Intelligent advertisement putting system and method

Abstract

The invention provides an intelligent advertisement putting system and method, and relates to the technical field of advertisement putting. Processing the user face image and the interactive behavior data to obtain an attention state score of the user and a high interested area and a low interested area of the user on the advertisement putting interface; when the attention state score is greater than an upper limit of a preset score range, is within the preset score range and is less than a lower limit of the preset score range, respectively selecting a detailed advertisement template, a standard advertisement template and a simple advertisement template as to-be-displayed templates; and enhancing the high interested area on the to-be-displayed template and weakening the low interested area, and then putting the to-be-displayed template. The method has the beneficial effects that a personalized advertisement template selection mechanism is provided for the user, so that the advertisement putting is more targeted, the information receiving requirements of different users in different states can be better matched, the user is effectively guided to pay attention to key information of the advertisement, and the conversion rate of the advertisement is improved.

Description

Technical Field

[0001] This invention relates to the field of advertising delivery technology, and in particular to an intelligent advertising delivery system and method. Background Technology

[0002] In today's era of booming digital marketing, advertising has become an important means for businesses to promote their products and services and enhance brand awareness. With the rapid development of internet technology, advertising channels have become increasingly diversified, covering many fields such as social media platforms, search engines, and video websites, and users are exposed to a massive amount of advertising information every day.

[0003] However, current advertising delivery models have many problems in practical applications. One prominent issue is the inability to flexibly adjust ad templates based on the user's current state during ad delivery. Specifically, users exhibit significantly different levels of interest in different ads. Some users may only browse quickly, needing only a basic understanding of the ad content; while others may show strong interest in specific products or services, expecting more detailed information. However, existing ad delivery systems struggle to accurately identify the user's current attention and interest level. Regardless of the user's state, the displayed ad template remains fixed, unable to dynamically switch based on real-time user feedback.

[0004] Furthermore, existing ad delivery technologies fail to promptly detect and enhance specific areas of interest in ads, such as highlighting, enlarging, or providing more relevant details. This lack of flexibility and targeting results in a significant gap between ads and user needs, reducing user acceptance and engagement, and ultimately diminishing the effectiveness of ad delivery and failing to maximize the value of advertising in marketing and promotion. Summary of the Invention

[0005] To address the problems existing in the prior art, the present invention provides an intelligent advertising delivery system, comprising:

[0006] The template storage module is used to save pre-generated detailed, standard, and simplified ad templates.

[0007] The status analysis module is used to process real-time environmental data, facial images, and interactive behavior data of the user to obtain the user's attention status score and the user's high interest area and low interest area on the advertising interface.

[0008] The template selection module is connected to the template storage module and the status analysis module respectively. It is used to select the detailed version of the advertisement template as the template to be displayed when the attention status score is greater than the upper limit of the preset score range, select the standard version of the advertisement template as the template to be displayed when the attention status score is within the preset score range, and select the simplified version of the advertisement template as the template to be displayed when the attention status score is less than the lower limit of the preset score range.

[0009] The intelligent delivery module is connected to the status analysis module and the template selection module respectively, and is used to enhance the high interest area and weaken the low interest area on the template to be displayed before delivery.

[0010] Preferably, the environmental data includes the vibration acceleration of the mobile terminal held by the user;

[0011] The status analysis module includes:

[0012] The image processing unit is used to extract and process the image features of the user's facial image to obtain the coordinates of the user's gaze area, gaze time fluctuation, gaze stability, and the user's smile score and frown score on the advertising interface.

[0013] An environmental processing unit is used to process the vibration acceleration to obtain the equipment stability.

[0014] The first calculation unit, connected to the image processing unit, is used to process the user's interest score based on the coordinates of the gaze area and the gaze time fluctuation item.

[0015] The second calculation unit is connected to the image processing unit and the environment processing unit respectively, and is used to calculate the device stability based on the vibration acceleration, and to obtain the user's attention score based on the line of sight stability and the device stability.

[0016] The third calculation unit, connected to the image processing unit, is used to process the user's emotional tendency score based on the smile score and the frown score.

[0017] The comprehensive scoring unit is connected to the first calculation unit, the second calculation unit, and the third calculation unit, respectively, and is used to perform a weighted summation of the interest score, the focus score, and the emotional tendency score to obtain the attention status score.

[0018] Preferably, the image processing unit includes:

[0019] The feature extraction subunit is used to identify the coordinates of the center points of the user's two eyes and the coordinates of the tip of the nose using a facial recognition model;

[0020] The attitude analysis subunit, connected to the feature extraction subunit, is used to establish a reference gaze vector based on the coordinates of the center points of the two eyes and the coordinates of the tip of the nose, and to calculate the horizontal yaw angle and vertical pitch angle of the user's head.

[0021] The coordinate mapping subunit, connected to the attitude analysis subunit, is used to superimpose the horizontal compensation amount generated by the horizontal yaw angle and the vertical compensation amount generated by the vertical pitch angle on the reference line-of-sight vector to obtain the final line-of-sight vector, and to map the final line-of-sight vector to the screen according to the pre-acquired mapping relationship to obtain the coordinates of the line-of-sight region.

[0022] Preferably, the image processing unit further includes:

[0023] The acquisition subunit is used to continuously acquire the gaze duration of multiple gaze events, calculate the standard deviation of each gaze duration, convert the standard deviation into a relative value, and then calculate the gaze duration fluctuation term based on an exponential function.

[0024] Preferably, the image processing unit further includes a scoring subunit connected to the feature extraction subunit, and the feature extraction subunit is further used to identify the left and right corners of the user's mouth feature points and track the center of the eyebrows feature points;

[0025] The scoring subunit is used to calculate the angle between the line connecting the left and right corner feature points and the horizontal plane to obtain the corner curvature of the mouth, and to normalize the corner curvature of the mouth and the longitudinal movement distance of the center of the eyebrows feature point to obtain the smile score and the frown score.

[0026] Preferably, the first computing unit includes:

[0027] The percentage calculation subunit is used to calculate the percentage of the overlapping area between the coordinates of the viewing area and the advertising elements on the advertising interface.

[0028] The first weighted subunit, connected to the proportion calculation subunit, is used to perform a weighted summation of the overlapping area proportion and the gaze time fluctuation item to obtain the interest score.

[0029] Preferably, the interaction behavior data includes the number of times the user touches the screen, and the state analysis module further includes:

[0030] The region of interest analysis unit is used to obtain the user's gaze duration and the number of touches, and to define the coordinates of the gaze region corresponding to the gaze duration being greater than a first threshold and the number of touches being greater than a second threshold as the high region of interest, and the coordinates of the gaze region corresponding to the gaze duration being less than a third threshold and the number of touches not being defined as the low region of interest.

[0031] Preferably, enhancing the highly interesting area includes enlarging the main image on the template to be displayed, adding a breathing light animation to the price tag therein, and fixing the action button at the bottom of the interface;

[0032] The process of weakening the low-interest area includes shrinking the main image on the template to be displayed, adjusting the price tag to have a gray border, and hiding the action button in the secondary menu.

[0033] Preferably, the interaction behavior data also includes the user's active time periods;

[0034] The template selection module includes:

[0035] The threshold adjustment unit is used to lower the upper limit of the preset rating range to a preset value when the user is in the active period.

[0036] The present invention also provides an intelligent advertising delivery method, applied to the above-mentioned intelligent advertising delivery system, wherein the intelligent advertising delivery system stores pre-generated detailed advertising templates, standard advertising templates, and simplified advertising templates;

[0037] The intelligent advertising delivery method includes:

[0038] Step S1: The intelligent advertising delivery system processes real-time environmental data of the user, as well as the user's facial image and interactive behavior data, to obtain the user's attention status score and the user's high interest area and low interest area on the advertising delivery interface.

[0039] Step S2: When the attention status score is greater than the upper limit of the preset score range, the intelligent advertising delivery system selects the detailed version of the advertising template as the template to be displayed; when the attention status score is within the preset score range, it selects the standard version of the advertising template as the template to be displayed; and when the attention status score is less than the lower limit of the preset score range, it selects the simplified version of the advertising template as the template to be displayed.

[0040] Step S3: The intelligent advertising delivery system enhances the high-interest areas and weakens the low-interest areas on the template to be displayed before delivery.

[0041] The above technical solution has the following advantages or beneficial effects:

[0042] 1) By analyzing users' attention status scores, we can provide users with a personalized ad template selection mechanism based on their attention status scores, making ad delivery more targeted and better matching the information reception needs of different users in different states.

[0043] 2) By accurately identifying the areas of high and low interest for users on the ad placement interface, and then enhancing the areas of high interest and weakening the areas of low interest, it is possible to effectively guide users to focus on the key information of the ad and improve the conversion rate of the ad. Attached Figure Description

[0044] Figure 1 A schematic diagram of the structure of an intelligent advertising delivery system is shown in a preferred embodiment of the present invention.

[0045] Figure 2 This is a flowchart illustrating a preferred embodiment of the present invention for a smart advertising delivery method. Detailed Implementation

[0046] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The present invention is not limited to this embodiment; other embodiments that conform to the spirit of the present invention may also fall within the scope of the present invention.

[0047] In a preferred embodiment of the present invention, based on the above-mentioned problems existing in the prior art, an intelligent advertising delivery system is provided, such as... Figure 1 As shown, it includes:

[0048] Template storage module 1 is used to store pre-generated detailed version advertising templates, standard version advertising templates, and simplified version advertising templates;

[0049] The status analysis module 2 is used to obtain the user's attention status score and the user's high interest area and low interest area on the advertising interface based on the real-time acquired environmental data of the user, as well as the user's facial image and interactive behavior data.

[0050] Template selection module 3 is connected to template storage module 1 and status analysis module 2 respectively. It is used to select the detailed version of the advertisement template as the template to be displayed when the attention status score is greater than the upper limit of the preset score range, select the standard version of the advertisement template as the template to be displayed when the attention status score is within the preset score range, and select the concise version of the advertisement template as the template to be displayed when the attention status score is less than the lower limit of the preset score range.

[0051] The intelligent delivery module 4 is connected to the status analysis module 2 and the template selection module 3 respectively. It is used to enhance the high interest areas and weaken the low interest areas on the template to be displayed before delivery.

[0052] Specifically, in this embodiment, it is preferable to acquire the user's facial image through the front-facing camera of the advertising delivery device and acquire the user's environmental data through the environmental perception sensor integrated into the advertising delivery device. The status analysis module 2 is deployed on an edge server or in the cloud to reduce the system's computing power requirements on the advertising delivery device.

[0053] Furthermore, when applying this system to mobile terminals, energy efficiency can be optimized in the following two ways:

[0054] The first method: data sampling strategy

[0055] When the screen of the advertising device is lit, full parameter sampling is performed, such as a sampling frequency of 1Hz;

[0056] After detecting a face, the sampling frequency can be increased to 3Hz;

[0057] During the ad display, a sampling frequency of 10Hz can be used to sample key parameters, where the key parameter is the user's facial image.

[0058] The second type: Dynamic power management

[0059] When the screen of the advertising device is lit, it first detects whether a user is approaching the advertising device. If an approach is detected, other sensors are activated. If the user does not interact within 30 seconds after activating other sensors, the screen of the advertising device is turned off.

[0060] More preferably, when the advertising device integrates a light acquisition sensor, to improve the user viewing experience, when the real-time ambient light level G collected by the light acquisition sensor is too low, the screen can be triggered to supplement the light, such as briefly increasing the brightness by 10%, or enabling the pre-configured CNN low-light enhancement model (MobileNet-v3) for facial recognition. Alternatively, when the screen of the advertising device is an OLED screen, based on the fact that OLED screens can emit near-infrared light waves and can still actively emit invisible infrared light when displaying a black screen, and that the infrared reflection characteristics are not affected by ambient visible light, the screen can also be switched to an infrared feature analysis mode, that is, the screen switches to an infrared emission mode, and the camera exposure is synchronized to obtain an infrared video stream for analysis.

[0061] The above-mentioned detailed version of the ad template preferably includes a complete product description and multiple images, the above-mentioned standard version of the ad template preferably includes basic text and image information and core selling points, and the above-mentioned simplified version of the ad template preferably includes only the main image, product price and minimize button.

[0062] Furthermore, when analyzing and obtaining the user's attention status score, a pre-stored ad template can be selected as the template to be displayed based on the attention status score and a preset score range. For example, with a preset score range of [0.4, 0.6]:

[0063] 1. When the attention status score is greater than 0.6, select the detailed version of the ad template as the template to be displayed, so as to provide users with richer and more detailed information and meet their needs to understand the ad content in depth.

[0064] For example, in the detailed version of the ad template, the main image can be enlarged to 115% and a floating effect can be added, and the "Buy Now" button can be fixed at the bottom and highlighted.

[0065] 2. When the attention status score is within [0.4, 0.6], select the standard version of the ad template as the template to be displayed to balance the level of detail and simplicity of the information display;

[0066] For example, the standard version of the ad template can keep the main image at its original size and display the action button in a normal style.

[0067] 3. When the attention status score is less than 0.4, select the simplified ad template as the template to be displayed to avoid showing users too much redundant information and improve user experience.

[0068] For example, the simplified ad template can hide auxiliary images and text and reduce the price tag to 90%.

[0069] As a preferred implementation, a higher evaluation threshold, such as 0.8, can be set so that when the attention status score is greater than the evaluation threshold of 0.8, a high-definition image is loaded for displaying the advertising product, and when the attention status score is between 0.6 and 0.8, a compressed image is loaded for displaying the advertising product.

[0070] This personalized template selection mechanism combines predefined templates with real-time parameter adjustments, ensuring consistency in advertising content while achieving precise personalized presentation. This makes advertising more targeted, better matching the information reception needs of different users in different states, effectively improving advertising conversion rates, reducing the production cost of advertising templates, and upgrading advertising from "personalized to personalized" to "personalized to personalized at different times and states," significantly improving advertising effectiveness and user experience.

[0071] Furthermore, after selecting a template to be displayed, before the display takes place, the template can be dynamically adjusted.

[0072] Specifically, for areas of high interest, the system will enhance them, such as enlarging the main image on the template to be displayed, highlighting the core content, and guiding users to further understand the product; adding a breathing light animation to the price tag to attract users' attention; and fixing the action button at the bottom of the interface for easy user operation.

[0073] For areas of low interest, the system will weaken the display, such as shrinking the main image to reduce its interference with the user; adjusting the price tag to a gray border to reduce its visual prominence; and hiding the action button in a secondary menu, making the ad interface more concise and clear, and avoiding distracting the user from the key content. This precise strategy of strengthening and weakening different areas can effectively guide users to focus on the key information in the ad, thereby improving the conversion rate.

[0074] Furthermore, regarding the state analysis module 2, it includes four core processing steps: user facial image recognition, environmental data processing, attention state score calculation, and high and low interest region localization. These will be analyzed in detail below.

[0075] 1. User facial image recognition

[0076] Status analysis module 2 includes:

[0077] The image processing unit 21 is used to extract and process image features from the user's facial image to obtain the coordinates of the user's gaze area, gaze time fluctuation, gaze stability, and the user's smile score and frown score on the advertising interface.

[0078] The image processing unit 21 includes:

[0079] The feature extraction subunit 211 is used to identify the coordinates of the center points of the user's two eyes and the coordinates of the tip of the nose using a facial recognition model;

[0080] The attitude analysis subunit 212 is connected to the feature extraction subunit 211 and is used to establish a reference line of sight vector based on the coordinates of the center points of the two eyes and the coordinates of the tip of the nose, and to calculate the horizontal yaw angle and vertical pitch angle of the user's head.

[0081] The coordinate mapping subunit 213 is connected to the attitude analysis subunit 212. It is used to superimpose the horizontal compensation amount generated by the horizontal yaw angle and the vertical compensation amount generated by the vertical pitch angle on the reference line-of-sight vector to obtain the final line-of-sight vector, and to map the final line-of-sight vector to the screen according to the pre-acquired mapping relationship to obtain the coordinates of the line-of-sight area.

[0082] Specifically, in this embodiment, the facial recognition model includes, but is not limited to, the lightweight EfficientNet-B0 model or the MediaPipe Face Mesh model, which are used to detect multiple facial feature points on the user's facial image in real time, including but not limited to the inner corner of the left eye, the inner corner of the right eye, and the tip of the nose.

[0083] Considering that the anatomical positions of the user's nose tip and corner of the eye are fixed, but the coordinate relationship of the image captured by the device's camera changes with head rotation, this embodiment achieves real-time gaze estimation by establishing a "head posture-gaze direction" mapping model. Specifically, this includes:

[0084] The coordinates of the center points of both eyes (center point x, center point y) are calculated based on the coordinates of the inner corners of the left and right eyes (left eye x, left eye y), where:

[0085] Center point x = (left eye x + right eye x) / 2

[0086] Center point y = (right eye x + right eye y) / 2

[0087] Subsequently, a reference gaze vector is established based on the coordinates of the nose tip (nose tip x, nose tip y) and the coordinates of the center points of the two eyes (center point x, center point y). The x-component of the reference gaze vector is equal to the nose tip x minus the center point x, and the y-component of the reference gaze vector is equal to the nose tip y minus the center point y.

[0088] It can also analyze the user's three-dimensional head posture based on the nose tip coordinates (nose tip x, nose tip y) and the coordinates of the center points of the two eyes (center point x, center point y). The three-dimensional head posture includes the horizontal yaw angle, which reflects the degree of left and right rotation of the user's head, and the vertical pitch angle, which reflects the degree of up and down nodding of the user's head.

[0089] Where, horizontal yaw angle = arctan(nose tip x - center point x) * 180 / π

[0090] Vertical pitch angle = arctan(center point y - nose tip y) * 180 / π

[0091] Since the horizontal deviation of the line of sight is proportional to the horizontal yaw angle when the head turns left and right, and the vertical deviation of the line of sight is proportional to the vertical pitch angle when the head nods up and down, the horizontal compensation and vertical compensation are calculated based on the horizontal yaw angle and the vertical pitch angle, respectively.

[0092] Horizontal compensation = Calibration coefficient K1 * Horizontal yaw angle

[0093] Vertical compensation = Calibration coefficient K2 * Vertical pitch angle

[0094] Final line-of-sight vector = Baseline line-of-sight vector + Horizontal compensation amount + Vertical compensation amount

[0095] Wherein, the calibration coefficient K1 represents the horizontal pixel offset caused by each degree of horizontal yaw angle, and the calibration coefficient K2 represents the vertical pixel offset caused by each degree of vertical pitch angle, which can be obtained through pre-calibration. Preferably, the calibration process can be triggered, and five calibration points can be displayed on the ad delivery interface (such as the four corners and the center of the ad delivery interface). Then, the user is guided to look at and click on each calibration point in sequence, and the horizontal yaw angle, vertical pitch angle, and reference line-of-sight vector are recorded accordingly when the user clicks on each calibration point.

[0096] For each calibration point, the following system of equations is established based on the screen coordinate mapping relationship:

[0097] screen_x=0.5+(base_x+K1*yaw+K2*pitch) / (2×D)

[0098] screen_y=0.3+(base_y+K1*yaw+K2*pitch) / (3×D)

[0099] Where (screen_x, screen_y) represents the coordinates of the line of sight on the screen, (base_x, base_y) are the x and y components of the base line of sight vector, yaw represents the horizontal compensation amount, pitch represents the vertical compensation amount, and D represents the virtual distance from the center point of the two eyes to the screen, which can be estimated by the distance between the two eyes.

[0100] The above 5 calibration points can construct 10 equations with only 2 unknowns K1 and K2. Based on this, the least squares method can be used to solve the equations and find K1 and K2 that minimize the sum of squared prediction errors. The equation solving process is existing technology and is not the inventive point of this invention, so it will not be described in detail here.

[0101] Furthermore, the aforementioned final gaze vector, used to represent the direction of the gaze, needs to be mapped onto the screen based on the mapping relationship between the two to obtain the coordinates of the gaze area. This mapping relationship is the aforementioned set of equations. This calculation method can more accurately determine the user's gaze point on the advertising interface, providing precise location information for analyzing the user's area of ​​interest.

[0102] In a preferred embodiment of the present invention, the image processing unit 21 further includes:

[0103] The acquisition subunit 214 is used to continuously acquire the gaze duration of multiple gaze events, calculate the standard deviation of each gaze duration, convert the standard deviation into a relative value, and then calculate the gaze duration fluctuation term based on the exponential function.

[0104] Specifically, in this embodiment, the user's continuous gaze data is first acquired, such as the user's continuous gaze data over the past 3 seconds, as follows:

[0105] gaze_durations = [420, 380, 510, 290, 680, 310], in milliseconds. It can be seen that 6 gaze events were detected in the past 3 seconds.

[0106] Based on the above 6 fixation events, the standard deviation of each fixation duration can be calculated to be 130, and the average value is 432. The ratio of the standard deviation to the average value is calculated as the relative value. In this embodiment, the relative value is approximately 0.3.

[0107] Considering that when user A's gaze duration is [500, 520, 510] ms, the relative value is approximately 0.02; and when user B's gaze duration is [200, 1000, 300] ms, the relative value is approximately 0.87, directly incorporating these values ​​into subsequent weighted calculations could lead to uncontrolled numerical ranges (a difference of 43 times). Furthermore, since interest is positively correlated with stability when human gaze fluctuations are relatively stable, and approaches zero when highly unstable, an exponential function better fits this non-linear relationship than a linear relative value. This allows the calculated index to reflect the stability of users' gaze duration while viewing advertisements, helping the system more accurately assess users' interest levels and providing an important basis for calculating interest scores.

[0108] Based on this, the following exponential function exp(-relative value / a) is introduced to calculate the fixation event fluctuation term, compressing the output to between 0 and 1, and making it insensitive to outliers. Here, 'a' is an adjustment coefficient used to control the decay rate, which can be adjusted according to the application scenario. For example, for regular advertising, 'a' can be set to 0.2; for high-value advertising, 'a' can be set to 0.16 to specify stricter stability requirements; and for brand image advertising, 'a' can be relaxed to 0.25. It is understood that the above adjustment rules for the adjustment coefficient are only examples and are not intended to limit the invention.

[0109] More preferably, after obtaining the multiple gaze events contained in the continuous gaze data, the method also includes data validity filtering to remove abnormal gaze data and obtain valid gaze duration, which is then used in subsequent calculations to improve the scoring accuracy.

[0110] Abnormal gaze data includes:

[0111] Abnormally short fixations (e.g., fixation duration <200ms) may indicate unconscious saccades by the user.

[0112] An unusually long gaze (e.g., gaze duration > 200ms) may indicate that the user has temporarily left the room.

[0113] In a preferred embodiment of the present invention, the image processing unit 21 further includes a scoring subunit 215 connected to the feature extraction subunit 211. The feature extraction subunit 211 is also used to identify the feature points of the user's left and right corners of the mouth and track the feature points of the center of the eyebrows.

[0114] The scoring subunit 215 is used to calculate the angle between the line connecting the feature points of the left and right corners of the mouth and the horizontal plane to obtain the curvature of the corners of the mouth. The curvature of the corners of the mouth and the longitudinal movement distance of the feature point between the eyebrows are normalized to obtain the smile score and the frown score.

[0115] Specifically, in this embodiment, the smile score = min(1, corner radius of mouth / 15°), and the frown score = min(1, vertical movement distance / 3mm). These scores can reflect the user's emotional tendencies, providing emotional dimension information for a comprehensive assessment of the user's attention status, making advertising more aligned with the user's emotional state.

[0116] 2. Environmental data processing

[0117] In a preferred embodiment of the present invention, the environmental data includes the vibration acceleration of the mobile terminal held by the user;

[0118] Status analysis module 2 also includes:

[0119] The environmental processing unit 22 is used to obtain the equipment stability based on the vibration acceleration.

[0120] Specifically, in this embodiment, the acceleration modulus is first calculated based on the collected vibration acceleration, then normalized, and finally the equipment stability is obtained by subtracting the normalized acceleration modulus from 1.

[0121] 3. Calculation of Attention Status Score

[0122] Status analysis module 2 also includes:

[0123] The first calculation unit 23 is connected to the image processing unit 21 and is used to obtain the user's interest score based on the coordinates of the gaze area and the gaze time fluctuation item.

[0124] This calculation method takes into account both the duration of the user's gaze on the advertising element and the stability of the gaze time, and can more scientifically assess the user's level of interest in the advertising element.

[0125] The second calculation unit 24 is connected to the image processing unit 21 and the environment processing unit 22 respectively. It is used to calculate the device stability based on the vibration acceleration and to obtain the user's attention score based on the line of sight stability and the device stability.

[0126] The third calculation unit 25 is connected to the image processing unit 21 and is used to process the user's emotional tendency score based on the smile score and frown score.

[0127] The comprehensive scoring unit 26 is connected to the first calculation unit 23, the second calculation unit 24 and the third calculation unit 25 respectively, and is used to perform a weighted summation of interest score, focus score and emotional tendency score to obtain attention status score.

[0128] This multi-dimensional data fusion analysis enables the system to understand users' status and needs more comprehensively and accurately, providing reliable data support for the selection of advertising templates and the regional reinforcement and weakening.

[0129] In a preferred embodiment of the present invention, the first computing unit 23 includes:

[0130] The percentage calculation subunit 231 is used to calculate the percentage of the overlapping area between the coordinates of the viewing area and the advertising elements on the advertising interface, that is, the intersection area of ​​the gaze point and the advertising elements divided by the total advertising area.

[0131] The first weighted subunit 232 connects to the proportion calculation subunit 231, which is used to perform a weighted summation of the overlap area proportion and the gaze time fluctuation term to obtain the interest score.

[0132] Specifically, in this embodiment, the attention score is obtained by weighted summation of gaze stability and device stability, where gaze stability is the duration of a single gaze event. The emotion tendency score is calculated as 0.5 + first weight * smile score + second weight * frown score, where the first weight is positive, the second weight is negative, and 0.5 limits the emotion tendency score to the range of [0,1].

[0133] 4. Location of regions of high and low interest

[0134] In a preferred embodiment of the present invention, the interaction behavior data includes the number of times the user touches the screen, and the state analysis module 2 further includes:

[0135] The region of interest analysis unit 27 is used to obtain the user's gaze duration and touch count, and to define the coordinates of the gaze region corresponding to a gaze duration greater than a first threshold and a touch count greater than a second threshold as the region of high interest, and the coordinates of the gaze region corresponding to a gaze duration less than a third threshold and no touch count as the region of low interest.

[0136] Specifically, in this embodiment, the number of touches is preferably the number of clicks per unit area. By identifying regions of high interest and regions of low interest, it is possible to accurately determine the degree of user attention to different areas of the advertisement, providing clear target areas for subsequent region strengthening and weakening.

[0137] In a preferred embodiment of the present invention, the interaction behavior data also includes the user's active time period;

[0138] Template selection module 3 includes:

[0139] The threshold adjustment unit 31 is used to lower the upper limit of the preset rating range to a preset value when the user is active.

[0140] Specifically, in this embodiment, the user's historical interaction behavior data is first collected, including but not limited to user behavior type, interaction coordinates, operation duration and corresponding timestamp, wherein user behavior type includes but is not limited to clicking, swiping and hovering.

[0141] The day is then divided into multiple time periods, either by hour (24 periods) or by 15 minutes (95 periods), with no specific limitation here. Historical interaction data is then integrated based on timestamps to form an interaction sequence.

[0142] Active time period prediction is performed based on a pre-trained bidirectional LSTM model that takes the sequence of interactive behaviors as input and the probability of activity in each time period as output. In other words, it predicts whether the user is currently in an active time period, or whether the current time period is the user's active time period.

[0143] If the corresponding activity probability is greater than the preset activity threshold, the user is considered to be in an active period. In this case, it is preferable to lower the upper limit of the preset rating range to enhance the intensity of ad delivery. For example, if the original upper limit of the preset rating range is 0.6, it can be adjusted to 0.4, making it easier to trigger the selection of the detailed ad template, thus meeting the user's need to receive more detailed information when active. This mechanism of adjusting the threshold based on the user's active period allows ad delivery to better adapt to changes in the user's status at different times, improving the accuracy and effectiveness of ad delivery.

[0144] To reduce the load on the terminal, TCN (Temporal Convolutional Network) can be used instead of the bidirectional LSTM model. Compared with the bidirectional LSTM model, it has a lower computational cost and is more suitable for continuous background operation on mobile terminals.

[0145] Furthermore, when users are active, in addition to lowering the upper limit of the preset rating range, the main image in the detailed ad template can be enlarged by 10%, and the animation intensity in the detailed ad template can also be enhanced.

[0146] As a preferred embodiment, the advertising device may also be equipped with an ambient light sensor to collect the illumination value of the user's current environment in real time. When the illumination value is greater than a preset illumination threshold, the advertising template can be displayed in a high-contrast bright color scheme, and when the illumination value is not greater than the preset illumination threshold, the advertising template can be displayed in a dark mode.

[0147] The present invention also provides an intelligent advertising delivery method, which is applied to the above-mentioned intelligent advertising delivery system. The intelligent advertising delivery system stores pre-generated detailed advertising templates, standard advertising templates, and simplified advertising templates.

[0148] like Figure 2 As shown, intelligent ad delivery methods include:

[0149] Step S1: The intelligent advertising delivery system processes real-time environmental data, facial images, and interactive behavior data to obtain a user attention status score and the user's high and low interest areas on the advertising delivery interface.

[0150] Step S2: When the attention status score is greater than the upper limit of the preset score range, the intelligent advertising delivery system selects the detailed version of the advertising template as the template to be displayed; when the attention status score is within the preset score range, it selects the standard version of the advertising template as the template to be displayed; and when the attention status score is less than the lower limit of the preset score range, it selects the concise version of the advertising template as the template to be displayed.

[0151] Step S3: The intelligent advertising delivery system enhances the high-interest areas and weakens the low-interest areas on the template to be displayed before delivery.

[0152] The above description is merely a preferred embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made using the content of this specification and illustrations should be included within the protection scope of the present invention.

Claims

1. An intelligent advertising delivery system, characterized in that, include: The template storage module is used to save pre-generated detailed, standard, and simplified ad templates. The status analysis module is used to process real-time environmental data, facial images, and interactive behavior data of the user to obtain the user's attention status score and the user's high interest area and low interest area on the advertising interface. The template selection module is connected to the template storage module and the status analysis module respectively. It is used to select the detailed version of the advertisement template as the template to be displayed when the attention status score is greater than the upper limit of the preset score range, select the standard version of the advertisement template as the template to be displayed when the attention status score is within the preset score range, and select the simplified version of the advertisement template as the template to be displayed when the attention status score is less than the lower limit of the preset score range. The intelligent delivery module is connected to the status analysis module and the template selection module respectively, and is used to enhance the high interest area and weaken the low interest area on the template to be displayed before delivery.

2. The intelligent advertising delivery system according to claim 1, characterized in that, The environmental data includes the vibration acceleration of the mobile terminal held by the user; The status analysis module includes: The image processing unit is used to extract and process the image features of the user's facial image to obtain the coordinates of the user's gaze area, gaze time fluctuation, gaze stability, and the user's smile score and frown score on the advertising interface. An environmental processing unit is used to process the vibration acceleration to obtain the equipment stability. The first calculation unit, connected to the image processing unit, is used to process the user's interest score based on the coordinates of the gaze area and the gaze time fluctuation item. The second calculation unit is connected to the image processing unit and the environment processing unit respectively, and is used to calculate the device stability based on the vibration acceleration, and to obtain the user's attention score based on the line of sight stability and the device stability. The third calculation unit, connected to the image processing unit, is used to process the user's emotional tendency score based on the smile score and the frown score. The comprehensive scoring unit is connected to the first calculation unit, the second calculation unit, and the third calculation unit, respectively, and is used to perform a weighted summation of the interest score, the focus score, and the emotional tendency score to obtain the attention status score.

3. The intelligent advertising delivery system according to claim 2, characterized in that, The image processing unit includes: The feature extraction subunit is used to identify the coordinates of the center points of the user's two eyes and the coordinates of the tip of the nose using a facial recognition model; The attitude analysis subunit, connected to the feature extraction subunit, is used to establish a reference gaze vector based on the coordinates of the center points of the two eyes and the coordinates of the tip of the nose, and to calculate the horizontal yaw angle and vertical pitch angle of the user's head. The coordinate mapping subunit, connected to the attitude analysis subunit, is used to superimpose the horizontal compensation amount generated by the horizontal yaw angle and the vertical compensation amount generated by the vertical pitch angle on the reference line-of-sight vector to obtain the final line-of-sight vector, and to map the final line-of-sight vector to the screen according to the pre-acquired mapping relationship to obtain the coordinates of the line-of-sight region.

4. The intelligent advertising delivery system according to claim 2, characterized in that, The image processing unit further includes: The acquisition subunit is used to continuously acquire the gaze duration of multiple gaze events, calculate the standard deviation of each gaze duration, convert the standard deviation into a relative value, and then calculate the gaze duration fluctuation term based on an exponential function.

5. The intelligent advertising delivery system according to claim 3, characterized in that, The image processing unit further includes a scoring subunit, which is connected to the feature extraction subunit. The feature extraction subunit is also used to identify the feature points of the user's left and right corners of the mouth and track the feature points of the center of the eyebrows. The scoring subunit is used to calculate the angle between the line connecting the left and right corner feature points and the horizontal plane to obtain the corner curvature of the mouth, and to normalize the corner curvature of the mouth and the longitudinal movement distance of the center of the eyebrows feature point to obtain the smile score and the frown score.

6. The intelligent advertising delivery system according to claim 2, characterized in that, The first computing unit includes: The percentage calculation subunit is used to calculate the percentage of the overlapping area between the coordinates of the viewing area and the advertising elements on the advertising interface. The first weighted subunit, connected to the proportion calculation subunit, is used to perform a weighted summation of the overlapping area proportion and the gaze time fluctuation item to obtain the interest score.

7. The intelligent advertising delivery system according to claim 1, characterized in that, The interaction behavior data includes the number of times the user touches the screen, and the state analysis module further includes: The region of interest analysis unit is used to obtain the user's gaze duration and the number of touches, and to define the coordinates of the gaze region corresponding to the gaze duration being greater than a first threshold and the number of touches being greater than a second threshold as the high region of interest, and the coordinates of the gaze region corresponding to the gaze duration being less than a third threshold and the number of touches not being defined as the low region of interest.

8. The intelligent advertising delivery system according to claim 1, characterized in that, The enhancement of the highly interesting area includes enlarging the main image on the template to be displayed, adding a breathing light animation to the price tag therein, and fixing the action button to the bottom of the interface; The process of weakening the low-interest area includes shrinking the main image on the template to be displayed, adjusting the price tag to have a gray border, and hiding the action button in the secondary menu.

9. The intelligent advertising delivery system according to claim 1, characterized in that, The interaction behavior data also includes the user's active time periods; The template selection module includes: The threshold adjustment unit is used to lower the upper limit of the preset rating range to a preset value when the user is in the active period.

10. A method for intelligent advertising delivery, characterized in that, Applied to the intelligent advertising delivery system as described in any one of claims 1-9, wherein the intelligent advertising delivery system stores pre-generated detailed version advertising templates, standard version advertising templates, and simplified version advertising templates; The intelligent advertising delivery method includes: Step S1: The intelligent advertising delivery system processes real-time environmental data of the user, as well as the user's facial image and interactive behavior data, to obtain the user's attention status score and the user's high interest area and low interest area on the advertising delivery interface. Step S2: When the attention status score is greater than the upper limit of the preset score range, the intelligent advertising delivery system selects the detailed version of the advertising template as the template to be displayed; when the attention status score is within the preset score range, it selects the standard version of the advertising template as the template to be displayed; and when the attention status score is less than the lower limit of the preset score range, it selects the simplified version of the advertising template as the template to be displayed. Step S3: The intelligent advertising delivery system enhances the high-interest areas and weakens the low-interest areas on the template to be displayed before delivery.

Images