AI-based user demand analysis method and related equipment

Abstract

The invention relates to an AI-based user demand analysis method and related equipment, and the method comprises the following steps: continuously collecting facial expressions of a target user through a video collection device, and obtaining a video sequence; extracting emotion features to obtain a user emotion state feature set; performing time sequence analysis to obtain an emotion change track map; analyzing demand tendency according to the atlas to obtain a demand tendency vector set; and finally, dynamic service matching is carried out based on the vector set, a personalized service recommendation scheme including a primary service, a secondary service combination and an execution time sequence is generated, and user requirements are accurately met. The technical problems that in the prior art, an emotion change dynamic process is lack of effective analysis, accurate association between emotions and demands cannot be established, and actual demands of an enterprise for prospective analysis and accurate service recommendation on user demands are difficult to meet are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of user requirements, and particularly relates to an AI-based user requirement analysis method and related devices. Background Art

[0002] With the rapid development of Internet technology and the increasing diversification of user requirements, how to accurately grasp user requirements and provide personalized services has become the key for enterprises to enhance competitiveness and user experience. Traditional user requirement analysis methods mostly rely on methods such as questionnaires and user interviews. These methods have problems such as high data collection costs, poor timeliness, and strong subjectivity, and are difficult to adapt to the rapidly changing market environment and user requirements. At the same time, existing analysis technologies based on text or simple behavioral data cannot comprehensively capture the emotional information contained in the interaction process of users, resulting in insufficient and inaccurate understanding of user requirements and inability to provide truly personalized services that meet the needs of users.

[0003] In the digital age, the interaction scenarios between users and various products and services are becoming increasingly complex. A large amount of unstructured emotional data contains rich clues of user requirements, but this data has not been fully explored and utilized. When dealing with user emotional data, existing technologies often lack effective analysis of the dynamic process of emotional changes and cannot establish an accurate association between emotions and requirements, resulting in poor demand prediction effects based on emotional analysis and difficulty in meeting the actual needs of enterprises for forward-looking analysis of user requirements and accurate service recommendations.

[0004] In addition, in the face of massive and rapidly updated user data, traditional requirement analysis models are difficult to achieve a balance between processing efficiency and analysis depth, and cannot achieve real-time and dynamic tracking and response to user requirements. Therefore, there is an urgent need for a new type of user requirement analysis method that can integrate multi-dimensional user data, especially emotional data, deeply analyze user requirements through intelligent means, and achieve dynamic service matching, so as to solve many problems existing in current user requirement analysis and improve the accuracy of services and user satisfaction. Summary of the Invention

[0005] The main object of the present invention is to provide an AI-based user requirement analysis method and related devices, which solve the technical problem that existing technologies often lack effective analysis of the dynamic process of emotional changes, cannot establish an accurate association between emotions and requirements, and are difficult to meet the actual needs of enterprises for forward-looking analysis of user requirements and accurate service recommendations.

[0006] To achieve the above object, the present invention provides an AI-based user requirement analysis method, including the following steps: Continuously collect video of the facial expressions of the target user through a video acquisition device to obtain a target user video sequence; Extract emotional features of the target user based on the target user video sequence to obtain a user emotional state feature set; Conduct emotional pattern time series analysis on the target user based on the user emotional state feature set to obtain a user emotional change trajectory map; Conduct demand tendency analysis on the target user based on the user emotional change trajectory map to obtain a user demand tendency vector set; Conduct dynamic service matching on the target user based on the user demand tendency vector set to obtain a personalized service recommendation plan.

[0007] Furthermore, the extracting emotional features of the target user based on the target user video sequence to obtain a user emotional state feature set includes: Perform state link tracking based on the emotional state transition probability map to obtain a set of emotional evolution paths, and perform multi-dimensional projection on the set of emotional evolution paths to obtain an emotional trajectory projection map; Extract feature points from the emotional trajectory projection map to obtain a sequence of key emotional nodes, and perform trajectory segmentation based on the sequence of key emotional nodes to obtain an emotional stage division map; Perform trajectory reconstruction based on the emotional stage division map to obtain an emotional change profile curve, and perform curvature analysis on the emotional change profile curve to obtain an emotional change feature vector; Perform spatio-temporal dimension fusion on the emotional change feature vector to obtain an emotional trajectory feature surface, and perform multi-dimensional feature combination based on the emotional trajectory feature surface to obtain a user emotional change trajectory map, where the user emotional change trajectory map includes emotional development trends, emotional fluctuation patterns, and emotional conversion rules.

[0008] Furthermore, the conducting emotional pattern time series analysis on the target user based on the user emotional state feature set to obtain a user emotional change trajectory map includes: Perform time window segmentation on the user emotional state feature set to obtain an emotional state sequence matrix, and perform time series correlation calculation on the emotional state sequence matrix to obtain an emotional state transition probability map; Perform state link tracking based on the emotional state transition probability map to obtain a set of emotional evolution paths, and perform multi-dimensional projection on the set of emotional evolution paths to obtain an emotional trajectory projection map; Extract feature points from the emotional trajectory projection map to obtain a sequence of key emotional nodes, and perform trajectory segmentation based on the sequence of key emotional nodes to obtain an emotional stage division map; Based on the emotional stage division graph, trajectory reconstruction is carried out to obtain an emotional change profile curve, and curvature analysis is performed on the emotional change profile curve to obtain an emotional change feature vector; Perform spatio-temporal dimension fusion on the emotional change feature vector to obtain an emotional trajectory feature surface, and perform multi-dimensional feature combination based on the emotional trajectory feature surface to obtain a user emotional change trajectory map, which includes emotional development trends, emotional fluctuation patterns, and emotional conversion rules.

[0009] Furthermore, perform demand tendency analysis on the target user based on the user emotional change trajectory map to obtain a set of user demand tendency vectors, including: Perform multi-dimensional feature aggregation on the user emotional change trajectory map to obtain an emotion-behavior correlation matrix, and perform behavior decision feature extraction on the emotion-behavior correlation matrix to obtain a set of behavior trend features; Perform context matching analysis on the user emotional change trajectory map based on the set of behavior trend features to obtain a context demand mapping matrix, and perform constraint condition analysis on the context demand mapping matrix to obtain a set of context demand features; Perform multi-dimensional decomposition on the set of context demand features to obtain a demand priority distribution map, and perform hierarchical processing on the set of context demand features based on the demand priority distribution map to obtain a demand hierarchy feature matrix; Perform dynamic weight adjustment on the context demand mapping matrix based on the demand hierarchy feature matrix to obtain an initial dynamic demand weight matrix, and perform time series analysis on the initial dynamic demand weight matrix to obtain a demand tendency trajectory curve; Perform vectorization modeling on the demand tendency trajectory curve to obtain a set of user demand tendency vectors.

[0010] Furthermore, perform dynamic weight adjustment on the context demand mapping matrix based on the demand hierarchy feature matrix to obtain a dynamic demand weight matrix, including: Perform hierarchical weight assignment on the demand hierarchy feature matrix to obtain a demand hierarchy weight distribution table, and perform multi-dimensional correlation analysis on the demand hierarchy weight distribution table to obtain a demand hierarchy correlation matrix; Perform context adaptability analysis on the context demand mapping matrix based on the demand hierarchy correlation matrix to obtain a context demand adaptation matrix, and perform dynamic weight correction on the context demand adaptation matrix to obtain a demand dynamic adaptation weight table; Perform time series feature extension on the demand dynamic adaptation weight table to obtain a dynamic demand weight matrix, and perform time series trend analysis on the dynamic demand weight matrix to obtain a demand weight evolution trajectory map; Normalize the dynamic demand weight matrix based on the demand weight evolution trajectory diagram to obtain a weight normalization matrix, and perform multi-dimensional feature fusion on the weight normalization matrix to obtain a dynamic demand weight matrix.

[0011] Further, the dynamic service matching of the target user based on the user demand tendency vector set to obtain a personalized service recommendation scheme includes: Perform multi-scale tensor decomposition on the user demand tendency vector set to obtain a demand dimension component matrix, and perform cross-modal correlation analysis on the demand dimension component matrix to obtain a demand-service coupling map; Based on the demand-service coupling map, perform service resource topology mapping on the demand dimension component matrix to obtain a service resource matching network, and perform dynamic manifold analysis on the service resource matching network to obtain a service scheduling trajectory diagram; Extract spatio-temporal features from the service scheduling trajectory diagram to obtain a service response feature surface, and perform multi-objective optimization calculation based on the service response feature surface to obtain a service combination strategy matrix; Perform personalized feature fusion based on the service combination strategy matrix to obtain a service recommendation feature cube, and perform dynamic pruning optimization on the service recommendation feature cube to obtain a personalized service recommendation scheme, where the personalized service recommendation scheme includes main service content, secondary service combination, and service execution timing.

[0012] Further, the service resource topology mapping of the demand dimension component matrix based on the demand-service coupling map to obtain a service resource matching network includes: Perform tensor product decomposition on the demand-service coupling map to obtain a coupling relation basis vector group, and perform eigenvalue decomposition on the demand dimension component matrix to obtain a demand eigen-subspace matrix; Based on the coupling relation basis vector group, perform vector space mapping on the demand eigen-subspace matrix to obtain a demand-service joint feature embedding matrix; Perform local neighborhood weighting on the demand-service joint feature embedding matrix through a Gaussian kernel function to obtain a weighted neighborhood feature matrix, and perform Delaunay triangulation on the weighted neighborhood feature matrix to obtain an initial service resource topology structure diagram; Based on the initial service resource topology structure diagram, perform topological constraint projection on the demand-service joint feature embedding matrix to obtain a topological constraint feature matrix, and perform minimum spanning tree construction on the topological constraint feature matrix to obtain a service resource core connection network; Perform a multi-scale dilation operation on the core connection network of the service resources to obtain an extended connection graph of the service resources, and perform topological connectivity analysis based on the extended connection graph of the service resources to obtain a service resource matching network.

[0013] The present invention also provides an AI-based user demand analysis system, including: A collection module for continuously collecting videos of the facial expressions of a target user through a video collection device to obtain a target user video sequence; An extraction module for extracting emotional feature of the target user based on the target user video sequence to obtain a user emotional state feature set; A first analysis module for performing an emotional pattern time series analysis on the target user based on the user emotional state feature set to obtain a user emotional change trajectory map; A second analysis module for performing a demand tendency analysis on the target user based on the user emotional change trajectory map to obtain a user demand tendency vector set; A matching module for performing dynamic service matching on the target user based on the user demand tendency vector set to obtain a personalized service recommendation scheme.

[0014] The present invention also provides a computer device, including a memory and a processor, where a computer program is stored in the memory, and when the processor executes the computer program, the steps of the method described in any one of the above are implemented.

[0015] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described in any one of the above are implemented.

[0016] A method for analyzing user needs based on AI provided by the present invention includes the following steps: continuously collecting the facial expressions of a target user through a video acquisition device to obtain a video sequence; extracting emotional features therefrom to obtain a set of user emotional state features; performing time series analysis on the set to obtain an emotional change trajectory map; analyzing the demand tendency based on the map to obtain a set of demand tendency vectors; and finally performing dynamic service matching based on the set of vectors to generate a personalized service recommendation scheme including a primary service, a secondary service combination, and an execution time sequence, so as to accurately meet the user's needs. The technical problem that the prior art often lacks an effective analysis of the dynamic process of emotional changes, cannot establish an accurate association between emotions and needs, and is difficult to meet the actual needs of enterprises for forward-looking analysis of user needs and accurate service recommendations is solved. The demand tendency analysis is realized based on the user emotional change trajectory map, and a set of user demand tendency vectors is obtained. By combining emotional changes with demand tendencies, the potential and unexpressed needs of users can be deeply explored. Compared with traditional demand analysis methods, this emotion-based demand analysis can better touch the real needs of users and improve the accuracy and depth of demand analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a schematic diagram of the steps of a method for analyzing user needs based on AI in an embodiment of the present invention; Figure 2 is a block diagram of the structure of a system for analyzing user needs based on AI in an embodiment of the present invention; Figure 3 is a schematic block diagram of the structure of a computer device in an embodiment of the present invention.

[0018] The implementation, functional features, and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0020] As Figure 1 shown, Figure 1 is a schematic diagram of the steps of a method for analyzing user needs based on AI in an embodiment of the present invention; An embodiment of the present invention provides a method for analyzing user needs based on AI, including the following steps: Step S1, continuously collecting video of the facial expressions of a target user through a video acquisition device to obtain a video sequence of the target user.

[0021] Specifically, to continuously capture the facial expressions of the target user through a video acquisition device and obtain the target user video sequence, a suitable video acquisition device needs to be deployed first. This device should have the capabilities of high-definition shooting and continuous recording to ensure that it can clearly and uninterruptedly capture the details of facial expressions. In actual operation, the device is installed in an appropriate position. For example, in an intelligent retail scenario, it can be placed above the shelves or near the cashier desk to ensure that the camera is directly facing the customer's face. After the acquisition function is enabled, the device captures the face of the target user at a certain frame rate (such as 30 frames per second). Each frame of the image records the instantaneous state of the user's facial expression. As time goes by, these consecutive images are arranged in the order of shooting, forming the target user video sequence. For example, in the movie viewing feedback scenario in a cinema, the video acquisition device installed in the hall continuously captures the facial expressions of the audience throughout the movie, and finally obtains a video sequence containing the expression changes of the audience at different times during the movie viewing, providing the original data support for subsequent emotional feature extraction, demand analysis, etc. based on this sequence.

[0022] Step S2: Extract emotional features of the target user based on the target user video sequence to obtain a user emotional state feature set.

[0023] Specifically, after obtaining the target user video sequence, to extract emotional features based on this and obtain a user emotional state feature set, computer vision and deep learning technologies need to be used. First, use image recognition algorithms to process each frame of the video sequence to accurately locate the user's facial area. Through facial key point detection technology, the feature points of key parts such as eyes, eyebrows, and mouth are identified. For example, in an intelligent retail scenario, when a customer browses products, subtle movements such as the contraction of the eye muscles and the upward curvature of the mouth in the video frame can be reflected through the changes in the feature points. Then, input the feature point data into a pre-trained emotion analysis model. This model is trained based on a large amount of labeled facial expression data and can analyze the dynamic changes of the feature points, judge the emotion category corresponding to each frame, such as happy, surprised, dissatisfied, etc., and give the corresponding emotion intensity value. As the analysis of the entire video sequence progresses, information such as the emotion category and intensity of each frame is summarized and integrated, forming a user emotional state feature set containing the emotional states of the user at different times, laying a foundation for subsequent in-depth analysis of the user's emotional pattern and demand tendency.

[0024] Step S3: Conduct a temporal analysis of the emotional pattern of the target user based on the user emotional state feature set to obtain a user emotional change trajectory map.

[0025] Specifically, after obtaining the user's emotional state feature set, emotional pattern time series analysis can be carried out on the target user based on this, and then a user emotional change trajectory map can be obtained. First of all, each emotional feature in the user's emotional state feature set needs to be arranged in chronological order because these features are collected at different times and have a clear chronological order. Through this arrangement, the change of the target user's emotional state over time can be clearly seen. Then, the time series analysis method is used to deeply analyze the arranged emotional features. This method takes into account the continuity and correlation of emotional features in the time dimension and identifies patterns such as trends, periodicities, and outliers. For example, if it is found that the emotional features of the target user show a trend of gradually changing from calm to excited over a period of time, it can be judged that their mood is gradually rising; if there are periodic emotional fluctuations, they may be related to specific events or environmental factors. Taking the intelligent retail scenario as an example, when a customer enters a store, their emotional state will continuously change as they browse different products. Through emotional pattern time series analysis, if it is found that when the customer sees a certain type of product, the emotional features repeatedly show peaks of excitement and these peaks show a certain periodicity, it may mean that this type of product has a greater attraction to this customer. Finally, the emotional patterns and change rules obtained from the analysis are visually displayed in the form of a map, that is, the user emotional change trajectory map is obtained. In this map, the horizontal axis represents time and the vertical axis represents different emotional states, and the dynamic change process of the target user's emotion is intuitively presented in the form of curves or bar charts. This map can help us understand the emotional change rules of the target user more comprehensively and deeply, providing a strong basis for subsequent demand tendency analysis and dynamic service matching.

[0026] Step S4: Based on the user emotional change trajectory map, perform demand tendency analysis on the target user to obtain a user demand tendency vector set.

[0027] Specifically, after obtaining the user's emotional change trajectory map, the target user's demand tendency can be analyzed based on this, so as to obtain the user demand tendency vector set. First, it is necessary to establish an association model between emotions and demands. This model is constructed based on a large amount of historical data and domain knowledge, and it can reflect the possible demands corresponding to different emotional states and emotional change patterns. When obtaining the user's emotional change trajectory map, analysts will interpret the emotional states and changes presented in the map according to the association model. For example, in the intelligent retail scenario, if the map shows that the customer's emotional state gradually changes from calm to excited near a display area of a certain electronic product, and the excited state lasts for a period of time, according to the association model, this may mean that the customer has a strong interest in the electronic product and there is a potential demand to purchase the product. At the same time, analysts will comprehensively consider factors such as the amplitude, frequency, and duration of emotional changes. If the customer's emotions fluctuate slightly several times and last for a long time when seeing a certain series of clothing, this may imply that the customer has a certain degree of attention to this series of clothing, but there is still hesitation and may need more information or discounts to generate a purchase behavior. By comprehensively and carefully analyzing the user's emotional change trajectory map, each possible demand and the corresponding demand intensity are quantitatively represented, and finally a series of vectors are formed. These vectors constitute the user demand tendency vector set. In this vector set, each vector represents a demand tendency, and the dimension and value of the vector reflect the specific content and intensity of the demand. For example, in the above intelligent retail scenario, a vector may represent the purchase demand for a certain electronic product, and the larger the value of the vector, the stronger the purchase demand. This user demand tendency vector set provides an accurate basis for subsequent dynamic service matching and can help enterprises provide more accurate personalized service recommendation solutions for target users.

[0028] Step S5, perform dynamic service matching on the target user based on the user demand tendency vector set to obtain a personalized service recommendation solution.

[0029] Specifically, after obtaining the user demand tendency vector set, dynamic service matching can be performed on the target user based on this, so as to obtain a personalized service recommendation plan. First of all, a service resource library containing various service types and service contents needs to be constructed. This resource library covers various service information that may meet the needs of the target user. When the user demand tendency vector set is obtained, the demand tendency represented by each vector in it will be matched with the services in the service resource library. The matching process will use advanced algorithms and technologies, comprehensively considering factors such as the specific content and intensity of the demand, as well as the characteristics and applicable scope of the service. For example, in the intelligent retail scenario, if the user demand tendency vector set shows that the user has a strong demand to purchase sports equipment of a certain brand, then services related to the sports equipment of this brand, such as product introductions, purchase discount activities, and matching suggestions, will be screened out in the service resource library. At the same time, since the user's needs are dynamically changing, the service matching process is also dynamic. The services for matching will be adjusted in a timely manner according to the real-time emotional state and demand changes reflected by the user emotional change trajectory map. For example, if it is found that the user becomes a little hesitant when browsing a certain sports product, it may be due to doubts about the price or style. At this time, some services that offer price discounts or display more style options will be dynamically matched. After the matching is completed, a personalized service recommendation plan will be generated according to the matching results. This plan will comprehensively consider factors such as the user's demand tendency, the current emotional state, and historical interaction data to ensure that the recommended services not only meet the user's needs but also have high pertinence and practicality. For example, for the user who has a demand for sports equipment mentioned above, the personalized service recommendation plan may include time-limited discount information for the sports equipment of this brand, product matching recommendations suitable for the user's body shape and exercise habits, etc. Through this way of dynamic service matching, a personalized service recommendation plan that truly meets the needs of the target user can be provided, improving the user's satisfaction and experience.

[0030] In a specific embodiment, the emotional feature extraction of the target user based on the target user video sequence to obtain a user emotional state feature set includes: Based on the emotional state transition probability graph, state link tracking is performed to obtain a set of emotional evolution paths, and multi-dimensional projection is performed on the set of emotional evolution paths to obtain an emotional trajectory projection graph; Extract the feature points of the emotional trajectory projection graph to obtain a key emotional node sequence; Based on the key emotional node sequence, trajectory reconstruction is performed to obtain an emotional change profile curve, and curvature analysis is performed on the emotional change profile curve to obtain an emotional change feature vector; Perform spatio-temporal dimension fusion on the emotional change feature vector to obtain an emotional trajectory feature surface, and perform multi-dimensional feature combination based on the emotional trajectory feature surface to obtain a user emotional change trajectory map, where the user emotional change trajectory map includes an emotional development trend, an emotional fluctuation pattern, and an emotional conversion rule.

[0031] Specifically, after obtaining the target user's video sequence, a series of closely connected steps are required to extract the emotional features of the target user and obtain the user's emotional state feature set. First, perform face region localization and tracking on the target user's video sequence. With the help of advanced computer vision algorithms, this process can accurately identify the face region of the target user in each frame of the video and continuously track the position changes in the video sequence, thereby obtaining a sequence diagram of facial feature points. The sequence diagram of facial feature points records the change information of the key parts of the user's face, such as the positions of the eyes, eyebrows, mouth, etc. Subsequently, perform micro-expression dynamic analysis on this sequence diagram. Micro-expressions often contain rich emotional information. By analyzing the movement trajectories and change amplitudes of facial feature points, a facial muscle deformation matrix is calculated. For example, in the intelligent retail scenario, when a customer is browsing a certain product, the subtle contraction of the muscles around the eyes or the slight upward curve of the mouth can be reflected in the facial muscle deformation matrix, and each value in the matrix represents the deformation degree of a specific facial muscle. Based on the facial muscle deformation matrix, further carry out expression semantic analysis. By comparing with a large amount of data with labeled emotion categories and performing operations of machine learning algorithms, a facial emotion intensity map is obtained. This map visually shows the emotional intensity distribution of the user's face at different times, and different colors and gray values correspond to different emotional intensities. Then, extract the time-domain features of the map, converting the emotional intensity changes in the time dimension into a set of emotional fluctuation curves. For example, in a shopping experience, the customer's emotions, from curiosity when entering the store, to excitement when seeing a certain product, and then to hesitation when considering the price, the ups and downs of these emotions on the time axis will be clearly reflected in the set of emotional fluctuation curves. Data such as the slope and peak of the curves can intuitively reflect the change speed and degree of emotional intensity. To further explore the emotional features, perform frequency-domain decomposition operations on the set of emotional fluctuation curves, converting the emotional changes in the time domain to the frequency domain for analysis, and obtaining an emotional spectrum feature map. In the frequency domain, the distribution of different frequency components in the emotional changes can be found. Some emotional changes are high-frequency and short-term, while others are low-frequency and long-term. Based on the emotional spectrum feature map, perform emotional periodicity analysis to find the rules of the user's emotional changes, thereby obtaining an emotional rhythm feature vector. Taking the customer's shopping process in the mall as an example, if it is found that the emotional intensity fluctuates slightly every 10 minutes when browsing the clothing area, this may imply that there is a certain periodic thinking when choosing products, and this rule will be reflected in the emotional rhythm feature vector. Finally, based on the emotional rhythm feature vector, perform multi-dimensional feature combination, integrating multi-dimensional features such as facial muscle deformation, emotional intensity map, emotional fluctuation curves, and emotional rhythm, thereby obtaining the user's emotional state feature set.The feature set includes basic emotion categories such as happiness, sadness, anger, etc.; emotion intensity levels, which are divided into levels 1-10 through quantitative analysis. For example, when seeing a desired product, the excitement emotion intensity of a customer may reach level 8; and emotion duration stability, which determines whether the emotion changes rapidly or is relatively stable by analyzing the smoothness and periodic characteristics of the emotion fluctuation curve. For instance, when a customer stops in front of a certain product, if the emotion fluctuation curve is relatively flat, it indicates that the emotion duration stability is relatively high and the interest in the product is relatively stable. In the intelligent retail scenario, through this series of rigorous and comprehensive emotion feature extraction steps, the emotional state of customers can be accurately grasped, providing an accurate and rich data basis for subsequent emotion pattern time series analysis, demand tendency analysis, and dynamic service matching based on the user emotion state feature set, helping merchants provide more personalized services that better meet the needs of customers.

[0032] In a specific embodiment, the emotion pattern time series analysis of the target user based on the user emotion state feature set to obtain a user emotion change trajectory map includes: Segment the user emotion state feature set by time window to obtain an emotion state sequence matrix, and calculate the time series correlation of the emotion state sequence matrix to obtain an emotion state transition probability map; Based on the emotion state transition probability map, perform state link tracing to obtain an emotion evolution path set, and perform multi-dimensional projection on the emotion evolution path set to obtain an emotion trajectory projection map; Extract feature points from the emotion trajectory projection map to obtain a key emotion node sequence, and segment the trajectory based on the key emotion node sequence to obtain an emotion stage division map; Based on the emotion stage division map, perform trajectory reconstruction to obtain an emotion change profile curve, and perform curvature analysis on the emotion change profile curve to obtain an emotion change feature vector; Perform spatio-temporal dimension fusion on the emotion change feature vector to obtain an emotion trajectory feature surface, and perform multi-dimensional feature combination based on the emotion trajectory feature surface to obtain a user emotion change trajectory map, where the user emotion change trajectory map includes emotion development trends, emotion fluctuation patterns, and emotion conversion rules.

[0033] Specifically, after obtaining the user's emotional state feature set, to conduct a temporal analysis of the emotional pattern of the target user to obtain a map of the user's emotional change trajectory, a series of complex and rigorous processes are required. First, the user's emotional state feature set is segmented by time windows, that is, the entire emotional data is divided according to a fixed time interval (for example, taking 5 seconds as a time window), and the emotional data within each time window forms a unit. Numerous units are combined to form an emotional state sequence matrix. In this way, the continuous emotional data is discretized, facilitating subsequent analysis. Next, temporal correlation calculation is performed on the emotional state sequence matrix. By analyzing the degree of association of emotional states within adjacent time windows, an emotional state transition probability map is obtained. For example, in the intelligent retail scenario, if a customer is in an "interested" emotional state within the previous 5 seconds, through calculation, the probabilities of transferring to other emotional states such as "excited" and "hesitant" within the next 5 seconds can be obtained. These probability relationships are visually presented with numerical values in the transition probability map. For example, the probability of transferring from "interested" to "excited" is 60%, and the probability of transferring to "hesitant" is 30%. Based on the emotional state transition probability map, state link tracking work is carried out. Along the transfer paths with higher probabilities, a series of emotional evolution paths are sorted out, and then an emotional evolution path set is obtained. These paths reflect the possible trajectories of the user's emotions over time. Subsequently, multi-dimensional projection is performed on the emotional evolution path set to simplify and display the complex emotional evolution paths from multiple perspectives, obtaining an emotional trajectory projection map. For example, by projecting from the emotional intensity dimension and the time dimension, the change trend of emotional intensity on the time axis can be clearly presented. During the process of a customer selecting goods, the emotional trajectory projection map can show how the emotional intensity gradually rises from the initial low intensity (such as calm, intensity value of 2) to a high intensity (such as excited, intensity value of 8), and may also decline. Next, feature points are extracted from the emotional trajectory projection map to identify those representative key emotional nodes, such as the peak, valley, and turning points of emotional intensity, forming a key emotional node sequence. Based on this sequence, trajectory segmentation is carried out, and the entire emotional trajectory is divided into different stages according to the key emotional nodes, obtaining an emotional stage division map. For example, the entire process of a customer entering the store and leaving the store is divided into stages such as "initial browsing (calm)", "discovering interesting goods (interested)", "considering purchase (hesitant)", "deciding to purchase (excited)" based on the key emotional nodes, and each stage corresponds to different emotional characteristics and time intervals. Then, based on the emotional stage division map, trajectory reconstruction is carried out, integrating the emotional data of each stage to draw an emotional change contour curve, which intuitively shows the overall contour of the user's emotions changing over time. Curvature analysis is performed on the emotional change contour curve. By calculating the degree of curvature of the curve at different points, an emotional change feature vector is obtained.Places with large curvature indicate drastic emotional changes. For example, when a customer sees the price of a product, the curvature of the emotional change contour curve suddenly increases here, indicating that their emotion has changed significantly in a short period of time. The emotional change feature vector will record this change information. Finally, the emotional change feature vector is fused in the time and space dimensions, combining the emotional changes in the time dimension with the space dimension (such as emotions triggered by different scenarios and different products) to obtain the emotional trajectory feature surface. This surface can more comprehensively display the changes in the user's emotions within the time and space range. Based on the emotional trajectory feature surface, multi-dimensional feature combinations are performed, and finally the user's emotional change trajectory map is obtained. This map contains emotional development trends, such as whether the customer's emotion gradually rises, falls, or first rises and then falls during the entire shopping process; emotional fluctuation patterns, such as small fluctuations occurring at regular intervals; emotional conversion rules, such as the probability and conditions for converting from "hesitation" to "abandoning the purchase", etc. Through this series of analyses and processes, merchants can deeply understand the emotional change patterns of customers during the shopping process, laying a solid foundation for accurately grasping customer needs and providing personalized services.

[0034] In a specific embodiment, the target user is analyzed for demand tendency based on the user's emotional change trajectory map, and a set of user demand tendency vectors is obtained, including: Perform multi-dimensional feature aggregation on the user's emotional change trajectory map to obtain an emotion-behavior correlation matrix, and extract behavior decision-making features from the emotion-behavior correlation matrix to obtain a set of behavior trend features; Based on the set of behavior trend features, perform context matching analysis on the user's emotional change trajectory map to obtain a context demand mapping matrix, and analyze the constraint conditions of the context demand mapping matrix to obtain a set of context demand features; Perform multi-dimensional decomposition on the set of context demand features to obtain a demand priority distribution map, and based on the demand priority distribution map, perform hierarchical processing on the set of context demand features to obtain a demand hierarchy feature matrix; Based on the demand hierarchy feature matrix, perform dynamic weight adjustment on the context demand mapping matrix to obtain an initial dynamic demand weight matrix, and perform time series analysis on the initial dynamic demand weight matrix to obtain a demand tendency trajectory curve; Perform vectorization modeling on the demand tendency trajectory curve to obtain a set of user demand tendency vectors.

[0035] Specifically, after having the user emotion change trajectory map, in order to realize the demand tendency analysis of the target user and obtain the user demand tendency vector set, a series of progressive analysis processes are required. First, the user emotion change trajectory map is subjected to multi-dimensional feature aggregation, and the characteristics of the emotion development trend, fluctuation pattern, conversion law, etc. contained in the map are integrated with the actual behavior data of the target user during the shopping process, such as the stay time, the type of goods browsed, the number of click operations, etc., to form an emotion-behavior association matrix. For example, in the smart retail scenario, if a customer stays in a certain electronic product display area for 5 minutes, and the emotion intensity rises from the initial level 3 to level 7 during this period, the correlation between this emotion change and the stay behavior will be recorded in the matrix. Then, the key features related to the behavioral decision are extracted from the emotion-behavior association matrix to obtain a set of behavioral trend features. These features can reflect the potential trend of user behavior. For example, if a customer frequently browses a certain type of goods, it may imply that he has a purchase intention. Based on the set of behavioral trend features, the user emotion change trajectory map is subjected to context matching analysis. The different shopping situations in which users are located, such as in the promotion area, new product display area, rest area, etc., are combined with emotional changes and behavioral trends to obtain a situational demand mapping matrix. For example, when customers are in the promotion area, their emotions fluctuate greatly and they stay longer. The matrix will record the correspondence between the situation and the potential demand. Subsequently, the situational demand mapping matrix is ​​subjected to constraint analysis, and the impact of factors such as price, inventory, and personal preferences on demand is considered to obtain a situational demand feature set. For example, even if a customer shows strong interest in a certain product, if the price is too high or the inventory is insufficient, their actual demand will be restricted. These restrictions will be reflected in the feature set. In order to more clearly present the importance of demand, the situational demand feature set is decomposed in multiple dimensions. By analyzing the differences in different demands in dimensions such as time, space, and importance, a demand priority distribution map is obtained. For example, during the shopping process, customers may always give higher priority to product quality, while their priority to gifts is relatively lower. The distribution map will use different colors or values ​​to mark these priority differences. Based on this distribution diagram, the situational demand feature set is hierarchically processed, and the demands are divided into different levels such as core demands, secondary demands, and additional demands to form a demand hierarchy feature matrix. Next, based on the demand hierarchy feature matrix, the situational demand mapping matrix is ​​dynamically weighted. Different demands are assigned different weights according to the priority and level of the demands to obtain the initial dynamic demand weight matrix. For example, core demands are assigned higher weights (such as 0.7), secondary demands are assigned lower weights (such as 0.2), and additional demands are assigned even lower weights (such as 0.1). The initial dynamic demand weight matrix is ​​subjected to time series analysis to observe the changes in demand weights over time and obtain the demand tendency trajectory curve.At different stages of customer shopping, their demand tendencies change, and the curve can intuitively display this changing trend. For example, the demand weight for product styles is relatively high during the browsing stage, while the demand weight for price discounts increases during the checkout stage. Finally, vectorize and model the demand tendency trajectory curve, converting the demand change information contained in the curve into a set of vectors. Each vector represents a demand tendency and its intensity, thereby obtaining a set of user demand tendency vectors. For example, the vector (0.8, 0.3, 0.1) may represent the demand tendency intensities of the customer for product quality, price, and appearance respectively. Through this series of complex and systematic analysis steps, the demand tendency can be accurately mined from the user emotion change trajectory map, providing strong data support for subsequent dynamic service matching and personalized service recommendation solutions, helping merchants better meet customer needs and improve user experience and business benefits.

[0036] In a specific embodiment, the dynamic weight adjustment of the situation demand mapping matrix based on the demand hierarchy feature matrix to obtain a dynamic demand weight matrix includes: Perform hierarchical weight allocation on the demand hierarchy feature matrix to obtain a demand hierarchy weight distribution table, and perform multi-dimensional correlation analysis on the demand hierarchy weight distribution table to obtain a demand hierarchy correlation matrix; Based on the demand hierarchy correlation matrix, perform situation adaptability analysis on the situation demand mapping matrix to obtain a situation demand adaptation matrix, and perform dynamic weight correction on the situation demand adaptation matrix to obtain a demand dynamic adaptation weight table; Perform time-series feature extension on the demand dynamic adaptation weight table to obtain a dynamic demand weight matrix, and perform time-series trend analysis based on the dynamic demand weight matrix to obtain a demand weight evolution trajectory graph; Based on the demand weight evolution trajectory graph, perform normalization processing on the dynamic demand weight matrix to obtain a weight normalization matrix, and perform multi-dimensional feature fusion on the weight normalization matrix to obtain a dynamic demand weight matrix.

[0037] Specifically, in the process of dynamically adjusting the weights of the scenario demand mapping matrix based on the demand hierarchy feature matrix to obtain the dynamic demand weight matrix, the first step is to assign hierarchical weights to the demand hierarchy feature matrix. Different levels such as core demands, secondary demands, and additional demands divided in the demand hierarchy feature matrix are assigned corresponding weights according to their importance to form a demand hierarchy weight distribution table. For example, in the intelligent retail scenario, assume that the customer's demand for product quality belongs to the core demand and is assigned a weight of 0.6; the demand for product appearance is a secondary demand and is assigned a weight of 0.3; the demand for gifts is an additional demand and is assigned a weight of 0.1. Next, a multi-dimensional correlation analysis is performed on the demand hierarchy weight distribution table to consider the mutual influence and degree of association between different levels of demands, thereby obtaining a demand hierarchy correlation matrix. For example, there may be a certain positive correlation between product quality and appearance demands. If the quality meets the standard, a good appearance will further enhance the customer's purchase intention, and this correlation will be reflected in the matrix in numerical form. For example, the correlation coefficient between the two is 0.4. Based on the demand hierarchy correlation matrix, a scenario adaptability analysis is carried out on the scenario demand mapping matrix. Combining the emotional changes, behavior trends, and demand hierarchy characteristics of users in different shopping scenarios, the adaptability of various demands in different scenarios is judged to obtain a scenario demand adaptability matrix. For example, in the new product display area, the adaptability of the customer's demand for product innovation is relatively high, and the corresponding position in the matrix may be marked as 0.8; while in the promotion area, the adaptability of the customer's demand for price discounts is higher, marked as 0.9. Then, the weights of the scenario demand adaptability matrix are dynamically corrected according to the actual situation and data changes, comprehensively considering factors such as the market environment and user feedback, to obtain a demand dynamic adaptability weight table. Assume that the market competition is fierce during a certain period, and customers are more sensitive to prices. The original weight of 0.9 for the demand for price discounts in the promotion area may be adjusted to 0.95. To more comprehensively reflect the changes in demand weights, the demand dynamic adaptability weight table is extended with time series characteristics, incorporating the time dimension and recording the changes in demand weights at different time points, thereby obtaining a dynamic demand weight matrix. During different time periods of customer shopping, such as morning, afternoon, and evening, the demand weights may be different, and the matrix will fully present these changes. Based on the dynamic demand weight matrix, a time series trend analysis is performed to draw a demand weight evolution trajectory graph, intuitively showing the change trend of demand weights over time. For example, during the shopping peak period, the demand weight for service efficiency of customers may gradually increase, and the trajectory graph can clearly show this trend. Finally, based on the demand weight evolution trajectory graph, the dynamic demand weight matrix is normalized so that the weight values in different dimensions are in the same order of magnitude for easy analysis and comparison, obtaining a weight normalization matrix. For example, different demand weights originally in the range of 0-1 are adjusted to the range of 0-0.5 through normalization processing.Perform multi-dimensional feature fusion on the weight normalization matrix, integrate features from multiple dimensions such as time, context, and hierarchy of needs, and finally obtain a dynamic demand weight matrix. In this matrix, each element represents the weight of a certain level of need in a specific context and at a specific time, providing key data support for subsequent demand tendency analysis and accurate personalized service recommendation. Through this series of rigorous and interlocking operations, the demand weight can be dynamically adjusted according to the changes in the user's needs in different shopping contexts and times, and the user's demand tendency can be grasped more accurately.

[0038] In a specific embodiment, the dynamic service matching of the target user based on the user demand tendency vector set to obtain a personalized service recommendation scheme includes: Perform multi-scale tensor decomposition on the user demand tendency vector set to obtain a demand dimension component matrix, and perform cross-modal correlation analysis on the demand dimension component matrix to obtain a demand-service coupling map; Based on the demand-service coupling map, perform service resource topological mapping on the demand dimension component matrix to obtain a service resource matching network, and perform dynamic manifold analysis on the service resource matching network to obtain a service scheduling trajectory map; Extract spatio-temporal features from the service scheduling trajectory map to obtain a service response feature surface, and perform multi-objective optimization calculation based on the service response feature surface to obtain a service combination strategy matrix; Based on the service combination strategy matrix, perform personalized feature fusion to obtain a service recommendation feature cube, and perform dynamic pruning optimization on the service recommendation feature cube to obtain a personalized service recommendation scheme, where the personalized service recommendation scheme includes main service content, secondary service combination, and service execution timing.

[0039] Specifically, when performing dynamic service matching for a target user based on the user demand tendency vector set to obtain a personalized service recommendation solution, it is first necessary to perform multi-scale tensor decomposition on the user demand tendency vector set. The user demand tendency vector set contains multi-dimensional and multi-level demand information. Through multi-scale tensor decomposition, it is split into more detailed demand dimension component matrices, which respectively present the specific content and intensity of user demands from different perspectives. For example, in the intelligent retail scenario, the decomposed matrices may contain detailed information on dimensions such as product function requirements, price requirements, and appearance requirements. Then, cross-modal correlation analysis is performed on the demand dimension component matrices to correlate user demands with different modal information of service resources, such as product introduction videos, graphic and text descriptions, and customer service consultations, to obtain a demand-service coupling map. This map intuitively shows the degree of association between user demands and various service resources. For example, the map shows that the correlation between a user's performance requirements for a certain electronic product and the product technical parameter video reaches 0.8. Based on the demand-service coupling map, a service resource topological mapping is performed on the demand dimension component matrices to establish a topological connection relationship between each dimension of user demands and specific services in the service resource library, forming a service resource matching network. In this network, each node represents a service resource, and the edge represents the matching relationship between user demands and service resources. For example, for a user's comfort requirement for sports equipment, the service resource matching network will connect to relevant service nodes such as material introduction and fitting experience reservation. Then, dynamic manifold analysis is performed on the service resource matching network, considering the dynamic changes in user demands and the real-time status of service resources, simulating the flow path of services in the network, and obtaining a service scheduling trajectory map. For example, when user demands change, or some service resources are temporarily unavailable, the trajectory map will show how the service scheduling is adjusted to ensure that the service can respond to user demands in a timely manner. To more accurately grasp the adaptability between services and user demands, spatio-temporal feature extraction is performed on the service scheduling trajectory map, integrating the features of the time dimension and the space dimension (the locations of different service resources, the scenarios where the user is located, etc.), to obtain a service response feature surface. This surface comprehensively reflects the response of services to user demands under different spatio-temporal conditions. Based on the service response feature surface, multi-objective optimization calculations are performed, comprehensively considering multiple objectives such as service quality, cost, and user satisfaction, to obtain a service combination strategy matrix. For example, while meeting the user's demand for fast delivery of a certain product, it is also necessary to control the delivery cost. Through optimization calculations, the best delivery service combination strategy is obtained, and the matrix will record the target values corresponding to different service combinations, such as a service quality score of 85 points and the cost controlled within 90% of the budget. Based on the service combination strategy matrix, personalized feature fusion is performed, combining the user's personalized demand features, such as personal preferences and historical purchase records, with the service combination strategy, to obtain a service recommendation feature cube. This cube describes the service recommendation solution suitable for the user from multiple dimensions.For example, in combination with user preferences such as favorite brands and colors, the service combination strategy is further optimized. Finally, dynamic pruning optimization is performed on the service recommendation feature cube to remove service options that do not meet user needs or are less efficient, resulting in a personalized service recommendation solution. This solution includes the main service content, such as the recommendation of the goods most needed by the user; the secondary service combination, such as matching goods, after-sales service, etc.; and the service execution timing sequence, which clarifies the sequence and time nodes of each service. For example, in the solution developed for users purchasing electronic products, the main service is to recommend specific models of products, the secondary service combination includes accessory recommendation and extended warranty service, and the service execution timing sequence is to first display the product details, then make accessory recommendations, and finally introduce the extended warranty service, so as to provide users with a complete and personalized service recommendation solution that meets their needs.

[0040] In a specific embodiment, the service resource topology mapping of the demand dimension component matrix based on the demand-service coupling graph to obtain a service resource matching network includes: Perform tensor product decomposition on the demand-service coupling graph to obtain a coupling relation basis vector group, and perform eigenvalue decomposition on the demand dimension component matrix to obtain a demand eigen-subspace matrix; Perform vector space mapping on the demand eigen-subspace matrix based on the coupling relation basis vector group to obtain a demand-service joint feature embedding matrix; Perform local neighborhood weighting processing on the demand-service joint feature embedding matrix through a Gaussian kernel function to obtain a weighted neighborhood feature matrix, and perform Delaunay triangulation on the weighted neighborhood feature matrix to obtain an initial graph of the service resource topology structure; Perform topological constraint projection on the demand-service joint feature embedding matrix based on the initial graph of the service resource topology structure to obtain a topological constraint feature matrix, and perform minimum spanning tree construction on the topological constraint feature matrix to obtain a core connection network of service resources; Perform multi-scale dilation operations on the core connection network of service resources to obtain an extended connection graph of service resources, and perform topological connectivity analysis based on the extended connection graph of service resources to obtain a service resource matching network.

[0041] Specifically, in the process of mapping the demand dimension component matrix to service resource topology based on the demand-service coupling graph to obtain the service resource matching network, the demand-service coupling graph must first be subjected to tensor integral decomposition. The demand-service coupling graph records the complex association between user demand and service resources. Through tensor integral decomposition, these complex relationships can be decomposed into more basic coupling relationship basis vector groups, which are the basic units that constitute the coupling relationship. At the same time, the demand dimension component matrix is ​​subjected to eigenvalue decomposition to mine the potential key features in the demand dimension and obtain the demand feature subspace matrix, which represents the core feature distribution of user demand in different dimensions. For example, in the smart retail scenario, after the user's demand dimension component matrix for a certain smart watch is decomposed, subspace matrices such as functional requirements, appearance requirements, and price requirements may be obtained, while the basis vector group decomposed from the demand-service coupling graph covers the basic association relationship between product introduction, trial service, and preferential activities and demand. Based on the obtained coupling relationship basis vector group, the demand feature subspace matrix is ​​subjected to vector space mapping. The demand feature vectors in the demand feature subspace matrix are mapped to the vector space containing service resource information according to the association rules defined by the coupling relationship basis vector group, thereby obtaining the demand-service joint feature embedding matrix. This matrix integrates user demand features and service resource features in the same space, making the association between the two more intuitive and clear. For example, after mapping, the user's demand features for long battery life of smart watches are closely associated with service resource features such as battery technology introduction videos and battery life test reports in the joint feature embedding matrix, and the correlation may reach 0.7. Then, the demand-service joint feature embedding matrix is ​​subjected to local neighborhood weighting processing by the Gaussian kernel function. The Gaussian kernel function can assign different weights to adjacent feature vectors according to the distance between feature vectors. The closer the distance, the higher the weight of the feature vector, thereby obtaining a weighted neighborhood feature matrix. In this way, the service resource features that are closely related to user needs can be highlighted. The weighted neighborhood feature matrix is ​​subjected to Delaunay triangulation, and the feature vectors are connected into triangular meshes in space to form an initial diagram of the service resource topology structure. In this initial diagram, each node represents a service resource or user demand feature, and the edge represents the connection relationship between them. For example, in the scenario of smart watches, service resource nodes such as product function demonstration videos and user reviews are connected with user function demand nodes through triangulation to construct a preliminary topological structure. Based on the initial graph of the service resource topological structure, the demand-service joint feature embedding matrix is ​​topologically constrained. The feature vectors in the joint feature embedding matrix are projected according to the topological structure defined by the initial graph to obtain a topological constraint feature matrix, so that the distribution of feature vectors is more consistent with the actual association structure between service resources and user needs.Then, construct a minimum spanning tree for the topological constraint feature matrix to find a tree structure that can connect all key service resources and demand feature nodes with the minimum sum of edge weights, obtaining the core connection network of service resources. This network ensures the efficiency and necessity of the key connections between service resources and user needs. For example, in the scenario of purchasing a smartwatch, the core connection network will focus on connecting key service resources such as product core function introductions and price advantage explanations with the corresponding core needs of users. Finally, perform a multi-scale dilation operation on the core connection network of service resources. Based on the core connection network, gradually expand to connect more service resource nodes and association relationships to obtain the extended connection graph of service resources. By performing topological connectivity analysis on the extended connection graph of service resources, check whether the connections between each service resource node are smooth, whether there are redundancies or breaks, and finally obtain the service resource matching network. In this network, the complete matching relationship between user needs and various service resources is clearly presented, providing an accurate basis for subsequent dynamic service scheduling and personalized service recommendation. For example, for users purchasing smartwatches, the service resource matching network will comprehensively cover the matching relationships between a series of service resources such as product consultation, purchase process guidance, and after-sales repair services and user needs, ensuring that users can obtain services that meet their needs throughout the shopping process.

[0042] The above described the method for analyzing user needs based on AI in the embodiments of the present invention. Next, the system for analyzing user needs based on AI in the embodiments of the present invention will be described. Please refer to Figure 2 , an embodiment of the system for analyzing user needs based on AI in the embodiments of the present invention includes: An acquisition module 21, configured to continuously collect videos of the facial expressions of a target user through a video acquisition device to obtain a target user video sequence; An extraction module 22, configured to extract emotional feature of the target user based on the target user video sequence to obtain a user emotional state feature set; A first analysis module 23, configured to perform emotional pattern time series analysis on the target user based on the user emotional state feature set to obtain a user emotional change trajectory map; A second analysis module 24, configured to perform demand tendency analysis on the target user based on the user emotional change trajectory map to obtain a user demand tendency vector set; A matching module 25, configured to perform dynamic service matching on the target user based on the user demand tendency vector set to obtain a personalized service recommendation plan.

[0043] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to that described in the above method embodiment, and details will not be repeated here.

[0044] Refer toFigure 3 , an embodiment of the present invention further provides a computer device, and its internal structure may be as Figure 3 shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements the above method.

[0045] Those skilled in the art can understand that Figure 3 the structure shown in

[0046] is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0047] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0048] It should be noted that in this document, the terms "including", "comprising", or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that includes a series of elements includes not only those elements but also other elements not expressly listed, or elements that are inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, apparatus, article, or method that includes the element.

[0049] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall equally be included in the patent protection scope of the present invention.

Claims

1. An AI-based user requirement analysis method, characterized in that, Including the following steps: Continuously collect videos of the facial expressions of the target user through a video acquisition device to obtain a target user video sequence; Extract emotional features of the target user based on the target user video sequence to obtain a user emotional state feature set; Conduct a temporal analysis of the emotional pattern of the target user based on the user emotional state feature set to obtain a user emotional change trajectory map; Conduct a demand tendency analysis of the target user based on the user emotional change trajectory map to obtain a user demand tendency vector set; Conduct dynamic service matching for the target user based on the user demand tendency vector set to obtain a personalized service recommendation plan.

2. The AI-based user requirement analysis method according to claim 1, wherein The extracting of the emotional features of the target user based on the target user video sequence to obtain a user emotional state feature set includes: Locate and track the face region of the target user video sequence to obtain a facial feature point sequence map, and conduct a micro-expression dynamic analysis on the facial feature point sequence map to obtain a facial muscle deformation matrix; Conduct an expression semantic analysis based on the facial muscle deformation matrix to obtain a facial emotional intensity map, and extract time-domain features from the facial emotional intensity map to obtain an emotional fluctuation curve set; Conduct a frequency-domain decomposition operation on the emotional fluctuation curve set to obtain an emotional frequency spectrum feature map, and conduct an emotional periodicity analysis based on the emotional frequency spectrum feature map to obtain an emotional rhythm feature vector; Conduct multi-dimensional feature combination based on the emotional rhythm feature vector to obtain a user emotional state feature set, where the user emotional state feature set includes basic emotional categories, emotional intensity levels, and emotional duration stability.

3. The AI-based user requirement analysis method according to claim 1, wherein The conducting of the temporal analysis of the emotional pattern of the target user based on the user emotional state feature set to obtain a user emotional change trajectory map includes: Segment the user emotional state feature set by time window to obtain an emotional state sequence matrix, and calculate the temporal correlation of the emotional state sequence matrix to obtain an emotional state transition probability map; Conduct state link tracking based on the emotional state transition probability map to obtain an emotional evolution path set, and conduct multi-dimensional projection on the emotional evolution path set to obtain an emotional trajectory projection map; Extract feature points from the emotional trajectory projection map to obtain a key emotional node sequence, and conduct trajectory segmentation based on the key emotional node sequence to obtain an emotional stage division map; Conduct trajectory reconstruction based on the emotional stage division map to obtain an emotional change contour curve, and conduct curvature analysis on the emotional change contour curve to obtain an emotional change feature vector; Conduct spatio-temporal dimension fusion on the emotional change feature vector to obtain an emotional trajectory feature surface, and conduct multi-dimensional feature combination based on the emotional trajectory feature surface to obtain a user emotional change trajectory map, where the user emotional change trajectory map includes emotional development trends, emotional fluctuation patterns, and emotional conversion rules.

4. The AI-based user requirement analysis method according to claim 1, wherein The conducting of the demand tendency analysis of the target user based on the user emotional change trajectory map to obtain a user demand tendency vector set includes: Perform multi-dimensional feature aggregation on the user's emotional change trajectory map to obtain an emotion-behavior association matrix, and perform behavior decision feature extraction on the emotion-behavior association matrix to obtain a set of behavior trend features; Based on the set of behavior trend features, perform context matching analysis on the user's emotional change trajectory map to obtain a context demand mapping matrix, and perform constraint condition analysis on the context demand mapping matrix to obtain a set of context demand features; Perform multi-dimensional decomposition on the set of context demand features to obtain a demand priority distribution map, and perform hierarchical processing on the set of context demand features based on the demand priority distribution map to obtain a demand hierarchy feature matrix; Based on the demand hierarchy feature matrix, perform dynamic weight adjustment on the context demand mapping matrix to obtain a dynamic demand weight matrix, and perform time series analysis on the dynamic demand weight matrix to obtain a demand tendency trajectory curve; Perform vectorization modeling on the demand tendency trajectory curve to obtain a set of user demand tendency vectors.

5. The AI-based user requirement analysis method according to claim 4, wherein The performing dynamic weight adjustment on the context demand mapping matrix based on the demand hierarchy feature matrix to obtain a dynamic demand weight matrix includes: Perform hierarchical weight assignment on the demand hierarchy feature matrix to obtain a demand hierarchy weight distribution table, and perform multi-dimensional correlation analysis on the demand hierarchy weight distribution table to obtain a demand hierarchy correlation matrix; Based on the demand hierarchy correlation matrix, perform context adaptability analysis on the context demand mapping matrix to obtain a context demand adaptation matrix, and perform dynamic weight correction on the context demand adaptation matrix to obtain a demand dynamic adaptation weight table; Perform time series feature extension on the demand dynamic adaptation weight table to obtain an initial dynamic demand weight matrix, and perform time series trend analysis on the initial dynamic demand weight matrix to obtain a demand weight evolution trajectory map; Based on the demand weight evolution trajectory map, perform normalization processing on the dynamic demand weight matrix to obtain a weight normalization matrix, and perform multi-dimensional feature fusion on the weight normalization matrix to obtain a dynamic demand weight matrix.

6. The AI-based user requirement analysis method according to claim 1, wherein The performing dynamic service matching on the target user based on the set of user demand tendency vectors to obtain a personalized service recommendation scheme includes: Perform multi-scale tensor decomposition on the set of user demand tendency vectors to obtain a demand dimension component matrix, and perform cross-modal correlation analysis on the demand dimension component matrix to obtain a demand-service coupling map; Based on the demand-service coupling map, perform service resource topology mapping on the demand dimension component matrix to obtain a service resource matching network, and perform dynamic manifold analysis on the service resource matching network to obtain a service scheduling trajectory map; Perform spatio-temporal feature extraction on the service scheduling trajectory map to obtain a service response feature surface, and perform multi-objective optimization calculation based on the service response feature surface to obtain a service combination strategy matrix; Based on the service combination strategy matrix, personalized feature fusion is performed to obtain a service recommendation feature cube, and dynamic pruning optimization is performed on the service recommendation feature cube to obtain a personalized service recommendation scheme, where the personalized service recommendation scheme includes main service content, secondary service combination, and service execution timing sequence.

7. The AI-based user requirement analysis method according to claim 6, wherein The service resource topology mapping of the demand dimension component matrix based on the demand-service coupling graph includes: Performing tensor product decomposition on the demand-service coupling graph to obtain a coupling relationship basis vector group, and performing eigenvalue decomposition on the demand dimension component matrix to obtain a demand feature subspace matrix; Performing vector space mapping on the demand feature subspace matrix based on the coupling relationship basis vector group to obtain a demand-service joint feature embedding matrix; Performing local neighborhood weighting processing on the demand-service joint feature embedding matrix through a Gaussian kernel function to obtain a weighted neighborhood feature matrix, and performing Delaunay triangulation on the weighted neighborhood feature matrix to obtain an initial graph of the service resource topology structure; Performing topological constraint projection on the demand-service joint feature embedding matrix based on the initial graph of the service resource topology structure to obtain a topological constraint feature matrix, and constructing a minimum spanning tree for the topological constraint feature matrix to obtain a core connection network of service resources; Performing multi-scale dilation operations on the core connection network of service resources to obtain an extended connection graph of service resources, and performing topological connectivity analysis based on the extended connection graph of service resources to obtain a service resource matching network.

8. An AI-based user requirement analysis system, characterized in that, Including: An acquisition module, configured to continuously acquire videos of the facial expressions of a target user through a video acquisition device to obtain a target user video sequence; An extraction module, configured to extract emotional features of the target user based on the target user video sequence to obtain a user emotional state feature set; A first analysis module, configured to perform emotional pattern timing analysis on the target user based on the user emotional state feature set to obtain a user emotional change trajectory graph; A second analysis module, configured to perform demand tendency analysis on the target user based on the user emotional change trajectory graph to obtain a user demand tendency vector set; A matching module, configured to perform dynamic service matching on the target user based on the user demand tendency vector set to obtain a personalized service recommendation scheme.

9. A computer device, comprising a memory and a processor, wherein a computer program is stored in the memory, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images