A commodity display system based on AR technology

Abstract

The present invention relates to the field of commodity display technology, and discloses a commodity display system based on AR technology, including a dynamic environment adaptation module: automatically adjusting the display effect of commodities in different environments; a display background adjustment module: dynamically changing the display background of commodities according to different influencing factors; a user behavior analysis and commodity recommendation module: providing commodity recommendations to users by analyzing user behavior status, and displaying them in the commodity display system. The commodity display system based on AR technology proposed by the present invention can help users adjust their emotions by adjusting the background to adapt to the user's emotional state, thereby improving the user's satisfaction and comfort. Secondly, considering the user's preferences and emotions, it can provide a more personalized experience, making the user feel valued and understood, thereby increasing the effectiveness and conversion rate of commodity display, thereby bringing higher commercial value to the enterprise.

Description

Technical Field

[0001] The present invention relates to the technical field of commodity display, and in particular to a commodity display system based on AR technology. Background Art

[0002] The product display system based on AR technology is a technical application that uses digital technology to superimpose virtual information into the real-world environment to enhance the user experience. Such systems are usually used in retail, e-commerce, exhibitions and other fields to provide richer, more interactive and immersive product display methods. These product display systems based on AR technology not only enhance consumers' shopping experience, but also provide companies with new marketing methods and technical support.

[0003] Existing product display systems based on augmented reality (AR) technology are often designed to rely on static or semi-static background display modes. Specifically, such systems tend to use fixed background images or a set of pre-set background images for rotation. Although this display method can provide certain visual effects, it has obvious limitations, which are mainly reflected in the following aspects:

[0004] First, the single background setting makes the product display lack of change and freshness. When users visit the same product multiple times, if they always see the same background, they will gradually get bored, which reduces the user's attention and interest in the product. In today's era of information explosion, it is becoming increasingly difficult to attract users' attention. Therefore, this single display method is difficult to stand out among many competitors;

[0005] Secondly, this fixed background display method cannot be customized according to the personalized needs of users. Different users have different needs for product display. For example, some users may prefer bright and lively backgrounds, while others may prefer simple backgrounds. Since the current technology lacks sufficient flexibility to adapt to these personalized needs, the user experience is greatly reduced.

[0006] Furthermore, this method limits the diversity and creative space of display. The lack of interactive and participatory display mode will weaken the user's sense of participation. Modern consumers pay more and more attention to the experience and personal participation in the shopping process, and the single-background display method obviously cannot meet this demand, which will affect the user retention rate and the attention of the product, and thus have an adverse impact on the sales conversion rate;

[0007] Finally, for users who want to explore new experiences, especially young people and elderly users who pursue personalized services, this lack of interactivity and novelty in the display method may reduce their overall evaluation and loyalty to the product, limiting the potential value of this important consumer group;

[0008] In summary, the limitations of the existing AR-based product display system in terms of background display not only limit the diversity of user experience, but also affect users' interest and participation in the product, which may have an adverse impact on the sales conversion rate of the product. Therefore, it is necessary to design a product display system based on AR technology to solve the above problems. Summary of the invention

[0009] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a commodity display system based on AR technology.

[0010] In order to achieve the above object, the present invention adopts the following technical solutions:

[0011] A commodity display system based on AR technology includes a dynamic environment adaptation module: automatically adjusting the display effect of commodities in different environments;

[0012] Display background adjustment module: based on the background adjustment parameters, updates the product display background, and the background adjustment parameters are controlled by the background adjustment parameter model;

[0013] User behavior analysis and product recommendation module: By analyzing user behavior status, it provides product recommendations to users and displays them in the product display system;

[0014] Background library: Several product display backgrounds are set in the background library.

[0015] As a preferred technical solution of the present invention, the background adjustment parameter model is:

[0016] S=∑ i w i f i (x i );

[0017] Among them, i represents different influencing factors, w i is the weight of influencing factor i, f i (x i ) is the mapping function of influencing factor i, x i Indicates the computable parameters corresponding to different influencing factors;

[0018] S is a background update standard calculated based on relevant influencing factors. According to the calculated value S, the suitable background in the background library is found and replaced.

[0019] As a preferred technical solution of the present invention, the step of finding an adapted background in a background library and replacing it according to the calculated value S comprises:

[0020] Each background in the background library corresponds to a set of initial scores S1. The background with the smallest value obtained by subtracting the score S from the initial score S1 is used as the adapted background. When there are several display backgrounds with the same difference, the front one in the background library is selected by default.

[0021] As a preferred technical solution of the present invention, the user emotion influencing factor i E The mapping function is:

[0022] f E (x E )= 1-(1-x E ) β , x E ≤α;

[0023] f E (x E )=log(1+x E ), x E >α;

[0024] Among them, x E Factors affecting user emotions E , α is used to distinguish the emotional influencing factors x E The thresholds for different regions, β is the adjustment coefficient.

[0025] As a preferred technical solution of the present invention, the user emotion influencing factor i E The corresponding computable parameter x E Subject to the constraints of the user emotion parameter detection model, the user emotion parameter detection model is:

[0026] x E =w F x F +w V x V +w H x H ;

[0027] In the formula, x F 、x V and x H Respectively represent the detection values ​​of facial expressions, voice intonation and physiological signals, w F 、w V and w H They represent the weights of the detection values, and w F +w V +w H =1.

[0028] As a preferred technical solution of the present invention, the display background adjustment module further includes a feedback adjustment unit;

[0029] Set time t, after the system changes the display background, after t time, keep other f i (x i ) remains unchanged, and the user sentiment value is recalculated;

[0030] Set a change Δx E , Δx E =x Et -x E , where x Et is the user sentiment score at the end of time t, x E is the user sentiment score at the beginning of time t, perform the following steps:

[0031] In x E >α, if Δx E ≥0, keep the current background. If Δx E <0, then change the current background;

[0032] In x E ≤α, if Δx E >0, keep the current background. If Δx E ≤0, the current background will be changed.

[0033] As a preferred technical solution of the present invention, the feedback adjustment unit further includes a weight adjustment model, and the weight adjustment model is:

[0034] ,

[0035] In the formula, w i ′ is the adjusted weight of influencing factor i, η i is the learning rate, F i For weight w i The expected value of , J is the loss function, is the loss function J with respect to weight w i The gradient of , R is the reward, γ is the discount factor, Q is the state-action value function, Q(s, a) represents the Q function value of taking action a in the current state s, max a′ Q(s′, a′) represents the maximum value of the Q function among all possible actions a′ in the next state s′.

[0036] As a preferred technical solution of the present invention, seasonal factors i T The mapping function is:

[0037] f T (x T )=x T , 0 <x T ≤δ1;

[0038] f T (x T )∈[δ3-δ4],δ1<x T ≤δ2;

[0039] f T (x T )∈[0-δ1],δ2<x T ≤δ3;

[0040] f T (x T )∈[δ1-δ2],δ3<x T ≤δ4;

[0041] Among them, x T Factors affecting user emotions E The computable parameters, 0-δ1 represents the spring value range, δ1-δ2 represents the summer value range, δ2-δ3 represents the autumn value range, and δ3-δ4 represents the winter value range.

[0042] As a preferred technical solution of the present invention, the seasonal influencing factor i T The mapping function also includes:

[0043] f T (x T ) = δ3 + (δ2-x T ) ,δ1<x T ≤δ2;

[0044] f T (x T ) = (δ3-x T ) ,δ2<x T ≤δ3;

[0045] f T (x T ) = δ1 + (δ4-x T ) ,δ3<x T ≤δ4.

[0046] As a preferred technical solution of the present invention, it also includes:

[0047] Local cultural factors C The mapping relationship is:

[0048] f C (x C ) = γ;

[0049] Among them, x CLocal cultural factors C The computable parameter of , the value γ indicates that the product display context is moderately related to the current culture;

[0050] User preference factor P The mapping relationship is:

[0051] f P (x P ) = ε;

[0052] Among them, x P is the user preference factor i P A computable parameter of , with a value of ε representing a moderate fit with user preferences.

[0053] The present invention has the following beneficial effects:

[0054] 1. Improve user experience: By adjusting the background to suit the user's emotional state, it can help users regulate their emotions, thereby improving their satisfaction and comfort. Secondly, considering the user's preferences and emotions, it can provide a more personalized experience, making the user feel valued and understood;

[0055] 2. Enhance user engagement: By dynamically adjusting the background, users’ attention can be attracted and their interest in the product can be increased. Users may interact more with the product display due to the change of background, such as commenting and sharing. The system allows users to adjust the weight of various factors, so the system can adapt to the needs of different users more flexibly.

[0056] 3. Deepen memory and increase attention: Dynamically adjusting the product display background through different factors can bring freshness to users and help deepen their memory of the product. The unique display background can attract more users' attention and increase their attention to the product.

[0057] 4. Improve conversion rate: By providing a context that matches the user's emotions and preferences, the user's willingness to buy can be increased, and the user's trust in the brand will increase because the system can accurately capture and respond to their emotions and preferences. This can increase the effectiveness and conversion rate of product display by improving user satisfaction and engagement, thereby bringing higher commercial value to the company. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is an architecture diagram of a commodity display system based on AR technology proposed by the present invention;

[0059] Figure 2 To show the architecture diagram of the background adjustment module;

[0060] Figure 3 A flowchart showing the background adjustment module. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0062] Reference Figure 1 , a commodity display system based on AR technology, comprising:

[0063] Dynamic environment adaptation module: responsible for automatically adjusting the display effect of goods in different environments. The module includes an ambient light detection and adjustment unit, a sound adjustment unit and a product size adjustment unit. The ambient light detection and adjustment unit measures the light intensity of the surrounding environment through an ambient light sensor. Based on the measured light intensity, the brightness level that should be set for the product display is determined by a preset mapping relationship or threshold. Specifically, the unit includes an ambient light sensor, a microcontroller or a processor, and a brightness adjustment circuit. The ambient light sensor continuously monitors the surrounding light conditions and sends the detected data to the microcontroller. The microcontroller calculates the appropriate brightness level based on the received data and a preset algorithm. Then the microcontroller sends a signal to the brightness adjustment circuit to adjust the brightness of the product display to adapt to the current changes in ambient light. In this way, whether in a bright or dim environment, the product display can be clearly visible, reducing user eye fatigue and enhancing the realism and immersion of the AR experience. Furthermore, in the above algorithm, the brightness adjustment is preset as a linear relationship, and only three different light intensity intervals are considered: low, medium, and high. In this way, a piecewise linear function can be used to represent the product brightness adjustment strategy, and three ambient light intensity thresholds L are set. low , L medium and L high , and the corresponding virtual goods display brightness level B min , B medium and B max , when the light intensity L is lower than L low When the brightness is set to the minimum value B min ; When L low ≤L <L medium When the brightness is at B min and B medium When L medium ≤L <L high When the brightness is at B medium and B max When the light intensity L is greater than or equal to L high When the brightness is set to the maximum value B maxIn some other embodiments, the range and number of intervals of the ambient illumination intensity and the corresponding commodity display brightness can be flexibly changed, and should be considered within the protection scope defined by the present invention;

[0064] Among them, the sound adjustment unit uses a microphone to capture environmental noise, implements an algorithm to evaluate the level of noise, and adjusts the size of the product-related sound effects accordingly. Specifically, the unit includes a microphone, an environmental noise detection circuit, a processor and a speaker. The microphone continuously captures the surrounding environmental sound, the environmental noise detection circuit converts the audio signal into an electrical signal, and the processor runs a noise evaluation algorithm to analyze the noise level. According to the result of the noise evaluation, the volume of the speaker is automatically adjusted to ensure that the product display sound effect is clear and audible. In addition, the noise evaluation algorithm includes preprocessing, noise level evaluation, threshold setting and volume adjustment rules. First, the sound signal captured by the microphone is preprocessed, such as filtering, to remove unnecessary background noise, and then a signal processing technology such as root mean square energy or peak detection is used to evaluate the intensity of the environmental noise. Similar to the algorithm form of the above-mentioned ambient light detection and adjustment unit, different noise threshold intervals are first defined, such as quiet, moderate, noisy, etc., and then based on the result of the noise evaluation, a set of rules are defined to determine the amplitude of the volume adjustment. This automatic volume adjustment design can effectively reduce the interference caused by environmental noise, ensure that users can clearly hear the display sound effects of the product in any environment, and make users more focused on the product display, thereby improving the overall shopping experience;

[0065] Among them, the product size adjustment unit uses a camera to monitor whether there is any obstruction in the product display area, designs an algorithm to detect obstructions, and automatically adjusts the display size of the product. Specifically, the unit includes a camera, an image processing mechanism, an occlusion detection algorithm, a size adjustment algorithm and an AR engine. The camera continuously captures the video stream of the product display area, and the image processing structure runs the occlusion detection algorithm to analyze the occlusion in the video stream, which usually involves image segmentation technology to distinguish between products and obstructions. Subsequently, image processing technologies such as edge detection and shape recognition are used to analyze the position, size and shape of the obstructing object. According to the results of the occlusion analysis, the size adjustment algorithm is used to calculate the new display size to ensure that the product can be fully displayed even in the case of occlusion. Finally, the AR engine updates the AR model of the product according to the new display size to ensure that the product is presented to the user completely and clearly. In addition, the technologies used in the occlusion detection algorithm include but are not limited to convolutional neural networks or other deep learning models to identify occluded objects. The size adjustment algorithm needs to consider the original proportions of the product to maintain the naturalness and authenticity of the product appearance, and the AR engine needs to have efficient rendering capabilities to ensure that the resized product model can be displayed smoothly. Under the action of this unit, users can see the complete product display regardless of whether the display area of ​​the product is blocked, which improves the overall shopping experience. Automatic adjustment of the product size can avoid information loss caused by occlusion, allowing users to focus more on the product itself.

[0066] Display background adjustment module: based on the background adjustment parameters, update the product display background, the background adjustment parameters are controlled by the background adjustment parameter model, wherein the background adjustment parameter model is:

[0067] S=∑ i w i f i (x i );

[0068] In the formula, i represents different influencing factors, w i is the weight of influencing factor i, f i (x i ) is the mapping function of influencing factor i, x i It represents the computable parameters corresponding to different influencing factors. S is the background update standard calculated based on the relevant influencing factors. According to the calculated value S, the suitable background in the background library is found and replaced. Each background in the background library corresponds to a set of initial scores S1. The initial score S1 is subtracted from the score S, and the background with the smallest value is used as the suitable background. When there are several display backgrounds with the same difference, the front one in the background library is selected by default.

[0069] Influencing factors i include user emotion influencing factors i E , the mapping function of this factor is:

[0070] f E (x E )= 1-(1-x E ) β , x E ≤α;

[0071] f E (x E )=log(1+x E ), x E >α;

[0072] In the formula, x E Factors affecting user emotions E , α is used to distinguish the emotional influencing factors x E The thresholds of different regions, β is the adjustment coefficient, and further, α can be expressed as the user's emotion is neutral, x E When ≤α, it means that the current user is in a low mood, and the mapped mood is positive. The larger β is, the more positive the corresponding mood is after mapping. E When >α, that is, when the sentiment score is high, a logarithmic function is used to maintain the growth trend of the score, but the rate of increase gradually slows down. This can quickly mobilize the user's emotions when the user is depressed, and smoothly change the user's emotions when the user is positive, so that the user is more focused on the product itself;

[0073] Specifically, the user emotion factor detection method is manifested in facial expression unit detection, voice intonation unit detection and physiological signal unit detection;

[0074] First, we need to collect labeled facial expression datasets. These datasets should contain images of different expression categories, and each image has a corresponding expression label. Common expression categories include but are not limited to: neutral, happy, sad, surprised, angry, disgusted, and afraid. Each category can correspond to a numerical range. Use a camera to capture the user's facial image and extract useful features from the image to describe the facial expression. Common feature extraction methods include: manual features, such as local binary patterns, oriented gradient histograms, etc.; deep learning features, use pre-trained convolutional neural networks to extract features, such as VGGFace, ResNet, etc., select a convolutional neural network model to perform expression classification, correspond the current detection image to the expression label, and get the corresponding value;

[0075] Use a microphone to capture the user's voice. First, you need to extract useful features from the voice signal. These features can help the model identify different emotional states. Common features include: acoustic features, such as fundamental frequency, pitch, duration, energy, resonance peak, etc.; spectral features, such as Mel-frequency cepstral coefficients, linear prediction coding, cepstral coefficients, etc.; time domain features, such as zero-crossing rate, energy entropy, etc. The extracted features are preprocessed, such as standardization and normalization, to ensure that the features are comparable between different samples. Use machine learning-based methods, such as support vector machines, random forests, or deep learning models such as recurrent neural networks or long short-term memory, to classify voice emotions into several categories, such as happiness, sadness, anger, surprise, neutrality, etc. Each category can correspond to a numerical range;

[0076] For the physiological signal unit, first, it is necessary to collect labeled physiological signal data. This data should include physiological signal records under different emotional states, and each record has a corresponding emotional label. Common physiological signals include but are not limited to: heart rate, skin conductance, respiratory rate, body temperature, brain waves, etc. The collected physiological signal data is preprocessed to make it suitable for further analysis. The preprocessing steps include denoising, standardization, etc. Then, useful features such as mean, variance, peak, etc. are extracted from the original signal, and emotions are divided into neutral, happy, sad, surprised, angry, disgusted, afraid, etc., and a numerical value is defined for each category. The support vector machine model is selected to perform emotion classification. Finally, the user's physiological signals are analyzed to determine the user's emotional state and the physiological signals are converted into numerical scores;

[0077] So the factors affecting user emotions are E The corresponding computable parameter x E Subject to the constraints of the user emotion parameter detection model, the model is: E =w F x F +w V x V +w H x H , where x F 、x V and x H Respectively represent the detection values ​​of facial expressions, voice intonation and physiological signals, w F 、w V and w H They represent the weights of the detection values, and w F +w V +w H =1, set x F The value range is 0-1, 0 represents a very negative expression, 1 represents a very positive expression, set x VThe value range is 0-1, 0 represents a very negative tone and 1 represents a very positive tone. H The value range is 0-1, where 0 represents a very negative physiological state and 1 represents a very positive physiological state. Through these settings, the system can accurately grasp the user's emotions and then change the display background accordingly;

[0078] The display background adjustment module also includes a feedback adjustment unit;

[0079] Set time t. After the system changes the display background, since user emotions can directly and quickly reflect the user's satisfaction with the changed product display background, and other influencing factors such as seasonal factors change almost zero in a short period of time, after t time, the computable parameters x corresponding to other influencing factors are maintained. i No change, recalculate the user sentiment value;

[0080] Specifically, set a change Δx E , Δx E =x Et -x E , where x Et is the user sentiment score at the end of time t, x E is the user sentiment score at the beginning of time t, perform the following steps:

[0081] In x E >α, that is, the initial user sentiment is good, if Δx E ≥0, that is, the user's emotion remains unchanged or becomes more positive, then keep the current background. If Δx E <0, that is, the user's mood becomes worse, then x Et Bring the background adjustment parameter model into the model, find a suitable background in the background library, and then change the current product display background;

[0082] In x E ≤α, that is, the initial user emotion is poor or moderate, if Δx E >0, that is, the user's mood has improved, then keep the current background. If Δx E ≤0, that is, the user's mood has not improved or has become worse, then the above steps are used to change the current background.

[0083] Furthermore, the feedback adjustment unit also includes a weight adjustment model, and the weight adjustment model is:

[0084] ,

[0085] In the formula, w i ′ is the adjusted weight of influencing factor i, η i is the learning rate, which controls the size of the update step, Fi For weight w i The expected value or target value of , J is the loss function, is the loss function J with respect to weight w i The gradient of , R is the reward, which represents the immediate feedback brought by the current action, γ is the discount factor, which is used to balance the importance of immediate rewards and future rewards, Q is the state-action value function, Q (s, a) represents the Q function value of taking action a in the current state s, which represents the value of the current action, max a′ Q(s′, a′) represents the maximum value of the Q function among all possible actions a′ in the next state s′, indicating the value of the optimal action. The model is based on user feedback, gradient descent algorithm, and Q-Learning algorithm. User feedback can be direct, such as ratings or comments submitted through an interactive interface, or indirect, such as user behavior data, such as click-through rate, dwell time, etc. The model combines the ideas of supervised learning and reinforcement learning, so that the weights can be dynamically adjusted in different situations to optimize the overall performance of the system.

[0086] Influencing factor i also includes seasonal influencing factor i T , use the seasonal ARIMA model to quantify the seasonal factor and use x T Indicates that 0-δ1 represents the spring value range, δ1-δ2 represents the summer value range, δ2-δ3 represents the autumn value range, and δ3-δ4 represents the winter value range. <x T When ≤δ1, f T (x T )=x T , that is, when the product display time is in spring, the display background directly uses the spring value. When δ1<x T ≤δ2, then f T (x T ) is located between δ3-δ4, that is, when the external environment where the displayed product is located is summer, a background related to winter factors is used. When δ2<x T ≤δ3, then f T (x T ) is between 0 and δ1, that is, when the external environment is in autumn, a display background related to spring is used. When δ3<x T When ≤δ4, then f T (x T) is in δ1-δ2, that is, when the displayed product is in winter, use a background related to summer. In the spring value range of 0-δ1, the higher the value, the stronger the spring atmosphere. In the summer value range of δ1-δ2, the higher the value, the higher the temperature. In the autumn value range of δ2-δ3, the higher the value, the stronger the autumn atmosphere. In the winter value range of δ3-δ4, the higher the value, the colder it is. Further set the specific season mapping relationship:

[0087] When 0 <x T When ≤δ1, f T (x T )=x T , which means that the displayed product is in spring, so the spring value is used;

[0088] When δ1<x T When ≤δ2, f T (x T ) = δ3 + (δ2-x T ) , which means that the displayed product is in summer, using winter values, and x T The larger the value, the hotter the temperature. T (x T ) The larger the value, the lower the winter temperature;

[0089] When δ2<x T When ≤δ3, f T (x T ) = (δ3-x T ) , which means that the displayed product is in autumn, using spring values, and x T The larger the corresponding f T (x T ) is bigger;

[0090] When δ3<x T When ≤δ4, f T (x T ) = δ1 + (δ4-x T ) , which means that the displayed product is in winter, using the summer value, and x T The larger the value, the lower the temperature. T (x T ) The larger the value, the higher the summer temperature;

[0091] By setting such a season mapping relationship, it is possible to adjust the user's emotions when viewing products and bring a sense of comfort. For example, summer is often accompanied by high temperature and humidity, which may make people feel uncomfortable or even irritable. Using a winter display background can help users psychologically get rid of the feeling of heat, thereby achieving the effect of mood regulation. Secondly, it can bring a sense of freshness to the product display, help deepen the user's memory of the product, and increase user attention.

[0092] Influencing factors also include local cultural factors C , in this factor, set x C Local cultural factors C A computable parameter of γ, setting the value γ indicates that the background is moderately related to the current culture, 0<γ<1, the default value is f C (x C ) = γ;

[0093] Influencing factor i also includes user preference factor i P , in this factor, set x P is the user preference factor i P The computable parameter, setting the value ε represents a medium fit to user preferences, 0<ε<1, the default value f P (x P ) = ε, medium settings that meet user preferences allow users to flexibly adjust values, add new elements while ensuring user interests, ensure user privacy, and improve favorability;

[0094] Secondly, the product display system is equipped with an interactive interface, and the weights w corresponding to different influencing factors i i And the mapping function f i (x i ) can be set manually.

[0095] User behavior analysis and product recommendation module: Provides product recommendations to users by analyzing user behavior status. This module includes a behavior analysis unit and a personalized recommendation unit.

[0096] The behavioral analysis unit includes:

[0097] Attention tracking mechanism: Use eye tracking technology to monitor the user's line of sight direction and dwell time, record the product features that the user is concerned about, and record the user's interaction process with the product. Specifically, this unit includes eye tracking, behavior recording, data analysis, and interest model building. Use high-precision eye tracking hardware equipment, such as infrared eye trackers, to capture the user's line of sight direction, including gaze points, scanning paths and other information, and record the user's line of sight dwell time, gaze times, and interaction times when viewing products. Use statistics and machine learning techniques to analyze user behavior patterns and generate visual reports of user attention, such as line of sight heat maps, to understand which products and parts of the products attract the user's attention. Finally, based on the user's eye movement data and other behavior data such as clicks and slides, combined with cluster analysis, association rule mining and other algorithms to identify similarities and differences between user groups, build a personalized interest model, and understand the types and characteristics of products preferred by users. In this way, by capturing the user's eye movement data, it is possible to more accurately predict the user's interests and needs, thereby providing product recommendations that are more in line with the user's preferences.

[0098] Sentiment analysis mechanism: including facial expression analysis and voice emotion recognition, analyzing the user's facial expression to determine his attitude towards the product, using deep learning models to identify smiling and frowning expressions, analyzing the user's tone and emotions through voice recognition technology, and identifying the user's satisfaction and interest level. Specifically, the user's facial image and voice are captured by a camera and a microphone respectively, and the user's emotions are analyzed by deep learning models and natural language processing technology respectively. Combined with the emotion analysis results of expression and voice, the user's emotional state is scored, such as positive, negative, neutral, etc. The emotion score uses a machine learning model, such as a support vector machine or a deep learning model, to give a quantitative emotion score based on the emotion analysis results of expression and voice. The results of the emotion analysis are then integrated into the user's interest model to more accurately predict the user's preference. Suppose the user shows a satisfied expression when watching a smart watch and comments by voice: "This watch is really good!" In this case, the system will recognize the user's positive emotions and may recommend more similar smart watches to the user, providing the user with more accurate product recommendations and services;

[0099] Among them, the personalized recommendation unit includes a recommendation generation agency and a viewing route planning agency. It generates a recommendation list based on the user's behavioral data, uses collaborative filtering or deep learning methods to predict the user's preferences, formulates a viewing route based on the user's interests and location information, and guides the user to the exhibition area they may be interested in. By providing highly personalized product recommendations, it increases the user's chances of finding their favorite products, thereby improving user satisfaction. Customized viewing routes can reduce the user's blind search time and make it easier for users to discover content that interests them.

[0100] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A commodity display system based on AR technology, characterized in that: include: Dynamic environment adaptation module: automatically adjusts the display effect of products in different environments; Display background adjustment module: based on the background adjustment parameters, updates the product display background, and the background adjustment parameters are controlled by the background adjustment parameter model; User behavior analysis and product recommendation module: By analyzing user behavior status, it provides product recommendations to users and displays them in the product display system; Background library: set up several product display backgrounds in the background library; The background adjustment parameter model is: S=∑ i w i f i (x i ); Among them, i represents different influencing factors, w i is the weight of influencing factor i, f i (x i ) is the mapping function of influencing factor i, x i Indicates the computable parameters corresponding to different influencing factors; S is a background update standard calculated based on relevant influencing factors. According to the calculated value S, the suitable background in the background library is found and replaced; The step of finding a suitable background in the background library according to the calculated value S and replacing it comprises: Each background in the background library corresponds to a set of initial scores S1. The background with the smallest value obtained by subtracting the score S from the initial score S1 is used as the adapted background. When there are several display backgrounds with the same difference, the one at the front of the background library is selected by default; The mapping function of the user emotion influencing factor E is: f E (x E )= 1-(1-x E ) β ,x E ≤α; f E (x E )=log(1+x E ),x E >α; Among them, x E is a computable parameter of the user emotion influencing factor E, α is the threshold used to distinguish different regions of the emotion influencing factor E, α indicates that the user emotion is neutral, and x E When ≤α, it means that the current user is in a low mood, and the mapped mood is positive. The larger β is, the more positive the corresponding mood is after mapping. E When >α, that is, when the emotion score is high, a logarithmic function is used to maintain the growth trend of the score, but the rate of increase gradually slows down. This can quickly mobilize the user's emotions when the user is depressed, and smoothly change the user's emotions when the user is positive. β is the adjustment coefficient; The display background adjustment module also includes a feedback adjustment unit: Set time t, after the system changes the display background, after t time, keep the computable parameters x corresponding to other influencing factors i No change, recalculate the user sentiment value; Set a change Δx E , Δx E =x Et -x E , where x Et is the user sentiment score at the end of time t, x E is the user sentiment score at the beginning of time t, perform the following steps: In x E >α, if Δx E ≥0, keep the current background. If Δx E <0, then change the current background; In x E ≤α, if Δx E >0, keep the current background, if Δx E ≤0, the current background will be changed.

2. The product display system based on AR technology according to claim 1, characterized in that: The computable parameter x corresponding to the user emotion influencing factor E E Subject to the constraints of the user emotion parameter detection model, the user emotion parameter detection model is: x E =w F x F +w V x V +w H x H ; In the formula, x F 、x V and x H Respectively represent the detection values ​​of facial expressions, voice intonation and physiological signals, w F 、w V and w H They represent the weights of the detection values, and w F +w V +w H =1.

3. The product display system based on AR technology according to claim 1, characterized in that: The feedback adjustment unit also includes a weight adjustment model, which is: ， In the formula, w i ′ is the adjusted weight of influencing factor i, η i is the learning rate, F i For weight w i The expected value of , J is the loss function, is the loss function J with respect to weight w i The gradient of , R is the reward, γ is the discount factor, Q is the state-action value function, Q(s, a) represents the Q function value of taking action a in the current state s, max a′ Q(s′, a′) represents the maximum value of the Q function among all possible actions a′ in the next state s′.

4. The product display system based on AR technology according to claim 1, characterized in that: The mapping function of the seasonal influencing factor T is: f T (x T )=x T ,0<x T ≤δ1; f T (x T )∈[δ3-δ4],δ1<x T ≤δ2; f T (x T )∈[0-δ1],δ2<x T ≤δ3; f T (x) T )∈[δ1-δ2],δ3<x T ≤δ4; Among them, x T is a computable parameter of the seasonal influencing factor T, 0-δ1 represents the numerical range of spring, δ1-δ2 represents the numerical range of summer, δ2-δ3 represents the numerical range of autumn, and δ3-δ4 represents the numerical range of winter.

5. The product display system based on AR technology according to claim 4, characterized in that: The mapping function of the seasonal influencing factor T also includes: f T (x T )=δ3+(δ2-x T ) ,δ1<x T ≤δ2; f T (x T )=(δ3-x T ) ,δ2<x T ≤δ3; f T (x T )=δ1+(δ4-x T ) ,δ3<x T ≤δ4。 6. The product display system based on AR technology according to claim 1, characterized in that: Also includes: The mapping relationship of local cultural factor C is: f C (x) C )=γ′; Among them, x C is a computable parameter of the local cultural factor C, and the value γ′ indicates that the product display context is moderately related to the current culture; The mapping relationship of user preference factor P is: f P (x) P )=e; Among them, x P is a computable parameter of the user preference factor P, and the value ε represents a moderate fit with the user preference.

Images