only helps advertisers improve the effectiveness of their advertising, enabling product advertisements to be more targeted to the target audience, but also helps increase user goodwill towards advertisers and enhance the user experience.

[0044] This invention, at the start of an advertising update cycle, accumulates scores based on all user feedback received during each advertisement's playback within that cycle, calculating the average user approval rating for each advertisement. A higher average rating indicates a wider reach of the advertisement to its target audience; in other words, users frequently appearing in the current advertising area belong to the target group, and therefore the advertisement's frequency should be increased. Conversely, a lower average rating indicates a lower frequency of advertising. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating a method for creating an advertising analysis database in a VOC (Vehicle Owner Big Data) platform according to Embodiment 1 of the present invention.

[0046] Figure 2 This is a classification diagram of the types of feature data contained in the identity tags in the advertising analysis database of Embodiment 1 of the present invention;

[0047] Figure 3 This is a type distinction diagram of the feature data contained in the user profile dataset in Embodiment 1 of the present invention;

[0048] Figure 4 This is a schematic diagram of the modules of an advertising analysis database creation system according to Embodiment 2 of the present invention;

[0049] Figure 5 This is a flowchart of a user's acceptance evaluation method for advertisements based on feature recognition, as described in Embodiment 3 of the present invention.

[0050] Figure 6 This is a schematic diagram of the module connection of an intelligent media management system based on a VOC vehicle owner big data platform provided in Embodiment 4 of the present invention;

[0051] Figure 7 This is a flowchart of a user profile-based targeted advertising method in Embodiment 5 of the present invention;

[0052] Figure 8 This is a flowchart of a method for timely analysis of user needs in a business district scenario according to Embodiment 6 of the present invention;

[0053] Figure 9This is a flowchart of a method for matching user needs and advertising content in Embodiment 7 of the present invention;

[0054] Figure 10 This is a schematic diagram of a garage giant screen MAX smart terminal with intelligent voice interaction function provided in Embodiment 8 of the present invention;

[0055] Figure 11 This is a diagram showing the types of switching commands used by the human-computer interaction module in a garage giant screen MAX smart terminal with intelligent voice interaction function, as described in Embodiment 8 of the present invention.

[0056] Figure 12 This is a flowchart of an advertising reprinting method for an advertising delivery device provided in Embodiment 9 of the present invention;

[0057] Figure 13 This is a flowchart of a method for evaluating the effectiveness of media advertising in a smart business district, as described in Embodiment 10 of the present invention. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0059] Example 1

[0060] This embodiment provides a VOC (Vehicle Owner) big data platform, which stores an advertising analysis database containing data from multiple historical users. The data in this advertising analysis database enables precise user profiles based on their interests and preferences, thereby allowing for targeted advertising marketing to users.

[0061] The data in the advertising analytics database is primarily derived from user identification features and user feedback on video ads observed in scenarios such as elevators, parking garages, and shopping malls. The data in the advertising analytics database mainly includes the following aspects:

[0062] (1) User's facial features; This feature is mainly used to distinguish the identities of different users and serves as a unique identifier for users. At the same time, the advertising analysis database also assigns a unique user number to each user based on their unique identifier.

[0063] (2) User identity characteristics; This part of the data is relatively rich, covering all the characteristics that can be obtained to distinguish user identity characteristics, including age, height, body shape, clothing, physiological state, etc. These characteristics are of reference value for judging the user's work type, behavior habits, needs, hobbies, group affiliation, etc.

[0064] (3) User's preferred objects; This part of the data is obtained through user feedback on different types of advertisements. This part of the content is constantly updated and optimized; it can basically depict the objects that users care about and prefer in their current state.

[0065] (4) User's disliked objects; This part of the data is obtained through user feedback on different types of advertisements. This part of the content is also constantly updated and optimized; it can basically depict the objects that users do not care about or dislike in their current state.

[0066] In this embodiment, as Figure 1 As shown, the method for creating the advertising analysis database in the VOC car owner big data platform is as follows:

[0067] Step 1: Create user tags for each user

[0068] 1. During the advertisement playback, the facial features of each user are acquired sequentially, and facial recognition is performed on the facial features.

[0069] 2. Based on the facial recognition results, query the advertising analytics database to determine whether the current user's facial features match the facial features of a historical user in the advertising analytics database:

[0070] (1) If yes, skip the current user.

[0071] (2) Otherwise, create an empty user tag for the current user; the user tag includes identity tag, likes tag and dislikes tag.

[0072] 3. Obtain multi-angle images of each user, and supplement the feature data in the identity tags of each user based on the image recognition results of the multi-angle images.

[0073] This step allows for the creation of a profile for each user, whether a new or existing user. As long as the user is present in the target area and can be detected, a profile and analysis can be performed. This enables the advertising analysis database established in this embodiment to reach a high scale and have a sufficiently rich sample size, laying a data foundation for later application development using this database.

[0074] In this embodiment, as Figure 2 As shown, the supplementary feature data in the identity label includes user ID, gender, age group, clothing style, and other features; other features represent identifiable non-gender, age group, and clothing style features that are useful for distinguishing user identity features.

[0075] The age range in the identity tag is one of the following categories categorized based on image recognition results: 0-10 years old, 10-20 years old, 20-30 years old, 30-50 years old, 50-70 years old, and over 70 years old. The clothing style in the identity tag includes casual, business, sports, children's, or elderly. In this embodiment, considering the significant impact of age on user needs, age is one of the essential identity features to consider. Furthermore, since conventional image information collection cannot directly obtain a user's occupational characteristics, this embodiment, by classifying users' clothing styles, can, to some extent, roughly categorize users' occupations or social identities.

[0076] Meanwhile, other characteristics in the identity tag reflect whether someone wears glasses, a hat, has hair loss, wears lipstick, wears high heels, has a beard, or wears a watch. If a characteristic is present, feature data reflecting that characteristic is added to the other characteristics; otherwise, no feature data is added. These other characteristics in the identity tag are all very typical features that differentiate users, and these features are highly correlated with different users' consumption needs. For example, women wearing lipstick and high heels may have a higher affinity for clothing and cosmetic advertisements. People with beards are generally not very concerned about razors. People with hair loss may be more interested in hair growth products and health supplements, etc.

[0077] In fact, by applying more diverse feature extraction techniques, this embodiment can obtain more different types of identity features. The richer the number of features obtained, the more refined the feature classification of users becomes.

[0078] Step 2: Obtain the feature data of the currently playing advertisement.

[0079] 1. Obtain the playback duration T of each advertisement and the keyword dataset associated with each advertisement.

[0080] The keyword dataset contains a set of pre-defined keywords related to the content of the currently playing advertisement. The keyword dataset for each advertisement includes at least the following features:

[0081] (1) Keywords that reflect the advertised products.

[0082] (2) Keywords that reflect the target customer group of the advertisement.

[0083] (3) Keywords that reflect the spokesperson or character in the advertisement.

[0084] (4) High-frequency or special keywords in advertising copy.

[0085] (5) Classification of advertising duration.

[0086] (6) Advertising style classification.

[0087] In this embodiment, a rich set of keywords is set for each advertisement. These keywords include various types of information that a customer can receive from an advertisement. When a user approves of an advertisement or gives positive feedback on its content, it can be assumed that some or all of the features in the keyword dataset of that advertisement are of interest or preference to the user. Conversely, when a user expresses aversion to an advertisement or gives negative feedback, it can be assumed that the user is indifferent to or averse to certain features in the keyword dataset of that advertisement. In this way, when the sample size of the feedback data collected from users on different types of advertisements is large enough, it is possible to analyze customer preferences and thus create a user profile.

[0088] Step 3: Obtain feedback data from each user regarding ad playback.

[0089] 1. Acquire audio stream data generated by all users within the advertising area during ad playback, monitor video stream data of all users within the advertising area, and receive instructions from one or more users within the advertising area to switch the currently playing ad.

[0090] The methods for users to change the currently playing advertisement include key input, voice interaction, and gesture interaction. Voice interaction is achieved by recognizing the user's spoken keywords requesting a change in the currently playing advertisement; gesture interaction is achieved by recognizing the user's characteristic hand gestures requesting a change in the currently playing advertisement; key input refers to the user directly inputting the key input command to change the currently playing advertisement.

[0091] Voice keywords are obtained by a speech recognition algorithm based on real-time speech stream data; feature gestures are obtained by a video action recognition algorithm based on real-time video stream data; and key input commands are obtained by a physical switching key module installed at the advertising playback site.

[0092] In this embodiment, user feedback mainly includes the following aspects:

[0093] (1) Changes in the user’s facial expressions when watching the advertisement.

[0094] (2) Direct user discussions about the advertisement. For example, discussing a particular actor or spokesperson in the advertisement, or the effectiveness of the product, etc.

[0095] (3) Gestures made by users while watching advertisements. For example, if a user points directly at the advertisement playback device to prompt other users to pay attention, it reflects that the user is interested in the currently playing advertisement.

[0096] (4) The time a user's attention is focused when watching an advertisement.

[0097] (5) The user requests to switch the currently playing advertisement. This directly reflects that the user does not like the advertisement.

[0098] Furthermore, when the technology is mature, other types of feedback can also be extracted and applied to later data analysis, such as users' laughter and other detailed characteristic actions.

[0099] 2. Determine whether an instruction to switch the currently playing advertisement has been received. If yes, assign the value 1 to the feature quantity SW that reflects the instruction; otherwise, assign the value 0 to SW.

[0100] Step 4: Calculate each user's approval rating for the current advertisement.

[0101] 1. Perform speech recognition on the speech stream data, extract keywords that match the feature data in the keyword dataset, and count their number N1.

[0102] 2. Perform video action recognition on the video stream data; extract the gestures that represent the user's response to the currently playing advertisement, and count their number N2.

[0103] Among them, the user's gestures in response to the currently playing advertisement include nodding, clapping, pointing at the advertisement interface with their hand, and turning their head from a non-direct gaze to a direct gaze.

[0104] 3. Perform video action recognition on the video stream data; extract feature actions that reflect changes in the eye focus position of each user, and calculate the attention duration t of each user for the currently playing advertisement based on the feature actions. n Where n represents the user ID of the current user.

[0105] The duration t of the user with ID n's attention to the currently playing advertisement. n The calculation method is as follows:

[0106]

[0107] In the above formula, t 1n This represents the duration of direct eye contact for user number n during the current ad playback; t 2n This represents the duration of time user number n has their eyes closed during the current advertisement playback; t 3n This represents the duration of time user number n looks down during the current advertisement playback; t 4n This indicates the duration of head-turning for user number n during the current ad playback.

[0108] In this embodiment, when calculating the duration of user attention to an advertisement, both the time spent directly viewing the advertisement interface and the time spent not directly viewing it are considered. This embodiment primarily obtains a relatively accurate attention duration by removing the time spent in the non-attentive state and then averaging it with the time spent in the attentive state.

[0109] 4. Perform frame-by-frame sampling on the video stream data according to the sampling frequency; perform image recognition on the sampled frames; extract the facial expressions of each user and classify them into like, ignore, or dislike; count the number of each user's three-category expression classification results and calculate the proportion of each user's three-category expression classification results in the total sample size of that user.

[0110] 5. Obtain the value of SW.

[0111] 6. Calculate the user's approval rating E for the current advertisement using the following formula. n :

[0112]

[0113] In the above formula, n represents the user ID of the current user, and E n E represents the rating of user ID n for the currently playing advertisement. n ≥0, and E n The higher the value, the higher the user's approval of the currently playing multimedia. The expression represents the user's attention level to the currently playing advertisement (number n); k1 represents the influence factor of voice feedback on the overall acceptance rating; k2 represents the influence factor of posture feedback on the overall acceptance rating; k3 represents the influence factor of facial expression feedback on the overall acceptance rating; k4 represents the influence factor of attention level on the overall acceptance rating; m1 represents the score of a single keyword in the voice feedback; m2 represents the score of a single posture in the posture feedback; m3 represents the score of attention level; a represents the score of the "like" expression; p 1,n To represent the proportion of images sampled in every frame where user ID n is categorized as "liked"; b represents the score for ignoring expressions, p 2,n To represent the proportion of user ID n who categorizes their facial expressions as ignored in the total number of images sampled every other frame; c represents the score of the disgust expression, p 3,n This represents the proportion of images sampled every frame where the user with ID n is classified as having an expression of disgust.

[0114] In this embodiment, facial expression recognition can be performed using a neural network algorithm trained on a large number of samples. There are also numerous readily applicable products for speech recognition, video action recognition, etc., and these aspects will not be elaborated upon in this embodiment.

[0115] In this embodiment, speech recognition, image recognition, and video motion recognition technologies are used to extract various types of user feedback information on the displayed advertisements from the user's voice stream data and video stream data. This feedback information is quantified using the method provided in this embodiment to obtain an evaluation result reflecting the user's level of acceptance of the current advertisement. This result reflects the user's degree of liking or disliking for the current advertisement and can therefore be used to characterize the user's needs or interests.

[0116] Step 5: Establishing or updating the advertising analysis database

[0117] 1. Set E n A high threshold E h and a low threshold E l E h E represents the threshold at which a user likes the currently playing ad. l E represents the threshold value at which a user dislikes the currently playing advertisement. l >0.

[0118] 2. When E n ≥E h And p 1，n +p 2，n ≥p 3，n When the ad is playing, the feature data of the keyword dataset associated with the currently playing ad is added to the current user's corresponding preference tag, and the feature data of the supplemented preference tag is deduplicated; then the feature data of the current user's corresponding aversion tag that is the same as the feature data in the keyword dataset is deleted.

[0119] 3. When E≤E l And p 2，n +p 3，n ≥p 1，n When the ad is playing, the feature data from the keyword dataset associated with the currently playing ad is added to the aversion tag corresponding to the current user, and the feature data of the aversion tag is deduplicated; then the feature data in the preference tag corresponding to the current user that matches the feature data in the keyword dataset is deleted.

[0120] 4. Update the user tags for each user to obtain a new user profile dataset for each user, and create an advertising analysis database.

[0121] Among them, such as Figure 3 As shown, the user profile dataset includes the corresponding user's facial feature data and user tags.

[0122] The core content of the advertising analysis database consists of preference and aversion tags derived from user behavior analysis. This data serves as direct evidence for subsequent user needs analysis. In this embodiment, user feedback while viewing advertisements can directly estimate user preferences, which should align with some or all of the features in the advertisement's keyword dataset. Therefore, after each advertisement is played, this embodiment analyzes and statistically studies user feedback to determine the user's accurate attitude towards the advertisement. Then, under specific conditions, the advertisement's keyword dataset is used as a feature in the current user's preference or aversion tags.

[0123] To avoid misclassification, a more rigorous review of the identified user attitudes is required. This embodiment introduces a specific threshold determined based on expert experience as the basis for judging the user's true attitude. In this embodiment, the threshold E... h and E l It was determined after repeated verification to have a high degree of credibility. This ensures that the final user profile is accurate and reliable.

[0124] Example 2

[0125] This embodiment provides a system for creating an advertising analysis database. The system uses the method for creating an advertising analysis database included in Embodiment 1 to realize the process of creating and updating the advertising analysis database.

[0126] like Figure 4 As shown, the creation system includes: a historical user query module, an advertising feature data extraction module, a user feedback data extraction module, a facial recognition module, an image recognition module, a voice recognition module, a video action recognition module, a user tag creation module, a recognition rating calculation module, and a database creation module.

[0127] The historical user query module is used to query an advertising analysis database and extract the user profile dataset of historical users that has been collected. The user profile dataset includes facial feature data of each historical user and user tags, including identity tags, preference tags, and aversion tags.

[0128] The ad feature data extraction module is used to extract the playback duration T of each ad played in an ad delivery system, as well as a keyword dataset associated with that ad.

[0129] The user feedback data extraction module is used for: (1) obtaining voice information generated by users watching the advertisement within the advertising delivery area when the advertisement is played in the advertising delivery system, and obtaining voice stream data related to each advertisement. (2) obtaining multi-angle monitoring videos of all users watching the advertisement within the advertising delivery area when the advertisement is played in the advertising delivery system, and obtaining video stream data related to each advertisement. (3) obtaining a switching command issued by a user watching the advertisement when the advertisement is played in the advertising delivery system. The switching command includes keyboard input command, voice interaction command or gesture interaction command; and assigning the feature quantity SW representing the switching command to 1 when the acquisition is successful, otherwise assigning SW to 0.

[0130] The facial recognition module is used to obtain an image dataset from video stream data through frame segmentation, extract the facial features of each user appearing in the image dataset, and complete the comparison process between the facial features of the current user and the facial features of each historical user in the advertising analysis database to distinguish between new users and historical users.

[0131] The image recognition module is used to perform image recognition on the image dataset obtained by frame-segmenting video stream data, and then: (1) obtain various feature data reflecting the identity characteristics of new users. (2) extract the facial expressions of all users during the advertisement playback and classify the facial expressions into one of the following: like, ignore or dislike.

[0132] The speech recognition module is used to perform speech recognition on the speech stream data, and then: (1) obtain the voice interaction command issued by the user during the advertisement playback, which represents the user's request to switch the currently playing advertisement. (2) extract all words in the speech stream data and find the keywords that match the feature data in the keyword dataset.

[0133] The video action recognition module is used to perform video action recognition on video stream data, and then: (1) extract the gesture interaction command issued by a user in the video stream data, which represents the user's request to switch the currently playing advertisement. (2) extract the posture action issued by a user in the video stream data, which represents the user's response to the currently playing advertisement. (3) extract the feature actions that reflect the changes in the user's eye focus position during the current advertisement playback.

[0134] The user tag creation module is used to create an empty user tag for each newly identified user and supplement the corresponding user's identity tag with the feature data reflecting the identity characteristics of the new user obtained by the image recognition module.

[0135] The recognition evaluation value calculation module is used for: (1) obtaining keywords that match the feature data in the keyword dataset and are identified by the speech recognition module from the speech stream data, and counting their number N1. (2) obtaining the posture actions that reflect the user's feedback to the currently playing advertisement and are identified by the video action recognition module, and counting their number N2. (3) obtaining the feature actions that reflect the change in a user's eye focus position during the current advertisement playback and are identified by the video action recognition module, and calculating the current user's attention duration t for the currently playing advertisement based on the feature actions. n Where n represents the user ID of the current user. The approval rating calculation module calculates the attention duration t of user ID n for the currently playing advertisement. n The calculation formula is as follows:

[0136]

[0137] In the above formula, t 1n This represents the duration of direct eye contact for user number n during the current ad playback; t 2n This represents the duration of time user number n has their eyes closed during the current advertisement playback; t 3n This represents the duration of time user number n looks down during the current advertisement playback; t 4n This indicates the duration of head-turning for user number n during the current ad playback.

[0138] (4) Obtain the number of three-category expression classification results for each user identified by the image recognition module, and calculate the proportion of each user's three-category expression classification results in the total sample size. (5) Obtain the value of SW. (6) Calculate the approval rating value E of each user for the current advertisement using the following formula. n :

[0139]

[0140] In the above formula, n represents the user ID of the current user, and E n E represents the rating of user ID n for the currently playing advertisement. n ≥0, and E n The higher the value, the higher the user's approval of the currently playing multimedia. The expression represents the user's attention level to the currently playing advertisement (number n); k1 represents the influence factor of voice feedback on the overall acceptance rating; k2 represents the influence factor of posture feedback on the overall acceptance rating; k3 represents the influence factor of facial expression feedback on the overall acceptance rating; k4 represents the influence factor of attention level on the overall acceptance rating; m1 represents the score of a single keyword in the voice feedback; m2 represents the score of a single posture in the posture feedback; m3 represents the score of attention level; a represents the score of the "like" expression; p 1,n To represent the proportion of images sampled in every frame where user ID n is categorized as "liked"; b represents the score for ignoring expressions, p 2,n To represent the proportion of user ID n who categorizes their facial expressions as ignored in the total number of images sampled every other frame; c represents the score of the disgust expression, p 3,n This represents the proportion of images sampled every frame where the user with ID n is classified as having an expression of disgust.

[0141] The database creation module is used to: (1) set up E based on expert experience. n A high threshold E h and a low threshold E l Among them, E h E represents the threshold at which a user likes the currently playing ad. l E represents the threshold value at which a user dislikes the currently playing advertisement. l >0. (2) Make the following judgments and decisions for each user: (i) When E n ≥E h And p 1，n +p 2，n ≥p 3，n When the current ad is playing, the feature data from the keyword dataset associated with it is added to the current user's corresponding preference tag, and the feature data of the supplemented preference tag is deduplicated; then, the feature data in the current user's corresponding aversion tag that is identical to the feature data in the keyword dataset is deleted. (ii) When E≤E l And p 2，n +p 3，n ≥p 1，n At that time, the feature data of the keyword dataset associated with the currently playing advertisement is added to the aversion tag corresponding to the current user, and the feature data of the supplemented aversion tag is deduplicated; then the feature data of the current user's preference tag that matches the feature data in the keyword dataset is deleted. (3) Update the user tags of each user in turn to obtain the new user profile dataset of each user, and then complete the creation or update of the advertising analysis database. Among them, the user profile dataset includes the facial feature data and user tags of the corresponding user.

[0142] In this embodiment, the advertising analysis database is initially empty. After the first historical user's user profile dataset is entered into it, the advertising analysis database creation system compares the current user's facial features with the facial features of historical users in the advertising analysis database to determine whether the current user is a new user or a historical user. The system then enters the user profile dataset of the newly identified user into the advertising analysis database, or updates the user tags in the existing user profile dataset of historical users in the advertising analysis database.

[0143] Example 3

[0144] Based on Example 2, this example provides a method for evaluating user acceptance of advertisements based on feature recognition, such as... Figure 5 As shown, the method includes the following steps:

[0145] Step 1: Obtain the feature data of the currently playing advertisement.

[0146] Obtain the playback duration T of each advertisement and the keyword dataset associated with each advertisement.

[0147] The keyword dataset contains a set of pre-defined keywords related to the content of the currently playing advertisement. The keyword dataset for each advertisement includes at least the following features:

[0148] (1) Keywords that reflect the advertised products.

[0149] (2) Keywords that reflect the target customer group of the advertisement.

[0150] (3) Keywords that reflect the spokesperson or character in the advertisement.

[0151] (4) High-frequency or special keywords in advertising copy.

[0152] (5) Classification of advertising duration.

[0153] (6) Advertising style classification.

[0154] Step 2: Obtain feedback data from each user regarding ad playback.

[0155] 1. Acquire audio stream data generated by all users within the advertising area during ad playback, monitor video stream data of all users within the advertising area, and receive instructions from one or more users within the advertising area to switch the currently playing ad.

[0156] The methods for users to change the currently playing advertisement include key input, voice interaction, and gesture interaction. Voice interaction is achieved by recognizing the user's spoken keywords requesting a change in the currently playing advertisement; gesture interaction is achieved by recognizing the user's characteristic hand gestures requesting a change in the currently playing advertisement; key input refers to the user directly inputting the key input command to change the currently playing advertisement.

[0157] Voice keywords are obtained by a speech recognition algorithm based on real-time speech stream data; feature gestures are obtained by a video action recognition algorithm based on real-time video stream data; and key input commands are obtained by a physical switching key module installed at the advertising playback site.

[0158] In this embodiment, user feedback mainly includes the following aspects:

[0159] (1) Changes in the user’s facial expressions when watching the advertisement.

[0160] (2) Direct user discussions about the advertisement. For example, discussing a particular actor or spokesperson in the advertisement, or the effectiveness of the product, etc.

[0161] (3) Gestures made by users while watching advertisements. For example, if a user points directly at the advertisement playback device to draw the attention of other users, it reflects that the user is interested in the currently playing advertisement.

[0162] (4) The time a user's attention is focused when watching an advertisement.

[0163] (5) The user requests to switch the currently playing advertisement. This directly reflects that the user does not like the advertisement.

[0164] Furthermore, when the technology is mature, other types of feedback can also be extracted and applied to later data analysis, such as users' laughter and other detailed characteristic actions.

[0165] 2. Determine whether an instruction to switch the currently playing advertisement has been received. If yes, assign the value 1 to the feature quantity SW that reflects the instruction; otherwise, assign the value 0 to SW.

[0166] Step 3: Calculate each user's approval rating for the current advertisement.

[0167] 1. Perform speech recognition on the speech stream data, extract keywords that match the feature data in the keyword dataset, and count their number N1.

[0168] 2. Perform video action recognition on the video stream data; extract the gestures that represent the user's response to the currently playing advertisement, and count their number N2.

[0169] Among them, the user's gestures in response to the currently playing advertisement include nodding, clapping, pointing at the advertisement interface with their hand, and turning their head from a non-direct gaze to a direct gaze.

[0170] 3. Perform video action recognition on the video stream data; extract feature actions that reflect changes in the eye focus position of each user, and calculate the attention duration t of each user for the currently playing advertisement based on the feature actions. n Where n represents the user ID of the current user.

[0171] The duration t of the user with ID n's attention to the currently playing advertisement. n The calculation method is as follows:

[0172]

[0173] In the above formula, t 1n This represents the duration of direct eye contact for user number n during the current ad playback; t 2n This represents the duration of time user number n has their eyes closed during the current advertisement playback; t 3n This represents the duration of time user number n looks down during the current advertisement playback; t 4n This indicates the duration of head-turning for user number n during the current ad playback.

[0174] In this embodiment, when calculating the duration of user attention to an advertisement, both the time spent directly viewing the advertisement interface and the time spent not directly viewing it are considered. This embodiment primarily obtains a relatively accurate attention duration by averaging the time spent in the non-attention state with the time spent in the attention state.

[0175] 4. Perform frame-by-frame sampling on the video stream data according to the sampling frequency; perform image recognition on the sampled frames; extract the facial expressions of each user and classify them into like, ignore, or dislike; count the number of each user's three-category expression classification results and calculate the proportion of each user's three-category expression classification results in the total sample size of that user.

[0176] 5. Obtain the value of SW.

[0177] 6. Calculate the user's approval rating E for the current advertisement using the following formula. n :

[0178]

[0179] In the above formula, n represents the user ID of the current user, and E nE represents the rating of user ID n for the currently playing advertisement. n ≥0, and E n The higher the value, the higher the user's approval of the currently playing multimedia. The expression represents the user's attention level to the currently playing advertisement (number n); k1 represents the influence factor of voice feedback on the overall acceptance rating; k2 represents the influence factor of posture feedback on the overall acceptance rating; k3 represents the influence factor of facial expression feedback on the overall acceptance rating; k4 represents the influence factor of attention level on the overall acceptance rating; m1 represents the score of a single keyword in the voice feedback; m2 represents the score of a single posture in the posture feedback; m3 represents the score of attention level; a represents the score of the "like" expression; p 1,n To represent the proportion of images sampled in every frame where user ID n is categorized as "liked"; b represents the score for ignoring expressions, p 2,n To represent the proportion of user ID n who categorizes their facial expressions as ignored in the total number of images sampled every other frame; c represents the score of the disgust expression, p 3,n This represents the proportion of images sampled every frame where the user with ID n is classified as having an expression of disgust.

[0180] The method provided in this embodiment can identify various characteristics of user feedback during advertisement playback, and then derive the user's approval rating of the advertisement. This method can collect various types of user feedback, resulting in more accurate user approval ratings of the advertisement, which can serve as a basis for evaluating the effectiveness of advertising.

[0181] Example 4

[0182] This embodiment provides an intelligent media management system based on a VOC (Vehicle Owner Big Data) platform. This system obtains the matching degree between the current user and the advertisements to be displayed, and then adjusts the advertisement playback sequence. In implementing its functions, this intelligent media management system relies on the VOC platform described in Embodiment 1; more precisely, it requires data from the advertisement analysis database established within the VOC platform.

[0183] In this embodiment, as Figure 6 As shown, the intelligent media management system includes: a keyword extraction module, a historical user information query module, a user type classification module, a user tag creation module, an identity feature recognition module, a target profile dataset creation module, and an advertisement playback sequence adjustment module.

[0184] The keyword extraction module is used to extract the keyword dataset associated with each advertisement in the advertisement playback sequence table. The feature data in the keyword dataset consists of multiple pre-defined keywords related to the content of the advertisement.

[0185] In this embodiment, the feature data within the keyword dataset for each advertisement is the same as in Embodiment 1, including at least:

[0186] (1) Keywords that reflect the advertised products.

[0187] (2) Keywords that reflect the target customer group of the advertisement.

[0188] (3) Keywords that reflect the spokesperson or character in the advertisement.

[0189] (4) High-frequency or special keywords in advertising copy.

[0190] (5) Classification of advertising duration.

[0191] (6) Advertising style classification.

[0192] The historical user information query module is used to query a user profile dataset of each historical user from an advertising analysis database to obtain various feature data about each historical user from the user profile dataset. The advertising analysis database is stored in the VOC Car Owner Cloud Big Data Platform. This advertising analysis database is the database created in Example 1. The advertising analysis database contains the collected user profile dataset of historical users; the user profile dataset includes facial feature data and user tags for each historical user. User tags include identity tags, preference tags, and aversion tags; the identity tags store feature data reflecting the user's identity characteristics, including: gender, age group, clothing style, and other features; other features represent identifiable features that are not related to gender, age group, or clothing style and are useful for distinguishing user identity characteristics. The preference tags store feature data reflecting the objects the user likes, and the aversion tags store feature data reflecting the objects the user dislikes. Among the features extracted by the identity feature recognition module, the age groups are divided into 0-10 years old, 10-20 years old, 20-30 years old, 30-50 years old, 50-70 years old, or 70 years and above; the clothing styles include casual, business, sports, children's, or elderly. Other features reflected in the identity tag include whether someone wears glasses, a hat, has hair loss, wears lipstick, wears high heels, has a beard, or wears a wristwatch. If any of these features are present, feature data reflecting that feature is added to the other features; otherwise, no feature data is added to the other features.

[0193] The user type classification module is used to extract the facial features of all target users within an advertising delivery area; then, the extracted facial features are compared with the facial features of all historical users in the advertising analysis database to distinguish whether the current user is a historical user or a new user.

[0194] In this embodiment, the data sources for both the user type classification module and the identity feature recognition module are video stream data from multi-angle monitoring of the advertising placement area. The user type classification module includes a facial feature extraction unit, a facial feature comparison unit, and a user type classification unit. The facial feature extraction unit extracts the facial features of all users appearing in the video stream data. The facial feature comparison unit obtains all facial features extracted by the facial feature extraction unit and all facial features of historical users queried by the historical user information query module, and compares the two features. The user type classification unit classifies all users appearing in the video stream data into historical users or new users based on the results distinguished by the facial feature comparison unit.

[0195] The user tag creation module is used to create an empty user tag for each new user. The created user tag includes an identity tag, a like tag, and an dislike tag. The user tag creation module also adds a unique user ID to the identity tag of each new user.

[0196] The identity feature recognition module is used to extract the identity features of new users and add the extracted identity features to the corresponding identity tags of new users.

[0197] The target profile dataset creation module is used for:

[0198] (1) Set a critical value q0 for the proportion of historical users, and calculate the proportion q of current users identified as historical users in the current user group within the advertising area.

[0199] (2) Determine the relationship between q and q0, and make the following decision based on the result:

[0200] (i) When q≥q0, extract the feature data from the preference tags of all historical users, remove duplicates from the above feature data, and use it as the target profile dataset for the current user group.

[0201] (ii) When q < q0, first extract the feature data from the preference tags of all historical users. Then, calculate the overlap (Dc1) between the content of each new user's identity tag and the content of each historical user's identity tag in the advertising analysis database. The formula for calculating Dc1 is as follows:

[0202]

[0203] Extract feature data from the preference tags of the historical user with the highest overlap (Dc1) with the identity tags of each new user. Merge the two sets of feature data and remove duplicates; this will form the target profile dataset for the current user group.

[0204] The ad playback sequence adjustment module is used to: (1) calculate the overlap degree Dc2 between the feature data in the keyword dataset associated with each ad extracted by the keyword extraction module and the feature data in the target profile dataset. The formula for calculating Dc2 is as follows:

[0205]

[0206] (2) According to the Dc2 calculation results of each advertisement in descending order, the advertisements in the advertisement playback sequence table are reordered to obtain the adjusted advertisement playback sequence table.

[0207] In this embodiment, the intelligent media management system based on the VOC vehicle owner big data platform is applied to an advertising delivery system with multi-angle monitoring equipment. The advertising delivery system is used to play the advertisements to be delivered according to the advertisement playback sequence list; the multi-angle monitoring equipment is used to obtain video stream data of multi-angle monitoring of all target users in the advertising delivery area of ​​the advertising delivery equipment.

[0208] Of course, in other embodiments, the relevant advertising delivery system and multi-angle monitoring equipment can also be included as part of the intelligent media management system based on the VOC vehicle owner big data platform in this embodiment. This allows for integrated and coordinated control of the data collection, data processing, data analysis, and advertising delivery processes required in this embodiment.

[0209] Example 5

[0210] This embodiment provides a method for precise advertising targeting based on user profiles. This precise targeting method is applied in an intelligent media management system based on a VOC (Vehicle Owner Big Data) platform, as described in Embodiment 4; for example... Figure 7 As shown, this precise targeting method includes the following steps:

[0211] Step 1: Obtain the current user's user tags

[0212] 1. Obtain the facial features of each current user within the advertising area.

[0213] 2. Perform facial recognition on each current user sequentially. Based on the facial recognition results, query an advertising analysis database containing user profile datasets of multiple historical users (i.e., the advertising analysis database created in Example 1 or 2), and make the following judgments:

[0214] (1) When the facial features of the current user match the feature data in the facial feature data of one of the historical users, obtain all the feature data in the user tag of the historical user.

[0215] (2) When the facial features of the current user do not match the facial feature data of all historical users, the current user is determined to be a new user and an empty user label is created for the new user.

[0216] The user profile dataset contains facial feature data and user tags of corresponding historical users; user tags include identity tags, preference tags, and aversion tags.

[0217] 3. Obtain multi-angle images of the new user, perform image recognition on the multi-angle images, and supplement the feature data in the identity label of the new user based on the recognition results; the supplemented feature data in the identity label includes user ID, gender, age group, clothing style and other features; other features represent the non-gender, age group and clothing style features that can be identified and are useful for distinguishing user identity features.

[0218] Step 2: Establish a target profile dataset for the current user group

[0219] 1. Set a historical user ratio threshold q0, and calculate the proportion q of current users identified as historical users in the current user group within the advertising area.

[0220] 2. Determine the relationship between q and q0, and make the following decision based on the result:

[0221] (1) When q≥q0, extract the feature data from the preference tags of all historical users, remove duplicates from the above feature data, and use it as the target profile dataset of the current user group.

[0222] (2) When q < q0, first extract the feature data from the preference tags of all historical users; then calculate the overlap between the content of the identity tags of each new user and the content of the identity tags of each historical user, Dc1. The formula for calculating Dc1 is as follows:

[0223]

[0224] Extract the feature data from the preference tags of the historical user with the highest overlap (Dc1) with the identity tags of each new user; merge the two parts of feature data (the identified historical user and the preference tags of the historical user with the highest overlap with the identity tags of each new user) and after deduplication of the feature data, use them as the target profile dataset for the current user group.

[0225] Step 3: Adjust the playback order of advertisements in the advertisement playback sequence list.

[0226] 1. Obtain the keyword dataset associated with each advertisement in the advertisement playback sequence table. The feature data in the keyword dataset consists of multiple pre-defined keywords related to the content of the currently playing advertisement.

[0227] 2. Obtain the feature data from the target profile dataset, and calculate the overlap (Dc2) between the feature data from the keyword dataset associated with each advertisement and the feature data from the target profile dataset. The formula for calculating Dc2 is as follows:

[0228]

[0229] 3. Sort the advertisements in the advertisement playback sequence table in descending order of their Dc2 calculation results to obtain a readjusted advertisement playback sequence table.

[0230] The method for adjusting the ad playback sequence table in an ad delivery system provided in this implementation is based on the following principles and implementation logic:

[0231] Since this embodiment has already obtained the data from the advertising analysis database created in embodiment 1, when advertising is placed, facial recognition is performed on all users in the advertising placement area to distinguish whether these users belong to historical users in the advertising analysis database or are newly added users that have not been collected by the advertising analysis database.

[0232] Considering that advertising analytics data has already enabled the profiling of historical users—that is, the rich feature data within user tags—when the vast majority of users in the advertising target area belong to historical users, their needs and preferences can be considered representative of the current user group. By obtaining the corresponding historical users' preference tags and then extracting their feature data, a target profile dataset can be obtained to characterize the preferences or needs of the current user group.

[0233] When the number of new users in the advertising area reaches a certain level, it becomes insufficient to rely solely on historical users for user profiling. Real-time analysis of these new users at this point is clearly too slow. However, since this implementation has access to a sufficiently large and rich advertising analysis dataset, it can identify the identity features of these new users (achieved through image recognition technology). These identity features are then compared with user tags in the advertising analysis dataset to extract the most relevant historical user. This historical user's tag is temporarily used as the new user's tag, thereby obtaining the features in the new user's preference tags. Because user identity features (such as age, height, gender, clothing, and physiological characteristics) are highly correlated with user needs or preferences (features in preference tags), this approximate substitution in this implementation should have high reliability. This implementation, through this technical solution, can obtain a target user profile dataset containing a large number of new users.

[0234] After obtaining the target profile dataset of the user group in the advertising area, this embodiment further compares the feature data in the target profile dataset with the keyword dataset of each advertisement to be played. The overlap between the two can be found. The higher the overlap, the more likely the user group is the target customer of the advertisement. In this case, these advertisements should be placed in a priority position. Based on this logic, this embodiment realizes the reordering of the advertisement playback sequence list to ensure that the most suitable advertisement can be delivered to the target group with higher priority.

[0235] Additionally, it should be noted that, using completely similar hardware and software, in another embodiment, it is also possible to simultaneously adjust the advertisement playback sequence and update the user tags of new users based on user feedback on the played advertisements. This further enriches the data content of the advertisement analysis database in the VOC car owner cloud big data platform. Thus, the technical solutions in Embodiments 1 and 2 are combined.

[0236] Example 6

[0237] This embodiment provides a method for timely analysis of user needs in a business district setting. This method is a further development of the method in Embodiment 5, enabling the most direct and rapid prediction or assessment of the user needs of specific users. For example... Figure 8 As shown, the method includes the following steps:

[0238] Step 1: Obtain the facial features of the current user within the advertising area.

[0239] Step 2: Perform facial recognition on the current user sequentially. Based on the facial recognition results, query an advertising analysis database containing user profile datasets of multiple historical users (this advertising analysis database is the same as the advertising analysis database in the aforementioned embodiment) and make the following judgment:

[0240] (1) When the facial features of the current user match the feature data in the facial feature data of one of the historical users, obtain all the feature data in the user tag of the historical user.

[0241] (2) When the facial features of the current user do not match the facial feature data of all historical users, the current user is determined to be a new user and an empty user label is created for the new user.

[0242] The user profile dataset contains facial feature data and user tags of corresponding historical users; user tags include identity tags, preference tags, and aversion tags.

[0243] Step 3: Obtain multi-angle images of the new user, perform image recognition on the multi-angle images, and supplement the feature data in the identity label of the new user based on the recognition results; the supplemented feature data in the identity label includes user ID, gender, age group, clothing style and other features; other features represent the non-gender, age group and clothing style features that can be identified and are useful for distinguishing user identity features.

[0244] Step 4: Compare all feature data in the identity tag with the identity tags of each historical user in the advertising analysis database, and calculate the feature overlap degree Dc3 between the two. The formula for calculating Dc3 is as follows:

[0245]

[0246] Step 5: Extract the feature data from the like and dislike tags of the historical users with the highest feature overlap (Dc3 value) with the current user in the advertising analysis database, and populate the user profile dataset of the new user with the above feature data to complete the timely analysis of the user needs of the current user.

[0247] Analyzing the above process reveals that the method in this embodiment can analyze and identify users as soon as they appear on screen, thereby establishing a predicted profile dataset of characteristics and behaviors to predict users' likes and dislikes; and based on this prediction, timely analysis of user needs can be achieved. This analytical method is more timely and effective, eliminating the need for long-term "tracking" and evaluation of users. Therefore, it has high practical value. It should be noted that the accuracy of this timely analysis result is highly correlated with the sample size of the advertising analysis database, which contains user profile datasets of multiple historical users. The larger the sample size of the advertising analysis database, the more accurate the results of this timely analysis.

[0248] The logic of the method in this embodiment is as follows: First, the facial features of the user appearing in a specific scenario are acquired. It is then determined whether the user's data sample has been included in the advertising analysis database. If so, the user's like and dislike tags included in the advertising analysis database are directly extracted, forming the user profile dataset for that user. This dataset is then used to analyze and predict the user's needs. If the user's data sample is not included in the advertising analysis database, the user's identity features are first extracted. Then, the like and dislike tags from the identity tags of the historical users whose identity features are most similar to the current user (determined by Dc3) are extracted from the advertising analysis database and used as the current user's user profile dataset for analyzing the user's needs.

[0249] It should be noted that this embodiment addresses the problem of quickly creating a profile of a single user and rapidly analyzing user needs, while embodiment 5 addresses the problem of creating a profile of the entire user group and matching the user needs of the entire group with the advertisements to be delivered. The two are related but have significant differences.

[0250] Example 7

[0251] This embodiment provides a method for matching user needs with advertising content. This method, developed based on Embodiments 1 and 6, is used to select the advertisement that best matches the current user from the currently available advertisements; for example... Figure 9 As shown, the matching method includes the following steps:

[0252] Step 1: Obtain the keyword dataset of all ads to be delivered; this keyword dataset is the keyword dataset established in any of the aforementioned embodiments; the keyword dataset contains keywords that reflect various feature data of the ad content.

[0253] Step 2: Obtain the user profile dataset of the current user. The user profile dataset is the final result obtained by the real-time analysis method of user needs in a business district scenario provided in the aforementioned embodiment.

[0254] Step 3: Calculate the matching degree Dc4 between the feature data in the keyword dataset of each advertisement and the data in the current user profile dataset. The formula for calculating Dc4 is as follows:

[0255]

[0256] Step 4: Select the advertisement with the highest Dc4 value as the most suitable advertisement for the current user, thus completing the matching process between user needs and advertisement content.

[0257] The best-matched ads are those that best address the user's actual needs, thus achieving the best promotional results. In practice, for the identified user, the most relevant ad should be prioritized for delivery.

[0258] In this embodiment, the method for matching user needs and advertising content is feature matching. In this feature matching process, the features representing user needs (features in preference tags) are derived from user feedback during historical ad playback, and the feature data are the corresponding ad keywords. Therefore, feature matching with the actual ads to be delivered is usually very easy to achieve, and considering that user preferences are typically consistent and long-term, the results of this feature matching are also more accurate.

[0259] Example 8

[0260] This embodiment provides a garage giant screen MAX smart terminal with intelligent voice interaction function. This terminal updates the advertisement playback sequence table based on user interaction during advertisement playback. This embodiment represents a deeper development and application of the technical solutions and achievements described in the foregoing embodiments. The garage giant screen MAX smart terminal with intelligent voice interaction function in this embodiment employs some of the processing methods and device modules from the foregoing embodiments.

[0261] Specifically, such as Figure 10 As shown, the garage giant screen MAX smart terminal provided in this embodiment includes: an advertising playback module, a voice acquisition module, a video monitoring module, an advertising feature data extraction module, a user feedback data extraction module, an image recognition module, a voice recognition module, a video action recognition module, a human-computer interaction module, a recognition evaluation value calculation module, and an advertising playback sequence update module.

[0262] The advertising playback module is used to play each advertisement to be played sequentially according to the advertising playback sequence list, and to switch the currently playing advertisement upon receiving a switching command from the human-computer interaction module. The advertising playback module is a garage giant MAX display screen.

[0263] The voice acquisition module is used to collect voice information generated by users watching the advertisements around the advertising playback module when each advertisement is played. The voice acquisition module consists of multiple microphones installed around the Garage Giant Screen MAX display screen; the microphones are distributed on one side facing the display surface of the Garage Giant Screen MAX display screen.

[0264] The video surveillance module is used to monitor the user group watching the advertisements around the advertising playback module from multiple angles while each advertisement is playing. The video surveillance module's field of view is one side facing the MAX display screen in the garage, and it includes multiple surveillance cameras, each capturing images of the field of view from different angles.

[0265] The advertising feature data extraction module is used to extract the playback duration T of each advertisement played by the advertising playback module, as well as a keyword dataset associated with that advertisement.

[0266] The user feedback data extraction module is used to: (1) receive voice information collected by the voice acquisition module and obtain voice stream data related to each advertisement; (2) receive multi-angle monitoring video collected by the video monitoring module and obtain video stream data related to each advertisement; (3) obtain the switching instruction issued by the human-computer interaction module requesting the switching of the currently playing advertisement, and when the switching instruction is received, assign the feature quantity SW representing the switching instruction to 1, otherwise assign SW to 0.

[0267] The image recognition module performs image recognition on the image dataset obtained by frame-segmenting video stream data, thereby extracting all user facial expressions during advertisement playback and classifying them into one of four categories: like, ignore, or dislike. The image recognition module includes an expression recognition unit, which uses a neural network recognition algorithm trained on a large training set to classify user facial expressions in the images.

[0268] The speech recognition module is used to perform speech recognition on the speech stream data, and then: (1) obtain the voice interaction commands issued by the user during the advertisement playback, which represent the request to switch the currently playing advertisement. (2) extract all words in the speech stream data and find the keywords that match the feature data in the keyword dataset.

[0269] The speech recognition module includes a speech interaction command extraction unit and a keyword extraction unit. The speech interaction command extraction unit sends the extracted speech interaction commands to the speech interaction unit in the human-computer interaction module. The keyword extraction unit sends the extracted keywords that match the feature data in the keyword dataset to the recognition evaluation value calculation module.

[0270] The video action recognition module is used to perform video action recognition on video stream data, and then: (1) extract the gesture interaction command issued by a user in the video stream data that represents the request to switch the currently playing advertisement; (2) extract the posture action issued by a user in the video stream data that represents the feedback to the currently playing advertisement; (3) extract the feature action that reflects the change of a user's eye focus position during the current advertisement playback.

[0271] The video action extraction module includes a gesture interaction command extraction unit, a posture action feedback extraction unit, and an eye movement feature extraction unit. The gesture interaction command extraction unit sends the extracted gesture interaction commands to the gesture interaction unit in the human-computer interaction module. The posture action feedback extraction unit and the eye movement feature extraction unit send the extracted feature data to the approval rating calculation module.

[0272] The human-computer interaction module is used to receive user commands to switch the currently playing advertisement and to issue a switching command; for example, Figure 11 As shown, users can request to switch the currently playing advertisement through button input, voice interaction, and gesture interaction. The human-computer interaction module includes a physical button module for receiving button input commands from the user requesting to switch the currently playing advertisement. The module also includes a voice interaction unit and a gesture interaction unit. The voice interaction unit acquires the user's voice command requesting to switch the currently playing advertisement; this command is obtained by a voice recognition module based on real-time voice stream data. The gesture interaction unit acquires the user's gesture command requesting to switch the currently playing advertisement; this command is obtained by a video motion recognition module based on real-time video stream data.

[0273] The recognition evaluation value calculation module is used to: (1) obtain keywords that match the feature data in the keyword dataset and are identified by the speech recognition module, and count their number N1. (2) obtain the gestures and actions that represent the user's response to the currently playing advertisement and are identified by the video action recognition module, and count their number N2. (3) obtain the feature actions that reflect the change in a user's eye focus position during the current advertisement playback and are identified by the video action recognition module, and calculate the user's attention duration t for the currently playing advertisement based on the feature actions. n ; Duration of attention t n The calculation formula is as follows:

[0274]

[0275] In the above formula, t 1n This represents the duration of direct eye contact for user number n during the current ad playback; t2n This represents the duration of time user number n has their eyes closed during the current advertisement playback; t 3n This represents the duration of time user number n looks down during the current advertisement playback; t 4n (4) Obtain the number of three-category expression classification results for each user identified by the image recognition module, and calculate the proportion of the number of three-category expression classification results for each user in the total sample size. (5) Obtain the value of SW. (6) Calculate the approval rating value E of each user for the current advertisement using the following formula. n :

[0276]

[0277] In the above formula, n represents the user ID of the current user, E n E represents the rating of user ID n for the currently playing advertisement. n ≥0, and E n The higher the value, the higher the user's approval of the currently playing multimedia. The expression represents the user's attention level to the currently playing advertisement (number n); k1 represents the influence factor of voice feedback on the overall acceptance rating; k2 represents the influence factor of posture feedback on the overall acceptance rating; k3 represents the influence factor of facial expression feedback on the overall acceptance rating; k4 represents the influence factor of attention level on the overall acceptance rating; m1 represents the score of a single keyword in the voice feedback; m2 represents the score of a single posture in the posture feedback; m3 represents the score of attention level; a represents the score of the "like" expression; p 1,n The percentage of images sampled in every frame for user number n who are categorized as having favorite expressions; b represents the score for ignoring expressions, p 2,n The percentage of images sampled every other frame for user number n who are classified as ignoring facial expressions; c represents the score for the disgust expression, p 3,n The percentage of images sampled every frame for user number n who are classified as having expressions of disgust.

[0278] The ad playback sequence update module is used to: (1) obtain the average approval rating of each ad in the entire played ad sequence table within an update period. The calculation formula is as follows:

[0279]

[0280] In the above formula, i represents the number of each advertisement in the advertisement playback sequence list. (2) According to each advertisement's (3) Obtain the ads that need to be added and their quantity, delete the corresponding number of ads that are ranked lower in the rating ranking table from the ad playback sequence table, and add the ads that need to be added to the ad playback sequence table to complete the update process of the ad playback sequence table.

[0281] Example 9

[0282] This embodiment provides a method for updating advertisements on an advertising device. This method is applied to the garage giant screen MAX smart terminal with intelligent voice interaction function in Embodiment 8; as follows: Figure 12 As shown, the method for updating advertisements includes the following steps:

[0283] Step 1: Obtain the feature data of the currently playing advertisement.

[0284] 1. Obtain the playback duration T of the currently playing advertisement, and the keyword dataset associated with the advertisement.

[0285] The keyword dataset contains pre-defined keywords related to the content of the currently playing advertisement. The feature data associated with each advertisement within the keyword dataset includes at least the following:

[0286] (1) Keywords that reflect the advertised products.

[0287] (2) Keywords that reflect the target customer group of the advertisement.

[0288] (3) Keywords that reflect the spokesperson or character in the advertisement.

[0289] (4) High-frequency or special keywords in advertising copy.

[0290] (5) Classification of advertising duration: very short, short, medium, long, and extra long.

[0291] (6) Advertising style classification.

[0292] Step 2: Obtain feedback data from each user regarding ad playback.

[0293] 1. Acquire audio stream data generated by the current user group during ad playback, monitor video stream data of the current user group, and receive instructions from one or more users in the current user group to switch the currently playing ad.

[0294] 2. Determine whether an instruction to switch the currently playing advertisement has been received. If yes, assign the value 1 to the feature quantity SW that reflects the instruction; otherwise, assign the value 0 to SW.

[0295] Step 3: Calculate each user's approval rating for the current advertisement.

[0296] 1. Perform speech recognition on the audio stream data, extract keywords that match the feature data in the keyword dataset associated with the currently playing advertisement, and count their number N1.

[0297] 2. Perform video action recognition on the video stream data; extract the gestures that represent the user's response to the currently playing advertisement, and count their number N2.

[0298] 3. Perform video action recognition on the video stream data; extract feature actions that reflect changes in the current user's eye focus position, and calculate the current user's attention duration t for the currently playing advertisement based on the feature actions. n Where n represents the user ID of the current user.

[0299] 4. Perform frame-by-frame sampling on the video stream data according to the sampling frequency; perform image recognition on the sampled frames; extract the facial expressions of each user and classify them into like, ignore, or dislike; count the number of each user's three-category expression classification results and calculate the proportion of each user's three-category expression classification results in the total sample size of that user.

[0300] 5. Obtain the value of SW.

[0301] 6. Calculate the user's approval rating E for the current advertisement using the following formula. n :

[0302]

[0303] In the above formula, n represents the user ID of the current user, E n E represents the rating of user ID n for the currently playing advertisement. n ≥0, and E n The higher the value, the higher the user's approval of the currently playing multimedia. The expression represents the user's attention level to the currently playing advertisement (number n); k1 represents the influence factor of voice feedback on the overall acceptance rating; k2 represents the influence factor of posture feedback on the overall acceptance rating; k3 represents the influence factor of facial expression feedback on the overall acceptance rating; k4 represents the influence factor of attention level on the overall acceptance rating; m1 represents the score of a single keyword in the voice feedback; m2 represents the score of a single posture in the posture feedback; m3 represents the score of attention level; a represents the score of the "like" expression; p 1,nThe percentage of images sampled in every frame for user number n who are categorized as having favorite expressions; b represents the score for ignoring expressions, p 2,n The percentage of images sampled every other frame for user number n who are classified as ignoring facial expressions; c represents the score for the disgust expression, p 3,n The percentage of images sampled every frame for user number n who are classified as having expressions of disgust.

[0304] Step 4: Update the ad playback sequence table

[0305] 1. Obtain the average approval rating of each advertisement in the entire playable advertisement sequence list within an update period. The calculation formula is as follows:

[0306]

[0307] In the above formula, i represents the number of each advertisement in the advertisement playback sequence list.

[0308] 2. Based on each advertisement The scores are sorted from largest to smallest for all played ads within the update period to obtain a ranking table of played ads.

[0309] 3. Obtain the ads that need to be added and their quantity, delete the corresponding number of played ads that rank lower in the rating ranking table from the ad playback sequence table, and add the ads that need to be added to the ad playback sequence table to complete the ad sequence table update process.

[0310] Analyzing the technical solution in this embodiment reveals that it updates the ad playback sequence list in the advertising device based on the user feedback received for each played ad as analyzed in Embodiment 1. The logic of this method is as follows: when an ad update cycle arrives, the system accumulates scores for all user feedback received during each ad playback within that cycle, calculating the average user feedback score for each ad. The higher the average score, the wider the ad's coverage of the target customers; in other words, users frequently appearing in the current ad placement area belong to the target audience of the ad, thus the ad's placement frequency should be increased. Conversely, the placement frequency should be reduced. Based on this conclusion, the method in this embodiment effectively implements this process.

[0311] It should be noted that: the aforementioned Example 5 addresses the problem of sorting the ad playback sequence for different target groups, i.e., the problem of precise ad delivery, while Example 9 addresses the problem of updating and replacing some ads in the ad playback sequence within an ad delivery cycle, i.e., the problem of ad re-running. The two are completely different.

[0312] Example 10

[0313] This embodiment provides a method for evaluating the effectiveness of media advertising in smart business districts, such as... Figure 13 As shown, the evaluation method includes the following steps:

[0314] A method for evaluating the effectiveness of media advertising in smart business districts, the method comprising:

[0315] Step 1: Obtain the feature data of the currently playing advertisement. The feature data includes: the playback duration T of the currently playing advertisement and the keyword dataset associated with the advertisement; the feature data in the keyword dataset consists of multiple pre-defined keywords related to the content of the currently playing advertisement.

[0316] The feature data within the keyword dataset associated with each advertisement includes at least the following:

[0317] (1) Keywords that reflect the advertised products.

[0318] (2) Keywords that reflect the target customer group of the advertisement.

[0319] (3) Keywords that reflect the spokesperson or character in the advertisement.

[0320] (4) High-frequency or special keywords in advertising copy.

[0321] (5) Classification of advertising duration: very short, short, medium, long, and extra long.

[0322] (6) Advertising style classification.

[0323] Step 2: Obtain feedback data from each user regarding ad playback. The feedback data specifically includes: audio stream data generated by the current user group during ad playback, video stream data monitoring the current user group, and instructions from one or more users in the current user group to switch the currently playing ad.

[0324] Determine whether the instruction to switch the currently playing advertisement has been received. If yes, assign the value 1 to the feature quantity SW reflecting the instruction; otherwise, assign the value 0 to SW.

[0325] The methods by which one or more users in the current user group request to switch the currently playing advertisement include key input, voice interaction, and gesture interaction.

[0326] Step 3: Perform data analysis on the audio stream and video stream data. Extract the number of keywords in the audio stream data that match the keywords in the keyword dataset, as well as the number of facial expressions and characteristic actions representing responses to the advertisement. Combine this with the instructions from the current user group to switch the currently playing advertisement, and calculate the approval rating E of each user for the current advertisement. n .

[0327] Each user's rating of the current advertisement (E) n The specific calculation method is as follows:

[0328] (1) Perform speech recognition on the speech stream data, extract the keywords that match the feature data in the keyword dataset associated with the currently playing advertisement, and count their number N1.

[0329] (2) Perform video action recognition on the video stream data; extract the gestures that represent the user's response to the currently playing advertisement, and count their number N2. The gestures that represent the user's response to the currently playing advertisement include: nodding, clapping, pointing the hand at the advertisement playback interface, and head tilting or turning when the head switches from a non-direct gaze state to a direct gaze state.

[0330] (3) Perform video action recognition on the video stream data; extract the feature actions that reflect changes in the current user's eye focus position, and calculate the current user's attention duration t for the currently playing advertisement based on the feature actions. n Characteristic actions reflecting changes in the user's current gaze position include: the user looking directly at the advertisement playback interface, the user closing their eyes, the user looking down, and the user turning their head.

[0331] (4) The video stream data is sampled frame by frame according to the sampling frequency; image recognition is performed on the sampled frames; facial expressions of each user are extracted and classified into like, ignore, or dislike; the number of each user's three-category expression classification results is counted, and the proportion of each user's three-category expression classification results in the total sample size of that user is calculated. The process of extracting and classifying user expressions in the sampled frames is completed by a trained neural network recognition algorithm.

[0332] (5) Get the value of SW.

[0333] (6) Calculate the approval rating E of each user for the current advertisement using the following formula. n :

[0334]

[0335] In the above formula, n represents the user ID of the current user, E n E represents the rating of user ID n for the currently playing advertisement. n ≥0, and E n The higher the value, the higher the user's approval of the currently playing multimedia. The expression represents the user's attention level to the currently playing advertisement (number n); k1 represents the influence factor of voice feedback on the overall acceptance rating; k2 represents the influence factor of posture feedback on the overall acceptance rating; k3 represents the influence factor of facial expression feedback on the overall acceptance rating; k4 represents the influence factor of attention level on the overall acceptance rating; m1 represents the score of a single keyword in the voice feedback; m2 represents the score of a single posture in the posture feedback; m3 represents the score of attention level; a represents the score of the "like" expression; p 1,n The percentage of images sampled in every frame for user number n who are categorized as having favorite expressions; b represents the score for ignoring expressions, p 2,n The percentage of images sampled every other frame for user number n who are classified as ignoring facial expressions; c represents the score for the disgust expression, p 3,n The percentage of images sampled every frame for user number n who are classified as having expressions of disgust.

[0336] Step 4: Obtain the user satisfaction rating (E) for each ad in the sequence of all played ads within an update period. n The sum of these values ​​is used to calculate the average approval rating. The value is used to characterize the user rating of each advertisement within the update cycle.

[0337] The calculation formula is as follows:

[0338]

[0339] In the above formula, i represents the number of each advertisement in the advertisement playback sequence list.

[0340] The method provided in this embodiment can obtain the average approval rating of each advertisement based on user feedback. This result serves as an evaluation indicator of the effectiveness of media advertising in a business district. A higher average approval rating for an advertisement indicates a higher level of attention from the consumer group frequently appearing in that business district, thus suggesting a higher advertising effectiveness. This data can serve as a basis for advertisers and advertising publishers to continue business cooperation or extend the advertising campaign period.

[0341] Example 11

[0342] The methods or systems in the foregoing embodiments can be implemented individually or in conjunction with other embodiments to achieve different technical objectives. For example, this embodiment combines the methods of Embodiment 1 and Embodiment 5 to obtain a smart marketing method based on the Internet of Things and big data; that is, the latter simultaneously includes the implementation processes of the methods in Embodiment 1 and Embodiment 5. Alternatively, the systems of Embodiment 2 and Embodiment 4 can be combined to obtain a smart marketing system based on the Internet of Things and big data; that is, the latter includes at least the modules in Embodiment 2 and Embodiment 4, and simultaneously implements the process of a smart marketing method based on the Internet of Things and big data. Through this method or system, it is possible to update the data in the advertising analysis database while profiling users, and to deliver the most accurate advertisements to different user groups, thereby achieving the purpose of smart marketing.

[0343] Example 12

[0344] The methods or systems in Examples 1-10 can be implemented individually or in conjunction with other examples to achieve different technical objectives. This example combines the methods in Examples 1, 5, and 9 to obtain a garage full-scene marketing method based on consumer feature analysis. This method includes the steps of implementing the methods in Examples 1, 5, and 9. It realizes four aspects: consumer feature profiling, updating the advertising analysis database, accurate advertising push, and updating advertising content. Alternatively, in this example, the hardware devices or systems in Examples 2, 4, and 8 can be combined to obtain a garage full-scene marketing system based on consumer feature analysis. This system includes at least any one of the modules in Examples 2, 4, and 8 and can provide the steps of the garage full-scene marketing method based on consumer feature analysis as described above in real time.

[0345] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for updating advertisements on an advertising placement device, characterized in that, Includes the following steps: Step 1: Obtain the feature data of the currently playing advertisement, including: Obtain the playback duration T of the currently playing advertisement, and the keyword dataset associated with that advertisement; the feature data within the keyword dataset associated with each advertisement includes at least: (1) Keywords reflecting the advertised product; (2) Keywords that reflect the target customer group of the advertisement; (3) Keywords that reflect the spokesperson or character in the advertisement; (4) High-frequency or special keywords in advertising copy; (5) Classification of advertising duration: very short, short, medium, long, and extra long; (6) Classification of advertising styles; Step 2: Obtain feedback data from each user regarding ad playback, including: (1) Acquire audio stream data generated by the current user group during ad playback, monitor video stream data of the current user group, and receive instructions from one or more users in the current user group to switch the currently playing ad; the sources of these instructions include: A switching command issued by a human-computer interaction module to switch the currently playing advertisement; Voice interaction commands issued by the user to switch the currently playing advertisement; A gesture-based interaction command issued by a user, indicating a request to switch the currently playing advertisement; (2) Determine whether an instruction to switch the currently playing advertisement has been received. If yes, assign 1 to the feature quantity SW that reflects the instruction; otherwise, assign 0 to SW. Step 3: Based on the keyword dataset and the audio stream data, video stream data, and SW value in the feedback data, calculate the user's approval rating value E for the current advertisement. n E n The calculation method is as follows; (1) Perform speech recognition on the speech stream data, extract the keywords that match the feature data in the keyword dataset associated with the currently playing advertisement, and count their number N1. (2) Perform video action recognition on the video stream data; extract the gestures that represent the user's response to the currently playing advertisement, and count their number N2. (3) Perform video action recognition on the video stream data; extract the feature actions that reflect changes in the current user's eye focus position, and calculate the current user's attention duration t for the currently playing advertisement based on the feature actions. n ; t n The calculation method is as follows: In the above formula, n represents the user ID of the current user; t 1n This represents the duration of direct eye contact for user number n during the current ad playback; t 2n This represents the duration of time user number n has their eyes closed during the current advertisement playback; t 3n This represents the duration of time user number n looks down during the current advertisement playback; t 4n This indicates the duration of head-turning for user number n during the current ad playback; (4) Perform frame-by-frame sampling on the frame-by-frame images of the video stream data according to the sampling frequency; perform image recognition on the frame-by-frame sampled images; extract the facial expressions of each user and classify the facial expressions as like, ignore or dislike; count the number of the three types of expression classification results for each user and calculate the proportion of the number of the three types of expression classification results for each user in the total sample size of the user. (5) Get the value of SW; (6) Calculate the approval rating E of each user for the current advertisement using the following formula. n : In the above formula, n represents the user ID of the current user, E n E represents the rating of user ID n for the currently playing advertisement. n ≥0, and E n The higher the value, the higher the user's approval of the currently playing multimedia. The expression represents the user's attention level to the currently playing advertisement (number n); k1 represents the influence factor of voice feedback on the overall acceptance rating; k2 represents the influence factor of posture feedback on the overall acceptance rating; k3 represents the influence factor of facial expression feedback on the overall acceptance rating; k4 represents the influence factor of attention level on the overall acceptance rating; m1 represents the score of a single keyword in the voice feedback; m2 represents the score of a single posture in the posture feedback; m3 represents the score of attention level; a represents the score of the "like" expression; p 1,n The percentage of images sampled in every frame for user number n who are categorized as having favorite expressions; b represents the score for ignoring expressions, p 2,n The percentage of images sampled every other frame for user number n who are classified as ignoring facial expressions; c represents the score for the disgust expression, p 3,n The percentage of images sampled every frame for user number n that are classified as showing disgust; Step 4: Update the ad playback sequence table according to the following rules: (1) Obtain the average approval rating of each advertisement in the entire playable advertisement sequence list within an update period. (2) Based on each advertisement Sort all played ads in descending order of value to obtain a ranking table of played ads. (3) Obtain the ads that need to be added and their quantity, delete the corresponding number of played ads that rank lower in the rating ranking table from the ad playback sequence table, and add the ads that need to be added to the ad playback sequence table to complete the update process of the ad sequence table. Step 5: During the advertising reprint process, establish or update an advertising analysis database that profiles the user groups targeted by the advertisements; (1) Set E n A high threshold E h and a low threshold E l E h E represents the threshold at which a user likes the currently playing ad. l E represents the threshold value at which a user dislikes the currently playing advertisement. l >0; (2) When E n ≥E h And p 1，n +p 2，n ≥p 3，n When the current ad is playing, the feature data from the keyword dataset associated with the current ad is added to the current user's corresponding preference tag, and the feature data of the supplemented preference tag is deduplicated; then the feature data in the current user's corresponding aversion tag that is the same as the feature data in the keyword dataset is deleted. (3) When E≤E l And p 2，n +p 3，n ≥p 1，n When the ad is playing, the feature data from the keyword dataset associated with the currently playing ad is added to the aversion tag corresponding to the current user, and the feature data of the added aversion tag is deduplicated; then the feature data in the preference tag corresponding to the current user that matches the feature data in the keyword dataset is deleted. (4) Update the user tags of each user to obtain a new user profile dataset for each user and create an advertising analysis database; Step Six: After obtaining the target profile dataset of the user group in the advertising area, the feature data in the target profile dataset is further compared with the keyword dataset of each advertisement to be played to obtain the degree of overlap between the two, and the advertisement with the higher degree of overlap is placed in the priority placement position.

2. The advertising reprinting method for the advertising placement device as described in claim 1, characterized in that: In step four, the average approval rating results for each advertisement are analyzed. The calculation formula is as follows: In the above formula, i represents the number of each advertisement in the advertisement playback sequence list.

3. The advertising reprinting method for the advertising placement device as described in claim 1, characterized in that: The voice stream data is acquired through a voice acquisition module; when each advertisement is played, the voice acquisition module collects the voice information generated by the user group watching the advertisement; the voice acquisition module includes multiple microphones, which are installed on the advertisement delivery device.

4. The advertising reprinting method for the advertising placement device as described in claim 1, characterized in that: The video stream data is acquired through a video monitoring module; the video monitoring module monitors the user group watching the advertisement from multiple angles during each advertisement; the video monitoring module includes multiple monitoring cameras; the monitoring cameras are installed on the advertisement delivery device.

Images