System switching method and system for a tutoring learning machine
By obtaining user facial images and behavior data, dynamic behavior analysis and interest tendency mining, and generating user portraits, the problem of low switching efficiency of traditional tutoring learning machine systems is solved, and intelligent system switching and personalized content adaptation is realized.
Patent Information
- Application Number
- CN202311611353.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-28
- Publication Date
- 2025-07-22
- Estimated Expiration
- 2043-11-28
AI Technical Summary
The system switching of traditional tutoring learning machines relies on subjective methods, resulting in low switching efficiency and unclear content targeted, which cannot meet the intelligent needs of modern learning machines.
By obtaining user facial images and behavioral data, dynamic behavior analysis is carried out, mood fluctuation curves and eye trajectory diagrams are constructed, interest tendency analysis and personality differences analysis are used for deep learning algorithms, user portraits are generated, and system switching decisions are made based on behavior demand prediction evaluation index.
It realizes the intelligent system switching of the tutoring learning machine, improves the switching efficiency and personalized adaptability of content, and meets the user's learning needs and interests.
Smart Images

Figure CN117666786B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of tutoring learning machines, and particularly to a system switching method and system for a tutoring learning machine. Background Art
[0002] The main function of the tutoring learning machine system is to adjust teaching content and methods in a timely manner according to the student's schoolwork situation to achieve the effect of personalized teaching. With the development of artificial intelligence technology, how to use technologies such as machine learning to achieve intelligent switching between tutoring learning machine systems has become an important issue. The system switching of traditional tutoring learning machines often relies on subjective switching, resulting in problems such as low switching efficiency and unclear content targeting. Therefore, in order to meet the intelligent development of modern learning machines, an automated and intelligent system switching method and system for tutoring learning machines are needed. Summary of the Invention
[0003] In order to solve the above technical problems, the present invention proposes a system switching method and system for a tutoring learning machine to solve at least one of the above technical problems.
[0004] To achieve the above object, the present invention provides a system switching method for a tutoring learning machine, including the following steps:
[0005] Step S1: Obtain a user's facial image and user behavior data; perform dynamic behavior analysis on the user behavior data to generate dynamic behavior feature data; perform emotional fluctuation analysis on the user's facial image according to the dynamic behavior feature data to construct a user emotional fluctuation curve;
[0006] Step S2: Perform learning state detection on the dynamic behavior feature data through the user emotional fluctuation curve to generate learning state data; perform eye movement trajectory recognition on the user's facial image according to the learning state data to construct an eye movement trajectory map;
[0007] Step S3: Use the eye movement trajectory map to perform implicit association mining on the user emotional fluctuation curve to generate user behavior association data; perform user interest tendency analysis on the dynamic behavior feature data based on the user behavior association data to generate user interest tendency data; perform behavior tendency evolution processing on the user interest tendency data to construct a behavior interest tendency map;
[0008] Step S4: Perform personality difference analysis on the dynamic behavior feature data using the behavior interest tendency map to generate user personality difference data; construct a user portrait based on the user personality difference data to generate a user portrait; use the behavior interest tendency map to perform real-time dynamic optimization on the user portrait to generate a user dynamic portrait;
[0009] Step S5: Conduct emotional trend analysis on the user's emotional fluctuation curve through the user's dynamic portrait to generate user emotional trend data; conduct behavioral demand prediction on the dynamic behavioral characteristic data through the user emotional trend data to generate a behavioral demand prediction evaluation index;
[0010] Step S6: Compare the behavioral demand prediction evaluation index based on a preset behavioral demand prediction evaluation threshold index to generate interest tendency optimization data; use a deep learning algorithm to conduct interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine to execute system switching operations.
[0011] By obtaining the user's facial image and behavioral data, the present invention can acquire the user's facial expressions and behavioral information for subsequent analysis and modeling. Dynamic behavior analysis can deeply understand the user's behavior patterns and habits, including characteristics in aspects such as actions, postures, and languages, providing a comprehensive understanding of the user's behavior. Based on the dynamic behavior characteristic data, emotional fluctuation analysis of the user's facial image can infer the user's emotional state and fluctuations. Learning state detection can determine the user's current learning state, such as concentration, distraction, fatigue, etc., based on the user's emotional fluctuation curve, providing a basis for subsequent learning process analysis and optimization. Through eye movement trajectory recognition, eye movement characteristics such as the user's fixation points, fixation times, and fixation orders can be observed, providing data support for subsequent learning behavior analysis and cognitive process research. Through mining the implicit associations of the eye movement trajectory map, the potential relationship between the user's emotional fluctuation curve and eye movements can be discovered, revealing the connection between the user's emotions and cognition. According to the user's behavior association data, the user's interest tendencies and preferences can be analyzed, understanding the user's behavior patterns and interest areas, providing a basis for personalized recommendations and customized services. Behavioral tendency evolution processing can track the changes and evolution process of the user's interests, constructing a behavioral interest tendency map, providing support for the long-term analysis and prediction of the user's interests. The behavioral interest tendency map can reveal the personality differences and differences in behavior patterns among users. Through personality difference analysis of the dynamic behavior characteristic data, a basis can be provided for the user's personalized services and recommendation systems. The construction of the user portrait can comprehensively consider factors such as the user's interests, preferences, and behavior patterns, integrating the user's characteristics into a comprehensive description for better understanding and meeting the user's needs. The real-time dynamic optimization of the behavioral interest tendency map can update and adjust the user portrait in a timely manner according to the user's behavior and interest changes, maintaining the accuracy and practicality of the user portrait. Emotional trend analysis can use the user's dynamic portrait to analyze the long-term changes and trends of the user's emotional fluctuation curve, helping to understand the evolution and development of the user's emotions. Through the user's emotional trend data, behavioral demand prediction of the dynamic behavior characteristic data can predict the user's future behavioral tendencies and demands, providing a decision-making basis for personalized recommendations and services. By comparing the behavioral demand prediction evaluation index with the preset threshold index, it can be determined whether the user's interest tendency meets the expectations, thereby generating interest tendency optimization data for optimizing personalized recommendation and service strategies. Using deep learning algorithms to analyze and model the interest tendency switching data and interest tendency optimization data can generate a user interest switching decision engine for automatically executing system switching tasks to better meet the user's interests and needs.
[0012] Preferably, step S1 includes the following steps:
[0013] Step S11: Obtain the user's facial image and the user's behavioral data;
[0014] Step S12: Perform time series analysis on the user behavior data to generate behavior time series data;
[0015] Step S13: Perform behavior pattern analysis on the user behavior data according to the behavior time series data to generate user behavior pattern data;
[0016] Step S14: Perform dynamic behavior analysis on the user behavior pattern data to generate dynamic behavior feature data;
[0017] Step S15: Detect micro-expression features of the user's facial image to generate user micro-expression feature data;
[0018] Step S16: Perform emotional fluctuation analysis on the user micro-expression feature data according to the dynamic behavior feature data to construct a user emotional fluctuation curve.
[0019] By obtaining the user's facial image and behavior data, the present invention can obtain the user's facial expressions and behavior information, providing a data basis for subsequent analysis and modeling. Time series analysis can deeply understand the user's behavior patterns and habits, including features in aspects such as actions, postures, and languages, providing a comprehensive understanding of the user's behavior. Behavior pattern analysis can identify and extract the user's common behavior patterns, helping to understand the user's preferences and habits, and providing a basis for subsequent personalized services and recommendations. Dynamic behavior analysis can deeply analyze the user's behavior patterns, revealing the dynamic changes and trends of the user's behavior, and providing a basis for personalized services and behavior prediction. Micro-expression feature detection can capture the changes in the user's facial micro-expressions, revealing the subtle fluctuations of the user's emotions and feelings. Emotional fluctuation analysis can infer the user's emotional state and fluctuation conditions, correlating the user's behavior and emotions, and providing an emotional basis and a reference for personalized adjustment for the system switch of the learning machine.
[0020] Preferably, step S16 includes the following steps:
[0021] Step S161: Identify the position points of the facial features of the user's facial image to obtain the user's facial feature position data;
[0022] Step S162: Detect the facial muscle distortion frequency of the user's facial image based on the user's facial feature position data to generate facial muscle distortion data;
[0023] Step S163: Perform emotional intensity quantization processing on the user micro-expression feature data through the facial muscle distortion data to generate an emotional intensity quantization index;
[0024] Step S164: Perform emotional fluctuation analysis on the emotional intensity quantization index according to the dynamic behavior feature data to generate the user's emotional fluctuation law;
[0025] Step S165: Fit an emotional fluctuation curve to the emotional intensity quantization index using the user's emotional fluctuation law to construct a user's emotional fluctuation curve.
[0026] Through the present invention, by identifying the position points of the facial features in the user's facial image, the positions of the respective feature points of the user's face can be accurately determined, providing basic data for subsequent detection of the facial muscle distortion frequency and emotional analysis. The detection of the facial muscle distortion frequency can analyze the degree and frequency of the distortion of the user's facial muscles, capture the minute movement changes of the facial muscles, providing a data basis for emotional analysis and emotional intensity quantization. The emotional intensity quantization process can convert the user's micro-expression feature data into a quantization index of emotional intensity, describing the strength of the user's emotions, providing basic data for emotional fluctuation analysis and the construction of an emotional fluctuation curve. The emotional fluctuation analysis can analyze the fluctuation conditions and trends of the user's emotions according to the emotional intensity quantization index, revealing the ups and downs of the user's emotions. By fitting and modeling the emotional intensity quantization index, an emotional fluctuation curve of the user individual can be constructed, describing the variation law of the user's emotions over time, providing a more accurate emotional prediction and adjustment basis for the system switching and personalized adaptation of the learning machine.
[0027] Preferably, step S2 includes the following steps:
[0028] Step S21: Detect the learning state of the dynamic behavior feature data through the user's emotional fluctuation curve to generate learning state data;
[0029] Step S22: Identify the eye movement trajectory of the user's facial image according to the learning state data to generate eye movement trajectory data;
[0030] Step S23: Analyze the fixation frequency of the eye movement trajectory data to generate fixation frequency data;
[0031] Step S24: Analyze the fixation duration of the user's facial image according to the fixation frequency data to generate fixation duration data;
[0032] Step S25: Analyze the blink frequency of the eye movement trajectory data based on the fixation duration data to generate blink frequency data;
[0033] Step S26: Detect the range of eye movement of the eye movement trajectory data according to the blink frequency data to generate range-of-eye-movement data;
[0034] Step S27: Perform trajectory visualization of the range-of-eye-movement data according to the eye movement trajectory data to construct an eye movement trajectory map.
[0035] Through the analysis of the user's emotional fluctuation curve, the present invention can detect the user's learning state, such as concentration, excitement, etc., providing a basis for subsequent learning behavior analysis and personalized adjustment. Eye movement trajectory recognition can track the user's eye movement trajectory, including fixation points and fixation paths, providing a data basis for subsequent fixation frequency analysis and eye movement range detection. Fixation frequency analysis can calculate the user's fixation frequency within a specific time period, that is, the change frequency of fixation points, helping to understand the user's attention level and attention allocation to different learning contents. Fixation duration analysis can measure the length of time the user stays at different fixation points, revealing the user's degree of attention and interest points for learning contents, providing a basis for personalized recommendation and adaptive adjustment of learning contents. Blink frequency analysis can calculate the user's blink frequency, understanding the fatigue degree of the user's eyes and the change of attention. Eye movement range detection can analyze the movement range of the user's eyes on the screen, understanding the user's visual focus and attention concentration, providing a reference for the layout and interface design of learning contents. By visualizing the eye movement range data, the user's line-of-sight movement trajectory can be intuitively displayed, helping to understand the attention distribution and visual attention pattern of the user during the learning process.
[0036] Preferably, the specific steps of step S3 are as follows:
[0037] Step S31: Use the eye movement trajectory map to conduct implicit association mining on the user's emotional fluctuation curve to generate user behavior association data;
[0038] Step S32: Calculate the similarity of the dynamic behavior feature data based on the user behavior association data to generate dynamic behavior similarity data;
[0039] Step S33: Use the dynamic behavior similarity data to conduct user interest tendency analysis on the dynamic behavior feature data to generate user interest tendency data;
[0040] Step S34: Identify the interest points from the user interest tendency data to generate user interest point data;
[0041] Step S35: Conduct behavior tendency evolution processing on the user interest tendency data through the user interest point data to construct a behavior interest tendency map.
[0042] By associatively mining the eye movement trajectory map and the user's emotional fluctuation curve, the present invention can discover the potential correlation between the eye movement trajectory and the emotional fluctuation, providing data support for subsequent behavior analysis and personalized recommendation. By calculating the similarity of the user behavior correlation data, the similarity degree between different behaviors can be quantified, providing a basis for understanding the user's behavior pattern and behavior preference, so as to better perform personalized recommendation and learning content adaptation. By analyzing the dynamic behavior similarity data, the user's interest tendency and behavior preference can be revealed, and the preference degree of the user for different learning contents and tasks can be understood. Interest point recognition can classify and label the user's interest tendency, identify and extract the user's interest fields and concerns, providing basic data for the personalized recommendation and content filtering of the learning machine. The evolution processing of behavior tendency can analyze the change trend and evolution law of the user's interest points, construct a map of the user's behavior interest tendency, and reveal the evolution path of the user's interest change and learning preference.
[0043] Preferably, the specific steps of step S4 are as follows:
[0044] Step S41: Use the behavior interest tendency map to conduct a personalized preference analysis on the dynamic behavior feature data to generate personalized preference data;
[0045] Step S42: Conduct a personality difference analysis on the personalized preference data to generate user personality difference data;
[0046] Step S43: Identify the interest driving force of the behavior interest tendency map according to the user personality difference data to generate interest driving force data;
[0047] Step S44: Construct a user portrait based on the interest driving force data for the user personality difference data to generate a user portrait;
[0048] Step S45: Use the behavior interest tendency map to conduct an interest feature time series weight analysis on the user portrait to generate interest time series weight data;
[0049] Step S46: Perform real-time dynamic iterative optimization on the user portrait according to the interest time series weight data to generate a user dynamic portrait.
[0050] By analyzing the behavioral interest tendency map and dynamic behavioral feature data, the present invention can deeply understand the user's personality preferences and learning interests, providing a basis for personalized recommendation and content customization of the learning machine to meet the user's personalized needs. Personality difference analysis can identify the personality differences and preference differences among users, helping to understand the learning preferences, behavioral habits and preference characteristics of different users, and providing a basis for personalized recommendation and customized learning strategies. Interest driving force identification can identify the driving forces and motivational factors behind the user's learning behavior according to the user's personality differences and interest preferences, providing a basis for personalized recommendation and learning motivation of the learning machine, and enhancing the user's learning motivation. Interest driving force identification can identify the driving forces and motivational factors behind the user's learning behavior according to the user's personality differences and interest preferences, providing a basis for personalized recommendation and learning motivation of the learning machine, and enhancing the user's learning motivation. User portrait construction can comprehensively consider the user's personality differences, interest driving forces and behavioral characteristics to form a comprehensive description and feature generalization of the user. Interest feature time-series weight analysis can analyze the change weights of the user's interests at different time periods, reflecting the user's attention and priority to different interests, and providing a reference basis for real-time recommendation and dynamic adjustment of the learning machine. Real-time dynamic iterative optimization can update and optimize the user portrait in real time according to the user's interest time-series weight data to adapt to the change of the user's interests and the dynamic adjustment of the learning needs, and provide more accurate personalized learning support.
[0051] Preferably, the specific steps of step S5 are as follows:
[0052] Step S51: Perform emotional cycle analysis on the user's emotional fluctuation curve through the user's dynamic portrait to generate emotional cycle data;
[0053] Step S52: Detect emotional mutation points on the user's emotional fluctuation curve through the emotional cycle data and mark the emotional mutation points;
[0054] Step S53: Perform emotional trend analysis on the user's emotional fluctuation curve according to the emotional mutation points to generate user emotional trend data;
[0055] Step S54: Perform demand trend prediction on the dynamic behavioral feature data through the user's emotional trend data to generate demand trend prediction data;
[0056] Step S55: Use the behavioral demand trend prediction probability calculation formula to perform probability prediction evaluation on the demand trend prediction data to generate a behavioral demand prediction evaluation index.
[0057] Through emotional cycle analysis, the present invention can identify the periodic patterns of users' emotional fluctuations, understand the periodic changes in users' emotions, and provide a basis for the emotional regulation and mood management of learning machines, so as to improve users' emotional stability and learning effects. The emotional mutation point detection can identify the mutation points in the users' emotional fluctuation curves, mark the important turning points in the emotional fluctuations, and help understand the changes and triggering factors of users' emotions, providing a basis for the emotional regulation, personalized feedback and intervention of learning machines. The emotional trend analysis can analyze the trend changes in the users' emotional fluctuation curves, identify the rising, falling or stable trends of users' emotions, and provide a basis for the emotional regulation and personalized feedback of learning machines to adapt to the emotional needs and emotional states of users. The demand trend prediction can predict the trend changes in users' learning needs based on users' emotional trends and behavioral characteristic data, and understand the degree of users' demand for different learning contents and resources. The behavioral demand prediction evaluation index can perform a probability evaluation on the demand trend prediction data, quantify the credibility and accuracy of the prediction results, and provide a reference for the decision-making and behavior adjustment of learning machines to improve the intelligence and personalized adaptation ability of learning machines.
[0058] Preferably, the calculation formula for the behavioral demand trend prediction probability in step S55 is specifically:
[0059]
[0060] Where P is the behavioral demand prediction evaluation index, t is the time point of demand prediction, X is the weight index of the predicted behavioral characteristics, Y is the user behavior adaptability parameter, Z is the behavioral demand change rate, S is the decay rate of behavioral demand over time, N is the emotional fluctuation amplitude value, d is the interest driving force index, and λ is the fixation duration of the interest point.
[0061] The present invention passes through By calculating the time derivative of the variance of the weight index of behavioral characteristics multiplied by the mean value of the user behavior adaptability parameter divided by the square of the natural logarithm of the behavioral demand change rate, the change rate of the behavioral demand trend can be understood, which can provide a more refined prediction result and take into account the change of behavioral demand over time. By combining the variance of the weight index of behavioral characteristics, the mean value of the user behavior adaptability parameter and the natural logarithm of the behavioral demand change rate and performing a square operation, the influence of these factors on the behavioral demand can be comprehensively considered, which helps to more accurately predict the behavioral demand trend and provide a more reliable evaluation index. By calculating the natural logarithm of the ratio of the decay rate of behavioral demand over time to the emotional fluctuation amplitude value plus 1, the influence of emotional fluctuations on behavioral demand can be considered, which helps to more comprehensively predict the behavioral demand trend and make the prediction result more accurate. Calculating the interest driving force index divided by the square root of the fixation duration of the interest point can comprehensively consider the user's interest driving force and the degree of continuous attention to the interest point. This is of great significance for the accuracy of predicting behavioral demand trends, making the prediction results more in line with the user's actual interests and behavioral needs. By performing an exponential operation on the negative value of the time derivative and converting the result into a probability value, the behavioral demand prediction evaluation index can be restricted between 0 and 1. This helps to intuitively understand the prediction probability of behavioral demands and the credibility evaluation of behavioral demand trends.
[0062] Preferably, the specific steps of step S6 are as follows:
[0063] Step S61: Compare the behavioral demand prediction evaluation index with a preset behavioral demand prediction evaluation threshold index. When the preset behavioral demand prediction evaluation threshold index is less than the behavioral demand prediction evaluation index, switch the user's interest page through the user dynamic portrait to generate interest tendency switching data;
[0064] Step S62: When the preset behavioral demand prediction evaluation threshold index is greater than or equal to the behavioral demand prediction evaluation index, optimize the interest tendency of the user dynamic portrait to generate interest tendency optimization data;
[0065] Step S63: Use a machine learning algorithm to perform interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine to execute the system switching operation.
[0066] By comparing the behavioral demand prediction evaluation index with the preset threshold index, the present invention can judge the degree of difference between the user's current behavioral demand and the expectation. When the behavioral demand prediction evaluation index is higher than the threshold index, it means that the user's behavioral demand exceeds the expectation. At this time, switching the user's interest page can provide learning content or resources that better meet the user's needs and enhance the personalized adaptation ability of the learning machine. When the behavioral demand prediction evaluation index is lower than or equal to the preset threshold index, it indicates that the user's behavioral demand is in line with or lower than the expectation. At this time, optimizing the interest tendency of the user dynamic portrait can further analyze and understand the user's interest preferences, provide more accurate user portrait data for the learning machine, and optimize personalized recommendation and learning path planning. By analyzing the interest tendency switching data and the interest tendency optimization data through a machine learning algorithm, the interest change trend and preferences of the user can be identified, providing a decision engine for the learning machine, thereby realizing automatic system switching. In this way, according to the user's behavioral needs and interest changes, the learning content, learning method, or learning environment of the learning machine can be adjusted in a timely manner to provide a better learning experience and learning effect.
[0067] In this specification, a system switching system for a tutoring learning machine is provided, including:
[0068] An emotional fluctuation module, configured to obtain a user's facial image and the user's behavior data; perform dynamic behavior analysis on the user's behavior data to generate dynamic behavior characteristic data; perform emotional fluctuation analysis on the user's facial image according to the dynamic behavior characteristic data to construct a user emotional fluctuation curve;
[0069] An eye movement trajectory module, configured to perform learning state detection on the dynamic behavior characteristic data through the user emotional fluctuation curve to generate learning state data; perform eye movement trajectory recognition on the user's facial image according to the learning state data to construct an eye movement trajectory map;
[0070] An interest tendency module, configured to perform implicit association mining on the user emotional fluctuation curve by using the eye movement trajectory map to generate user behavior association data; perform user interest tendency analysis on the dynamic behavior characteristic data based on the user behavior association data to generate user interest tendency data; perform behavior tendency evolution processing on the user interest tendency data to construct a behavior interest tendency map;
[0071] A dynamic portrait module, configured to perform personality difference analysis on the dynamic behavior characteristic data by using the behavior interest tendency map to generate user personality difference data; construct a user portrait according to the user personality difference data to generate a user portrait; perform real-time dynamic optimization on the user portrait by using the behavior interest tendency map to generate a user dynamic portrait;
[0072] A behavior demand prediction module, configured to perform emotional trend analysis on the user emotional fluctuation curve through the user dynamic portrait to generate user emotional trend data; perform behavior demand prediction on the dynamic behavior characteristic data through the user emotional trend data to generate a behavior demand prediction evaluation index;
[0073] A decision engine module, configured to compare the behavior demand prediction evaluation index based on a preset behavior demand prediction evaluation threshold index to generate interest tendency optimization data; perform interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data by using a deep learning algorithm to generate a user interest switching decision engine to execute a system switching operation.
[0074] The present invention constructs a system switching system for a tutor learning machine, obtains user facial images and behavior data, and performs dynamic behavior analysis to understand the user's emotional fluctuations and behavioral characteristics. Through the emotional fluctuation analysis, the user's emotional fluctuation curve can be constructed to better understand the user's emotional state and changing trend. By learning the user's emotional fluctuation curve and combining the dynamic behavior feature data, the eye trajectory module can perform learning state detection and eye trajectory recognition, which can help determine the user's attention, concentration and eye movement trajectory during the learning process, provide information about the user's learning state, and provide basic data for subsequent interest tendency analysis. Through the analysis of the eye trajectory graph, the interest tendency module can mine the implicit association between the user's emotional fluctuation curve and behavior, and generate user behavior association data. Based on these data, the user's interest tendency analysis can be performed on the dynamic behavior feature data to further understand the user's interest preferences and behavioral tendencies. By constructing a behavioral interest tendency map, the user's learning interest can be better understood, providing a basis for personalized learning. The dynamic portrait module uses the behavioral interest tendency map to personalize the dynamic behavior feature data. The user personality difference analysis generates user personality difference data, which helps to gain an in-depth understanding of the unique characteristics and behavior patterns of each user, thereby building a user portrait. By dynamically optimizing the user portrait in real time, it can better adapt to the user's individual needs and provide personalized learning experience and learning recommendations. The behavior demand prediction module uses the user's dynamic portrait to perform emotional trend analysis on the emotional fluctuation curve and generate user emotional trend data. Through the emotional trend data, the dynamic behavior feature data can be used to predict behavioral needs, further understand the user's learning needs and behavioral tendencies, and the generated behavioral demand prediction evaluation index can be provided to the subsequent decision engine for further processing. The decision engine module compares the preset behavioral demand prediction evaluation threshold index with the behavioral demand prediction evaluation index to generate interest tendency optimization data. The deep learning algorithm is used to perform interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine to execute the system switching operation. In this way, the learning content and learning environment of the learning machine can be intelligently adjusted according to the user's emotional fluctuations, learning status and interest tendencies, providing a better learning experience and learning effect. BRIEF DESCRIPTION OF THE DRAWINGS
[0075] Figure 1 A schematic diagram of a system switching method and a system flow chart of a tutor learning machine of the present invention;
[0076] Figure 2 Detailed implementation flow chart of step S1;
[0077] Figure 3 Detailed implementation flow chart of step S2;
[0078] Figure 4 It is a schematic diagram of the detailed implementation steps of step S3. Specific implementation manner
[0079] 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.
[0080] The embodiments of the present application provide a system switching method and system for a tutoring learning machine. The execution subjects of the system switching method and system of the tutoring learning machine include but are not limited to: mechanical equipment, data processing platforms, cloud server nodes, network uploading devices, etc. that can be regarded as general computing nodes of the present application. The data processing platform includes but is not limited to: at least one of an audio and image management system, an information management system, and a cloud data management system.
[0081] Please refer to Figures 1 to 4 , the present invention provides a system switching method for a tutoring learning machine. The system switching method of the tutoring learning machine includes the following steps:
[0082] Step S1: Obtain the user's facial image and user behavior data; perform dynamic behavior analysis on the user behavior data to generate dynamic behavior feature data; perform emotional fluctuation analysis on the user's facial image according to the dynamic behavior feature data to construct a user emotional fluctuation curve;
[0083] Step S2: Perform learning state detection on the dynamic behavior feature data through the user emotional fluctuation curve to generate learning state data; perform eye movement trajectory recognition on the user's facial image according to the learning state data to construct an eye movement trajectory map;
[0084] Step S3: Use the eye movement trajectory map to perform implicit association mining on the user emotional fluctuation curve to generate user behavior association data; perform user interest tendency analysis on the dynamic behavior feature data based on the user behavior association data to generate user interest tendency data; perform behavior tendency evolution processing on the user interest tendency data to construct a behavior interest tendency map;
[0085] Step S4: Perform personality difference analysis on the dynamic behavior feature data using the behavior interest tendency map to generate user personality difference data; construct a user portrait based on the user personality difference data to generate a user portrait; perform real-time dynamic optimization on the user portrait using the behavior interest tendency map to generate a user dynamic portrait;
[0086] Step S5: Perform emotional trend analysis on the user emotional fluctuation curve through the user dynamic portrait to generate user emotional trend data; perform behavior demand prediction on the dynamic behavior feature data through the user emotional trend data to generate a behavior demand prediction evaluation index;
[0087] Step S6: Compare the behavior demand prediction evaluation index based on a preset behavior demand prediction evaluation threshold index to generate interest tendency optimization data; use a deep learning algorithm to perform interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine to execute the system switching operation.
[0088] By obtaining the user's facial image and behavioral data, the present invention can acquire the user's facial expressions and behavioral information for subsequent analysis and modeling. Dynamic behavior analysis can deeply understand the user's behavior patterns and habits, including features in aspects such as actions, postures, and languages, providing a comprehensive understanding of the user's behavior. Based on the dynamic behavior feature data, emotional fluctuation analysis of the user's facial image can infer the user's emotional state and fluctuations. Learning state detection can determine the user's current learning state, such as concentration, distraction, fatigue, etc., based on the user's emotional fluctuation curve, providing a basis for subsequent learning process analysis and optimization. Through eye movement trajectory recognition, eye movement features such as the user's fixation points, fixation times, and fixation orders can be observed, providing data support for subsequent learning behavior analysis and cognitive process research. Through implicit association mining of the eye movement trajectory map, the potential relationship between the user's emotional fluctuation curve and eye movements can be discovered, revealing the connection between the user's emotions and cognition. According to the user behavior association data, the user's interest tendencies and preferences can be analyzed, understanding the user's behavior patterns and interest areas, providing a basis for personalized recommendations and customized services. Behavioral tendency evolution processing can track the changes and evolution process of the user's interests, constructing a behavioral interest tendency map, providing support for long-term analysis and prediction of the user's interests. The behavioral interest tendency map can reveal the personality differences and differences in behavior patterns among users. Through personality difference analysis of the dynamic behavior feature data, it can provide a basis for user personalized services and recommendation systems. The construction of the user portrait can comprehensively consider factors such as the user's interests, preferences, and behavior patterns, integrating the user's characteristics into a comprehensive description for better understanding and meeting the user's needs. The real-time dynamic optimization of the behavioral interest tendency map can update and adjust the user portrait in a timely manner according to the user's behavior and interest changes, maintaining the accuracy and practicality of the user portrait. Emotional trend analysis can use the user's dynamic portrait to analyze the long-term changes and trends of the user's emotional fluctuation curve, helping to understand the evolution and development of the user's emotions. Through the user's emotional trend data, behavioral demand prediction of the dynamic behavior feature data can predict the user's future behavioral tendencies and demands, providing a decision-making basis for personalized recommendations and services. By comparing the behavioral demand prediction evaluation index with the preset threshold index, it can be determined whether the user's interest tendency meets the expectations, thereby generating interest tendency optimization data for optimizing personalized recommendation and service strategies. Using deep learning algorithms to analyze and model the interest tendency switching data and interest tendency optimization data can generate a user interest switching decision engine for automatically executing system switching tasks to better meet the user's interests and needs.
[0089] In an embodiment of the present invention, referring to Figure 1 as described, it is a schematic diagram of the step flow of a system switching method and system for a tutoring learning machine of the present invention. In this example, the steps of the system switching method for the tutoring learning machine include:
[0090] Step S1: Obtain the user's facial image and user behavior data; perform dynamic behavior analysis on the user behavior data to generate dynamic behavior feature data; perform emotional fluctuation analysis on the user's facial image according to the dynamic behavior feature data to construct a user emotional fluctuation curve;
[0091] In this embodiment, the user's facial image is obtained through the learning machine camera, and data related to the user's behavior is collected, which may include information such as the user's actions, postures, voices, gestures, etc. These data can be obtained in various ways, such as using sensors, monitoring devices or user input. The collected user behavior data is analyzed and processed to extract feature information about the user's dynamic behavior, which may include identifying features in aspects such as the user's action patterns, frequencies, durations, speeds, etc. This is achieved through facial expression recognition technology. By analyzing the features of facial expressions, such as the movement changes of the eyes, mouth, eyebrows, etc., the user's emotional state is inferred. Using computer vision and pattern recognition methods, feature extraction and emotion classification can be performed on the facial image to identify the user's emotional state. By associating the emotion data with time, a curve graph of the user's emotion changing over time is drawn. The curve graph can show the fluctuation trend of the user's emotion, such as information about the high and low peaks of emotion, duration, frequency, etc.
[0092] Step S2: Perform learning state detection on the dynamic behavior feature data through the user emotional fluctuation curve to generate learning state data; perform eye movement trajectory recognition on the user's facial image according to the learning state data to construct an eye movement trajectory graph;
[0093] In this embodiment, using the user emotional fluctuation curve data, the user's learning state is detected and classified. The emotional fluctuation curve data and the corresponding learning state labels are trained to establish a learning state classification model. This model is used to predict new emotional fluctuation curve data to determine the user's current learning state. According to the learning state data, eye movement trajectory recognition is performed on the user's facial image to capture the user's eye movement information. This is achieved through the use of computer vision and image processing technologies, such as methods like face detection, eye localization and tracking. The face detection algorithm is used to find the face area in the user's facial image. The eye localization algorithm or feature point detection method is used to locate the eye area. Technologies such as the eye movement tracking algorithm or optical flow method are used to track the eye area to obtain the eye movement trajectory information. By associating the eye movement trajectory data with time, a trajectory graph of the eye movement changing over time is drawn. The trajectory graph can show the user's eye movement trajectory, such as information about the position of the fixation point, fixation time, fixation order, etc.
[0094] Step S3: Conduct implicit correlation mining on the user's emotional fluctuation curve using the eye movement trajectory map to generate user behavior correlation data; perform user interest tendency analysis on the dynamic behavior feature data based on the user behavior correlation data to generate user interest tendency data; conduct behavior tendency evolution processing on the user interest tendency data to construct a behavior interest tendency map;
[0095] In this embodiment, the eye movement trajectory map of the user and the emotional fluctuation curve data are used for implicit correlation mining to discover the correlation between the two, which is achieved by using data mining and machine learning methods, such as association rule mining, clustering analysis, etc. The eye movement trajectory map and the emotional fluctuation curve data are subjected to appropriate feature representation and encoding, and the association rule mining algorithm or clustering analysis algorithm is applied to discover the implicit correlation between the two, such as the correlation between the eye movement behavior pattern and the emotional fluctuation. Using the obtained user behavior correlation data, the user's interest tendency is analyzed and inferred, which is achieved by using machine learning algorithms and data mining techniques, such as classification algorithms, recommendation systems, etc. The behavior correlation data and the corresponding interest labels are trained to establish an interest tendency analysis model, and this model is used to predict new dynamic behavior feature data to determine the user's interest tendency. The user interest tendency data is further processed to capture the evolution and change trend of the behavior tendency, and a behavior interest tendency map is constructed, which is achieved by analyzing the user's historical behavior data and time series data. Time series analysis is performed on the user interest tendency data to discover the evolution pattern and trend of the behavior tendency, and these evolution patterns and trends are integrated into the behavior interest tendency map to show the change of the user's behavior interest over time.
[0096] Step S4: Conduct personality difference analysis on the dynamic behavior feature data using the behavior interest tendency map to generate user personality difference data; construct a user portrait based on the user personality difference data to generate a user portrait; perform real-time dynamic optimization on the user portrait using the behavior interest tendency map to generate a user dynamic portrait;
[0097] In this embodiment, based on the behavior interest tendency map constructed previously, personality difference analysis is performed on the dynamic behavior feature data of users to discover the personality differences between different users, which is achieved by using data mining and statistical analysis methods, such as clustering analysis, feature extraction, etc. The dynamic behavior feature data is subjected to appropriate feature representation and encoding, and clustering analysis algorithms or other personality difference analysis methods are applied to divide users into different personality groups, discover the characteristics and patterns of personality differences. Using the obtained user personality difference data, user portraits are constructed to deeply understand the characteristics and preferences of users, which is achieved by using user modeling and feature extraction techniques, such as feature engineering, statistical analysis, etc. According to the key features in the personality difference data, the feature vectors or feature descriptions of users are extracted, and based on these feature vectors or descriptions, user portraits are constructed, including the user's interest fields, behavior preferences, characteristics, etc. Based on the behavior interest tendency map, real-time dynamic optimization of the user portraits is performed to track the changes in users' behaviors and interests. The current dynamic behavior feature data of users is compared and matched with the behavior interest tendency map, and according to the matching results, the relevant feature and preference information in the user portraits is updated to achieve the real-time dynamic optimization of the user portraits and ensure that the user portraits are consistent with the actual behaviors and interests of users.
[0098] Step S5: Perform emotional trend analysis on the user's emotional fluctuation curve through the user dynamic portrait to generate user emotional trend data; perform behavior demand prediction on the dynamic behavior feature data through the user emotional trend data to generate a behavior demand prediction evaluation index.
[0099] In this embodiment, using the previously generated user dynamic portrait, emotional trend analysis is performed on the user's emotional fluctuation curve to discover the changing trend of the user's emotions, which is achieved by using time series analysis and pattern recognition methods, such as trend analysis, periodic analysis, etc. The emotional fluctuation curve data of users is subjected to appropriate representation and encoding, and time series analysis methods are applied to explore the trends, seasonality, and periodic changes in the emotional fluctuation curve to discover the long-term and short-term trends of the user's emotions. Using the obtained user emotional trend data, behavior demand prediction is performed on the dynamic behavior feature data of users to predict the user's future behavior demands and preferences, which is achieved by using machine learning and prediction models, such as regression analysis, time series prediction, etc. The emotional trend data of users and the corresponding dynamic behavior feature data are subjected to feature representation and encoding, and appropriate machine learning algorithms or prediction models are applied to establish a behavior demand prediction model to predict the user's future behavior demand evaluation index.
[0100] Step S6: Compare the behavior demand prediction evaluation index based on a preset behavior demand prediction evaluation threshold index to generate interest tendency optimization data; use a deep learning algorithm to perform interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine for performing system switching operations.
[0101] In this embodiment, according to a preset behavior demand prediction evaluation threshold index, the behavior demand prediction evaluation index is compared with the threshold to determine the satisfaction degree of the user's behavior demand. The threshold can be set according to business requirements and user feedback and is used to judge whether the user is satisfied with the behavior demand of the current system. If the behavior demand prediction evaluation index is higher than the threshold, it means that the user's behavior demand is met; if it is lower than the threshold, it means that the user's behavior demand is not fully met. According to the comparison result of the behavior demand prediction evaluation index and the threshold, interest tendency optimization data is generated to optimize the user's interest tendency and behavior recommendations. If the behavior demand is met, the corresponding behavior feature data can be used as positive optimization data to strengthen the user's interest tendency and behavior preferences. If the behavior demand is not met, the corresponding behavior feature data can be used as negative optimization data to adjust the user's interest tendency and behavior preferences. Use a deep learning algorithm to analyze and model the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine, which can be implemented by using a deep neural network or other deep learning models, such as a multi-layer perceptron, a convolutional neural network, etc. Represent and encode the interest tendency switching data and the interest tendency optimization data, apply an appropriate deep learning algorithm, and train the model to learn the user's interest switching rules and decision-making strategies to generate a user interest switching decision engine for performing system switching operations according to the user's interest tendency and behavior demand, such as switching recommended content, adjusting the interface layout, etc.
[0102] In this embodiment, referring to Figure 2 As described, it is a schematic diagram of the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:
[0103] Step S11: Obtain the user's facial image and the user's behavior data;
[0104] Step S12: Perform time series analysis on the user's behavior data to generate behavior time series data;
[0105] Step S13: Perform behavior pattern analysis on the user's behavior data according to the behavior time series data to generate user behavior pattern data;
[0106] Step S14: Perform dynamic behavior analysis on the user's behavior pattern data to generate dynamic behavior feature data;
[0107] Step S15: Detect micro-expression features of the user's facial image to generate user micro-expression feature data;
[0108] Step S16: Analyze the emotional fluctuations of the user's micro-expression feature data based on the dynamic behavior feature data to construct a user emotional fluctuation curve.
[0109] Through obtaining the user's facial image and behavior data, the present invention can acquire the user's facial expressions and behavior information, providing a data basis for subsequent analysis and modeling. The time series analysis can deeply understand the user's behavior patterns and habits, including features in aspects such as actions, postures, and languages, providing a comprehensive understanding of the user's behavior. The behavior pattern analysis can identify and extract the user's common behavior patterns, helping to understand the user's preferences and habits, and providing a basis for subsequent personalized services and recommendations. The dynamic behavior analysis can deeply analyze the user's behavior patterns, revealing the dynamic changes and trends of the user's behavior, and providing a basis for personalized services and behavior prediction. The micro-expression feature detection can capture the changes in the user's facial micro-expressions, revealing the subtle fluctuations of the user's emotions and feelings. The emotional fluctuation analysis can infer the user's emotional state and fluctuation conditions, correlating the user's behavior and emotions, and providing an emotional basis and a reference for personalized adjustment for the system switching of the learning machine.
[0110] In this embodiment, a suitable device or sensor (such as a camera) is used to obtain the facial image data of the user, which is achieved by calling the corresponding image acquisition interface or using computer vision technology. Data related to the user's behavior is collected, such as the user's click, swipe, input and other behavior data, which is obtained by monitoring the way the user interacts with the system, such as recording the user's mouse movement, keyboard input, etc. The collected user behavior data is sorted and arranged in chronological order. Appropriate time series analysis methods, such as time series analysis, sequential pattern mining, etc., are applied to analyze the user behavior data. According to the results of the time series analysis, behavior time series data is generated, that is, the user behavior sequence arranged in chronological order. Data mining, machine learning or statistical analysis and other methods are used to perform pattern recognition and clustering analysis on the behavior time series data. According to the results of the behavior pattern analysis, user behavior pattern data is generated, that is, the common behavior patterns of the user are identified and recorded. Key dynamic behavior characteristics are extracted from the behavior pattern data to describe the dynamic behavior characteristics of the user. Characteristics such as the frequency, periodicity, and conversion speed of the user's behavior are extracted. According to the results of the dynamic behavior analysis and feature extraction, dynamic behavior feature data is generated, that is, the data set describing the dynamic behavior characteristics of the user. Computer vision technology is used to detect and extract the micro-expression features of the user's facial image. Micro-expression is a facial expression change that appears instantaneously and lasts for a very short time, which can reflect the user's emotions and internal state. Key micro-expression features are extracted from the user's facial image, such as the movements and wrinkles of the eyes and mouth. According to the results of the detection and extraction of the micro-expression features, user micro-expression feature data is generated, that is, the data set describing the micro-expression features of the user. The dynamic behavior feature data and the user micro-expression feature data are used to analyze the user's emotional fluctuations, including correlating and comparing the dynamic behavior features with the micro-expression features to understand the user's emotional changes. The emotional fluctuation curve can be a visual representation of the emotional state over time, reflecting the user's emotional change trend.
[0111] In this embodiment, step S16 includes the following steps:
[0112] Step S161: Identify the position points of the facial features in the user's facial image to obtain the user's facial feature position data;
[0113] Step S162: Detect the facial muscle distortion frequency of the user's facial image based on the user's facial feature position data to generate facial muscle distortion data;
[0114] Step S163: Quantify the emotional intensity of the user's micro-expression feature data through the facial muscle distortion data to generate an emotional intensity quantification index;
[0115] Step S164: Analyze the emotional fluctuations of the emotional intensity quantification index according to the dynamic behavior feature data to generate the user's emotional fluctuation law;
[0116] Step S165: Use the law of user emotional fluctuations to perform emotional fluctuation curve fitting on the emotional intensity quantization index to construct a user emotional fluctuation curve.
[0117] In the present invention, by identifying the position points of the facial features in the user's facial image, the positions of the respective feature points on the user's face can be accurately determined, providing basic data for subsequent detection of facial muscle distortion frequency and emotional analysis. The detection of facial muscle distortion frequency can analyze the degree and frequency of distortion of the user's facial muscles, capture the minute movement changes of the facial muscles, providing a data basis for emotional analysis and emotional intensity quantization. The emotional intensity quantization process can convert the user's micro-expression feature data into a quantization index of emotional intensity, describing the strength of the user's emotions, providing basic data for emotional fluctuation analysis and construction of emotional fluctuation curves. Emotional fluctuation analysis can analyze the fluctuation situation and trend of the user's emotions according to the emotional intensity quantization index, revealing the ups and downs of the user's emotions. By fitting and modeling the emotional intensity quantization index, an emotional fluctuation curve of the user individual can be constructed, describing the variation law of the user's emotions over time, providing a more accurate emotional prediction and adjustment basis for the system switching and personalized adaptation of the learning machine.
[0118] In this embodiment, the user's facial feature position data is used to detect the distortion of facial muscles by calculating the relative movement or deformation between the facial features. This involves performing time-domain or frequency-domain analysis on the changes in the positions of the facial features to capture the frequency characteristics of facial muscle distortion. These data can be represented as a time series, which contains the dynamic information of facial muscle distortion. By performing emotional analysis on the facial muscle distortion data, the intensity of the user's micro-expressions can be quantified, and a model or rule system can be established to map the facial muscle distortion data to a measure of emotional intensity, extracting emotional information from the facial muscle distortion and generating a numerical value representing the emotional intensity. Considering the user's dynamic behavior, time series analysis methods such as sliding window or time series model are used to perform fluctuation analysis on the emotional intensity quantization index, capture the dynamic changes of emotions, identify the high and low peaks and change trends of emotions, forming the law of user emotional fluctuations. Based on the results of the fluctuation analysis, curve fitting techniques such as polynomial fitting or spline interpolation can be used to perform curve fitting on the emotional intensity quantization index, establishing a smooth curve reflecting the overall fluctuation trend of the user's emotions. The fitted curve can be used as a visual representation of the user's emotional fluctuations, providing a deeper understanding of the user's emotional state.
[0119] In this embodiment, refer to Figure 3 As described, it is a schematic diagram of the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:
[0120] Step S21: Detect the learning state of the dynamic behavior feature data through the user's emotional fluctuation curve to generate learning state data;
[0121] Step S22: Perform eye movement trajectory recognition on the user's facial image according to the learning state data to generate eye movement trajectory data;
[0122] Step S23: Analyze the fixation frequency of the eye movement trajectory data to generate fixation frequency data;
[0123] Step S24: Analyze the fixation duration of the user's facial image according to the fixation frequency data to generate fixation duration data;
[0124] Step S25: Analyze the blink frequency of the eye movement trajectory data based on the fixation duration data to generate blink frequency data;
[0125] Step S26: Detect the range of eye movement of the eye movement trajectory data according to the blink frequency data to generate eye movement range data;
[0126] Step S27: Perform trajectory visualization on the eye movement range data according to the eye movement trajectory data to construct an eye movement trajectory map.
[0127] Through the analysis of the user's emotional fluctuation curve, the present invention can detect the user's learning state, such as concentration, excitement, etc., providing a basis for subsequent learning behavior analysis and personalized adjustment. Eye movement trajectory recognition can track the user's eye movement trajectory, including fixation points and fixation paths, providing a data basis for subsequent fixation frequency analysis and eye movement range detection. Fixation frequency analysis can calculate the fixation frequency of the user within a specific time period, that is, the change frequency of fixation points, helping to understand the user's attention degree and attention distribution to different learning contents. Fixation duration analysis can measure the length of time the user stays at different fixation points, revealing the user's emphasis and interest points on learning contents, providing a basis for personalized recommendation and adaptive adjustment of learning contents. Blink frequency analysis can calculate the blink frequency of the user to understand the fatigue degree of the user's eyes and the change of attention. Eye movement range detection can analyze the movement range of the user's eyes on the screen to understand the user's visual focus and attention concentration, providing a reference for the layout and interface design of learning contents. By visualizing the eye movement range data, the user's line-of-sight movement trajectory can be intuitively displayed, helping to understand the attention distribution and visual attention pattern of the user during the learning process.
[0128] In this embodiment, according to the user's emotional fluctuation curve, using a suitable learning state detection algorithm or model, the learning state of the dynamic behavior feature data is judged. The learning state can include states such as concentration, distraction, and fatigue. According to the results of the learning state detection, learning state data is generated, that is, a data set describing the user's learning state. Using computer vision technology, the selected eye image is identified and tracked for the eye movement trajectory. According to the results of the eye movement trajectory identification, eye movement trajectory data is generated, that is, a data set describing the user's eye movement trajectory. Based on the eye movement trajectory data, the distribution and duration of the fixation points are calculated to determine the degree of the user's attention concentration on a specific area. According to the results of the fixation frequency analysis, key fixation frequency data is extracted, that is, data describing the user's attention allocation to different areas. According to the extracted fixation frequency data, a fixation frequency data set is generated to represent the user's attention distribution to different areas. Based on the fixation frequency data and the eye movement trajectory data, the fixation duration of the user on different areas is calculated. According to the results of the fixation duration analysis, key fixation duration data is extracted, that is, data describing the user's fixation duration on different areas. Using the fixation duration data and the eye movement trajectory data, the user's blink frequency is calculated, that is, the number of blinks per unit time. According to the results of the blink frequency analysis, key blink frequency data is extracted to describe the user's blink frequency situation. Based on the blink frequency data and the eye movement trajectory data, the user's eye movement range is analyzed, that is, the activity range of the eyeball within the visual field. According to the results of the eye movement range detection, key eye movement range data is extracted to describe the user's eye movement range. Using the eye movement trajectory data and the eye movement range data, the user's eyeball movement process is visualized as an eye movement trajectory map. According to the results of the trajectory visualization, the user's eye movement trajectory map is constructed, which shows information such as the user's eye movement trajectory, fixation points, blink frequency, and eye movement range.
[0129] In this embodiment, referring to Figure 4 as described, it is a schematic diagram of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include:
[0130] Step S31: Use the eye movement trajectory map to mine the implicit association of the user's emotional fluctuation curve to generate user behavior association data;
[0131] Step S32: Calculate the similarity of the dynamic behavior feature data based on the user behavior association data to generate dynamic behavior similarity data;
[0132] Step S33: Use the dynamic behavior similarity data to analyze the user's interest tendency of the dynamic behavior feature data to generate user interest tendency data;
[0133] Step S34: Identify points of interest from the user interest tendency data to generate user point-of-interest data;
[0134] Step S35: Perform behavioral tendency evolution processing on the user interest tendency data through the user point-of-interest data to construct a behavioral interest tendency map.
[0135] Through the associated mining of the eye movement trajectory map and the user's emotional fluctuation curve, the present invention can discover the potential association relationship between the eye movement trajectory and the emotional fluctuation, providing data support for subsequent behavioral analysis and personalized recommendation. By calculating the similarity of the user behavior association data, the similarity degree between different behaviors can be quantified, providing a basis for understanding the user's behavior pattern and behavior preference, so as to better perform personalized recommendation and learning content adaptation. By analyzing the dynamic behavior similarity data, the user's interest tendency and behavior preference can be revealed, and the preference degree of the user for different learning contents and tasks can be understood. The point-of-interest identification can classify and label the user's interest tendency, identify and extract the user's interest fields and concerns, providing basic data for the personalized recommendation and content filtering of the learning machine. The behavioral tendency evolution processing can analyze the change trend and evolution law of the user's points of interest, construct a map of the user's behavioral interest tendency, and reveal the evolution path of the user's interest change and learning preference.
[0136] In this embodiment, the eye movement trajectory map and the emotional fluctuation curve are associated and mined to explore the implicit relationship therein. Data mining or machine learning algorithms are used to analyze the association between the eye movement trajectory map and the emotional fluctuation curve. According to the results of the implicit association mining, a user behavior association data set is generated, that is, the data describing the association between the user's behavior and emotion. Using the user behavior association data and the dynamic behavior feature data, the similarity between them is calculated. According to the results of the similarity calculation, a dynamic behavior similarity data set is generated, which is used to represent the similarity situation between the user's dynamic behavior features. Based on the dynamic behavior similarity data and the dynamic behavior feature data, the user's interest tendency is analyzed, that is, the preference degree of the user for different behavior features. According to the results of the user interest tendency analysis, a user interest tendency data set is generated, which is used to describe the interest tendency situation of the user for different behavior features. Based on the user interest tendency data, the user's points of interest are identified, that is, the specific behaviors or topics that the user is interested in. Techniques such as text mining, keyword extraction, or topic models can be used for point-of-interest identification. The evolution of the user's interest is explored through methods such as time series analysis, association rule mining, or deep learning models. The behavioral interest tendency map is a graph structure representing the evolution of the user's interest and the behavior association, where the nodes represent the user's points of interest and the edges represent the association relationships between the points of interest. By continuously collecting and analyzing the user data, the structure and content of the behavioral interest tendency map are continuously updated and improved.
[0137] In this embodiment, step S4 includes the following steps:
[0138] Step S41: Analyze the personality preferences of the dynamic behavior feature data by using the behavior interest tendency map to generate personality preference data;
[0139] Step S42: Analyze the personality differences of the personality preference data to generate user personality difference data;
[0140] Step S43: Identify the interest driving forces of the behavior interest tendency map according to the user personality difference data to generate interest driving force data;
[0141] Step S44: Construct a user portrait based on the interest driving force data for the user personality difference data to generate a user portrait;
[0142] Step S45: Analyze the temporal weights of the interest characteristics of the user portrait by using the behavior interest tendency map to generate interest temporal weight data;
[0143] Step S46: Perform real-time dynamic iterative optimization on the user portrait according to the interest temporal weight data to generate a user dynamic portrait.
[0144] Through the analysis of the behavior interest tendency map and dynamic behavior feature data, the present invention can deeply understand the user's personality preferences and learning interests, providing a basis for the personalized recommendation and content customization of the learning machine to meet the user's personalized needs. Personality difference analysis can identify the personality differences and preference differences between users, helping to understand the learning preferences, behavior habits and preference characteristics of different users, providing a basis for personalized recommendation and customized learning strategies. Interest driving force identification can identify the driving forces and motivational factors behind the user's learning behavior according to the user's personality differences and interest preferences, providing a basis for the personalized recommendation and learning motivation of the learning machine, enhancing the user's learning motivation. Interest driving force identification can identify the driving forces and motivational factors behind the user's learning behavior according to the user's personality differences and interest preferences, providing a basis for the personalized recommendation and learning motivation of the learning machine, enhancing the user's learning motivation. User portrait construction can comprehensively consider the user's personality differences, interest driving forces and behavior characteristics to form a comprehensive description and feature generalization of the user. Interest characteristic temporal weight analysis can analyze the change weights of the user's interests at different time periods, reflecting the user's attention and priority to different interests, providing a reference basis for the real-time recommendation and dynamic adjustment of the learning machine. Real-time dynamic iterative optimization can update and optimize the user portrait in real time according to the user's interest temporal weight data to adapt to the changes in the user's interests and the dynamic adjustment of learning needs, providing more accurate personalized learning support.
[0145] In this embodiment, based on the behavioral interest tendency map, the personality preferences of users are analyzed, that is, the preference degrees of users at different interest points. The personality preferences can be measured by calculating the correlation degree or path weight between the user and the interest point. Based on the personality preference data, the differences between user personalities are analyzed, and clustering algorithms, principal component analysis or other statistical methods are used to explore the user personality differences. According to the results of the personality difference analysis, a user personality difference data set is generated to represent the personality differences between users. By analyzing the correlation relationship between the user personality differences and the behavioral interest tendency map, the interest driving force is identified. According to the results of the interest driving force identification, an interest driving force data set is generated to describe the interest driving force of users. The interest driving force is used as one of the important features of the user portrait, and is segmented and described in combination with the personality difference data. Based on the behavioral interest tendency map and the user portrait data, the weight change of the user interest characteristics in different time periods is analyzed. The time series weight of the interest characteristics is calculated by considering indicators such as the attention degree and click-through rate of the user on specific interest points in different time periods. By adjusting the weight of the interest characteristics in the user portrait, the time series change of the user interest is adapted. According to the results of the real-time dynamic iterative optimization, a user dynamic portrait data set is generated to describe the dynamic interest characteristics and preferences of users in different time periods.
[0146] In this embodiment, the specific steps of step S5 are as follows:
[0147] Step S51: Perform an emotional cycle analysis on the user's emotional fluctuation curve through the user dynamic portrait to generate emotional cycle data;
[0148] Step S52: Detect the emotional mutation points on the user's emotional fluctuation curve through the emotional cycle data and mark the emotional mutation points;
[0149] Step S53: Perform an emotional trend analysis on the user's emotional fluctuation curve according to the emotional mutation points to generate user emotional trend data;
[0150] Step S54: Perform a demand trend prediction on the dynamic behavior characteristic data through the user emotional trend data to generate demand trend prediction data;
[0151] Step S55: Use the behavioral demand trend prediction probability calculation formula to perform a probability prediction evaluation on the demand trend prediction data to generate a behavioral demand prediction evaluation index.
[0152] Through emotional cycle analysis, the present invention can identify the periodic patterns of users' emotional fluctuations, understand the periodic changes in users' emotions, and provide a basis for the emotional regulation and mood management of learning machines, so as to improve users' emotional stability and learning effects. Emotional mutation point detection can identify the mutation points in the users' emotional fluctuation curves, mark the important turning points in the emotional fluctuations, help understand the changes and triggering factors of users' emotions, and provide a basis for the emotional regulation, personalized feedback and intervention of learning machines. Emotional trend analysis can analyze the trend changes in the users' emotional fluctuation curves, identify the rising, falling or stable trends of users' emotions, and provide a basis for the emotional regulation and personalized feedback of learning machines to adapt to the emotional needs and emotional states of users. Demand trend prediction can predict the trend changes of users' learning needs based on users' emotional trends and behavioral characteristic data, and understand the degree of users' demand for different learning contents and resources. The behavioral demand prediction evaluation index can conduct a probability evaluation on the demand trend prediction data, quantify the credibility and accuracy of the prediction results, and provide a reference for the decision-making and behavior adjustment of learning machines to improve the intelligence and personalized adaptation ability of learning machines.
[0153] In this embodiment, periodic analysis methods such as Fourier transform and autocorrelation function are used to identify the periodicity of emotional fluctuations, and periodic analysis is performed on the users' emotional data to discover the periodic changes in emotions. Mutation point detection algorithms such as threshold-based methods and statistics-based methods are used to detect the mutation points in the emotional fluctuation curves, and change point detection is performed on the emotional cycle data to determine the mutation points in the emotional fluctuation curves. Based on the emotional mutation points, the trend changes in the users' emotional fluctuation curves are analyzed to determine the overall trend of emotions. Trend analysis methods such as linear regression and moving average are used to analyze the trends of the emotional fluctuation curves. According to the results of the emotional trend analysis, a users' emotional trend dataset is generated to describe the overall trend changes of users' emotions. Based on the users' emotional trend data, demand trend prediction is performed on the dynamic behavioral characteristic data to infer the future demand trends of users. Prediction models such as time series analysis and machine learning are used to predict the trends of demands. The defined behavioral demand trend prediction probability calculation formula is used to perform probability prediction evaluation on the demand trend prediction data. According to the calculation results of the formula, the corresponding behavioral demand probability values of each demand trend prediction data point are obtained. Based on the behavioral demand probability values, a behavioral demand prediction evaluation index is generated to evaluate the behavioral demand credibility or importance degree of each demand trend prediction data point.
[0154] In this embodiment, the behavioral demand trend prediction probability calculation formula in step S55 is specifically:
[0155]
[0156] Among them, P is the behavior demand prediction evaluation index, t is the time point of demand prediction, X is the predicted behavior feature weight index, Y is the user behavior adaptability parameter, Z is the behavior demand change rate, S is the decay rate of behavior demand over time, N is the emotional fluctuation amplitude value, d is the interest driving force index, and λ is the fixation duration of the interest point.
[0157] Through the present invention By calculating the time derivative of the variance of the behavior feature weight index multiplied by the mean value of the user behavior adaptability parameter divided by the square of the natural logarithm of the behavior demand change rate, the change rate of the behavior demand trend can be understood. This can provide a more refined prediction result and take into account the change of behavior demand over time. By combining the variance of the behavior feature weight index, the mean value of the user behavior adaptability parameter, and the natural logarithm of the behavior demand change rate, and performing a square operation, the influence of these factors on the behavior demand can be comprehensively considered. This helps to more accurately predict the behavior demand trend and provide a more reliable evaluation index. By calculating the natural logarithm after adding 1 to the ratio of the decay rate of behavior demand over time to the emotional fluctuation amplitude value, the influence of emotional fluctuation on behavior demand can be considered. This helps to more comprehensively predict the behavior demand trend and make the prediction result more accurate. Calculating the interest driving force index divided by the square root of the fixation duration of the interest point can comprehensively consider the user's interest driving force and the degree of continuous attention to the interest point. This is of great significance for the accuracy of predicting the behavior demand trend and makes the prediction result more in line with the user's actual interests and behavior demands. By performing an exponential operation on the negative value of the time derivative and converting the result into a probability value, the behavior demand prediction evaluation index can be limited between 0 and 1. This helps to intuitively understand the prediction probability of behavior demand and the credibility evaluation of the behavior demand trend.
[0158] In this embodiment, the specific steps of step S6 are as follows:
[0159] Step S61: Compare the behavior demand prediction evaluation index based on a preset behavior demand prediction evaluation threshold index. When the preset behavior demand prediction evaluation threshold index is less than the behavior demand prediction evaluation index, the user interest page is switched through the user dynamic portrait to generate interest tendency switching data.
[0160] Step S62: When the preset behavior demand prediction evaluation threshold index is greater than or equal to the behavior demand prediction evaluation index, optimize the interest tendency of the user dynamic portrait to generate interest tendency optimization data.
[0161] Step S63: Use a machine learning algorithm to perform interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine to execute the system switching operation.
[0162] By comparing the behavioral demand prediction evaluation index with a preset threshold index, the present invention can determine the degree of difference between the user's current behavioral demand and the expectation. When the behavioral demand prediction evaluation index is higher than the threshold index, it means that the user's behavioral demand exceeds the expectation. At this time, switching the user interest page can provide learning content or resources that better meet the user's needs, enhancing the personalized adaptation ability of the learning machine. When the behavioral demand prediction evaluation index is lower than or equal to the preset threshold index, it indicates that the user's behavioral demand is in line with or lower than the expectation. At this time, optimizing the interest tendency of the user dynamic portrait can further analyze and understand the user's interest preferences, providing more accurate user portrait data for the learning machine to optimize personalized recommendation and learning path planning. By analyzing the interest tendency switching data and interest tendency optimization data through machine learning algorithms, the interest change trend and preference of the user can be identified, providing a decision-making engine for the learning machine, thereby realizing automatic system switching. In this way, according to the user's behavioral demand and interest changes, the learning content, learning method or learning environment of the learning machine can be adjusted in a timely manner, providing a better learning experience and learning effect.
[0163] In this embodiment, if the preset threshold is less than the behavior demand prediction evaluation index, it indicates that the credibility or importance of the behavior demand is higher than the preset threshold, and it is necessary to switch the user interest page. According to the user dynamic portrait data, the interest page is switched. Based on information such as the user's interest preferences and behavior history, a suitable page is selected for switching, and the relevant data of the user interest page switching is recorded, such as the page before switching, the page after switching, etc. If the preset threshold is greater than or equal to the behavior demand prediction evaluation index, it indicates that the credibility or importance of the behavior demand is lower than or equal to the preset threshold, and there is no need to switch the page. Instead, the interest tendency of the user dynamic portrait is optimized. According to the user dynamic portrait data, the user's interest preferences, behavior history and other information are optimized. More content and services that match the user's interests can be provided through methods such as personalized recommendation algorithms and content filtering. The input features are determined, such as the page before switching, the page after switching, the interest preferences before optimization, the interest preferences after optimization, etc., and the corresponding labels indicating whether to perform page switching are used. Machine learning algorithms, such as classification algorithms (such as decision trees, random forests, logistic regression, etc.) or regression algorithms (such as linear regression, support vector regression, etc.), are used to train the prepared training data set. During the training process, the algorithm will learn the correlation between the interest tendency switching data and the interest tendency optimization data, as well as their relationship with the user interest page switching. After training, a user interest switching decision engine, that is, a trained model or algorithm, is generated. The generated user interest switching decision engine is used to predict and make decisions on future user behaviors. When there is a new user behavior demand prediction evaluation index, it is input into the user interest switching decision engine, and the engine will judge whether to perform the page switching operation according to the prediction evaluation index and the trained model to meet the user's interest needs.
[0164] In this embodiment, a system switching system for a tutoring learning machine is provided, including:
[0165] An emotional fluctuation module, configured to obtain a user facial image and user behavior data; perform dynamic behavior analysis on the user behavior data to generate dynamic behavior feature data; perform emotional fluctuation analysis on the user facial image according to the dynamic behavior feature data to construct a user emotional fluctuation curve;
[0166] An eye movement trajectory module, configured to perform learning state detection on the dynamic behavior feature data through the user emotional fluctuation curve to generate learning state data; perform eye movement trajectory recognition on the user facial image according to the learning state data to construct an eye movement trajectory map;
[0167] An interest tendency module, which is used to mine the implicit association between the user's emotional fluctuation curve by using the eye movement trajectory diagram to generate user behavior association data; analyze the user's interest tendency of the dynamic behavior feature data based on the user behavior association data to generate user interest tendency data; perform behavior tendency evolution processing on the user interest tendency data to construct a behavior interest tendency map;
[0168] A dynamic portrait module, which is used to analyze the personality differences of the dynamic behavior feature data by using the behavior interest tendency map to generate user personality difference data; construct a user portrait based on the user personality difference data to generate a user portrait; use the behavior interest tendency map to perform real-time dynamic optimization on the user portrait to generate a user dynamic portrait;
[0169] A behavior demand prediction module, which is used to analyze the emotional trend of the user's emotional fluctuation curve through the user dynamic portrait to generate user emotional trend data; predict the behavior demand of the dynamic behavior feature data through the user emotional trend data to generate a behavior demand prediction evaluation index;
[0170] A decision engine module, which is used to compare the behavior demand prediction evaluation index based on a preset behavior demand prediction evaluation threshold index to generate interest tendency optimization data; use a deep learning algorithm to perform interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine to execute system switching operations.
[0171] The present invention constructs a system switching system for a tutor learning machine, obtains user facial images and behavior data, and performs dynamic behavior analysis to understand the user's emotional fluctuations and behavioral characteristics. Through the emotional fluctuation analysis, the user's emotional fluctuation curve can be constructed to better understand the user's emotional state and changing trend. By learning the user's emotional fluctuation curve and combining the dynamic behavior feature data, the eye trajectory module can perform learning state detection and eye trajectory recognition, which can help determine the user's attention, concentration and eye movement trajectory during the learning process, provide information about the user's learning state, and provide basic data for subsequent interest tendency analysis. Through the analysis of the eye trajectory graph, the interest tendency module can mine the implicit association between the user's emotional fluctuation curve and behavior, and generate user behavior association data. Based on these data, the user's interest tendency analysis can be performed on the dynamic behavior feature data to further understand the user's interest preferences and behavioral tendencies. By constructing a behavioral interest tendency map, the user's learning interest can be better understood, providing a basis for personalized learning. The dynamic portrait module uses the behavioral interest tendency map to personalize the dynamic behavior feature data. The user personality difference analysis generates user personality difference data, which helps to gain an in-depth understanding of the unique characteristics and behavior patterns of each user, thereby building a user portrait. By dynamically optimizing the user portrait in real time, it can better adapt to the user's individual needs and provide personalized learning experience and learning recommendations. The behavior demand prediction module uses the user's dynamic portrait to perform emotional trend analysis on the emotional fluctuation curve and generate user emotional trend data. Through the emotional trend data, the dynamic behavior feature data can be used to predict behavioral needs, further understand the user's learning needs and behavioral tendencies, and the generated behavioral demand prediction evaluation index can be provided to the subsequent decision engine for further processing. The decision engine module compares the preset behavioral demand prediction evaluation threshold index with the behavioral demand prediction evaluation index to generate interest tendency optimization data. The deep learning algorithm is used to perform interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine to execute the system switching operation. In this way, the learning content and learning environment of the learning machine can be intelligently adjusted according to the user's emotional fluctuations, learning status and interest tendencies, providing a better learning experience and learning effect.
[0172] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.
[0173] It should be understood that although the terms "first", "second", etc. are used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly, the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.
[0174] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather should be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A system switching method for a tutoring learning machine, characterized in that, It includes the following steps: Step S1: Obtain the user's facial image and user behavior data; Conduct dynamic behavior analysis on the user behavior data to generate dynamic behavior feature data; Conduct emotional fluctuation analysis on the user's facial image according to the dynamic behavior feature data to construct the user's emotional fluctuation curve; Step S2: Detect the learning state of the dynamic behavior feature data through the user's emotional fluctuation curve to generate learning state data; Conduct eye movement trajectory recognition on the user's facial image according to the learning state data to construct an eye movement trajectory map; Step S3: Use the eye movement trajectory map to mine implicit associations of the user's emotional fluctuation curve to generate user behavior association data; conduct user interest tendency analysis on the dynamic behavior feature data based on the user behavior association data to generate user interest tendency data; Conduct behavioral tendency evolution processing on the user interest tendency data to construct a behavioral interest tendency map; Step S4: Use the behavioral interest tendency map to conduct personality difference analysis on the dynamic behavior feature data to generate user personality difference data; Construct a user portrait based on the user personality difference data to generate a user portrait; Use the behavioral interest tendency map to perform real-time dynamic optimization on the user portrait to generate a user dynamic portrait; Step S5: Conduct emotional trend analysis on the user's emotional fluctuation curve through the user dynamic portrait to generate user emotional trend data; Conduct behavioral demand prediction on the dynamic behavior feature data through the user emotional trend data to generate a behavioral demand prediction evaluation index; Step S6: Compare the behavioral demand prediction evaluation index with a preset behavioral demand prediction evaluation threshold index to generate interest tendency optimization data; use a deep learning algorithm to conduct interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine to execute system switching operations.
2. The method according to claim 1, wherein The specific steps of Step S1 are as follows: Step S11: Obtain the user's facial image and user behavior data; Step S12: Conduct time series analysis on the user behavior data to generate behavioral time series data; Step S13: Conduct behavior pattern analysis on the user behavior data according to the behavioral time series data to generate user behavior pattern data; Step S14: Conduct dynamic behavior analysis on the user behavior pattern data to generate dynamic behavior feature data; Step S15: Conduct micro-expression feature detection on the user's facial image to generate user micro-expression feature data; Step S16: Conduct emotional fluctuation analysis on the user micro-expression feature data according to the dynamic behavior feature data to construct the user's emotional fluctuation curve.
3. The method according to claim 2, wherein The specific steps of Step S16 are as follows: Step S161: Identify the position points of the facial features of the user's facial image to obtain user facial feature position data; Step S162: Conduct facial muscle distortion frequency detection on the user's facial image based on the user facial feature position data to generate facial muscle distortion data; Step S163: Conduct emotional intensity quantification processing on the user micro-expression feature data through the facial muscle distortion data to generate an emotional intensity quantification index; Step S164: Perform emotional fluctuation analysis on the emotional intensity quantization index based on the dynamic behavior characteristic data to generate the user's emotional fluctuation pattern; Step S165: Fit an emotional fluctuation curve to the emotional intensity quantization index using the user's emotional fluctuation pattern to construct the user's emotional fluctuation curve.
4. The method according to claim 1, wherein The specific steps of Step S2 are as follows: Step S21: Detect the learning state of the dynamic behavior characteristic data through the user's emotional fluctuation curve to generate learning state data; Step S22: Identify the eye movement trajectory of the user's facial image based on the learning state data to generate eye movement trajectory data; Step S23: Analyze the fixation frequency of the eye movement trajectory data to generate fixation frequency data; Step S24: Analyze the fixation duration of the user's facial image based on the fixation frequency data to generate fixation duration data; Step S25: Analyze the blink frequency of the eye movement trajectory data based on the fixation duration data to generate blink frequency data; Step S26: Detect the range of eye movement of the eye movement trajectory data based on the blink frequency data to generate eye movement range data; Step S27: Visualize the trajectory vision of the eye movement range data based on the eye movement trajectory data to construct an eye movement trajectory map.
5. The method according to claim 1, wherein The specific steps of Step S3 are as follows: Step S31: Mine the implicit association between the user's emotional fluctuation curve and the eye movement trajectory map to generate user behavior association data; Step S32: Calculate the similarity of the dynamic behavior characteristic data based on the user behavior association data to generate dynamic behavior similarity data; Step S33: Analyze the user's interest tendency of the dynamic behavior characteristic data using the dynamic behavior similarity data to generate user interest tendency data; Step S34: Identify the interest points of the user interest tendency data to generate user interest point data; Step S35: Process the behavior tendency evolution of the user interest tendency data through the user interest point data to construct a behavior interest tendency map.
6. The method according to claim 1, characterized in that, The specific steps of Step S4 are as follows: Step S41: Analyze the personality preferences of the dynamic behavior characteristic data using the behavior interest tendency map to generate personality preference data; Step S42: Analyze the personality differences of the personality preference data to generate user personality difference data; Step S43: Identify the interest driving force of the behavior interest tendency map based on the user personality difference data to generate interest driving force data; Step S44: Construct a user portrait based on the user personality difference data using the interest driving force data to generate a user portrait; Step S45: Analyze the temporal weights of the interest characteristics of the user portrait using the behavior interest tendency map to generate interest temporal weight data; Step S46: Perform real-time dynamic iterative optimization on the user portrait based on the interest temporal weight data to generate a user dynamic portrait.
7. The method according to claim 1, wherein The specific steps of Step S5 are as follows: Step S51: Analyze the emotional cycle of the user's emotional fluctuation curve through the user dynamic portrait to generate emotional cycle data; Step S52: Detect the emotional mutation points of the user's emotional fluctuation curve through the emotional cycle data and mark the emotional mutation points; Step S53: Perform emotional trend analysis on the user's emotional fluctuation curve based on the emotional mutation points to generate user emotional trend data; Step S54: Perform demand trend prediction on the dynamic behavior characteristic data through the user emotional trend data to generate demand trend prediction data; Step S55: Use the behavior demand trend prediction probability calculation formula to perform probability prediction evaluation on the demand trend prediction data to generate a behavior demand prediction evaluation index.
8. The method according to claim 7, characterized in that The specific behavior demand trend prediction probability calculation formula in Step S55 is as follows: Where P is the behavior demand prediction evaluation index, t is the time point of demand prediction, X is the predicted behavior characteristic weight index, Y is the user behavior adaptability parameter, Z is the behavior demand change rate, S is the decay rate of behavior demand over time, N is the emotional fluctuation amplitude value, d is the interest driving force index, and λ is the interest point fixation duration.
9. The method according to claim 1, characterized in that The specific steps of Step S6 are as follows: Step S61: Compare the behavior demand prediction evaluation index with a preset behavior demand prediction evaluation threshold index. When the preset behavior demand prediction evaluation threshold index is less than the behavior demand prediction evaluation index, switch the user's interest page through the user dynamic portrait to generate interest tendency switching data; Step S62: When the preset behavior demand prediction evaluation threshold index is greater than or equal to the behavior demand prediction evaluation index, optimize the interest tendency of the user dynamic portrait to generate interest tendency optimization data; Step S63: Use machine learning algorithms to perform interest switching decision analysis on the interest tendency switching data and the interest tendency optimization data to generate a user interest switching decision engine to execute the system switching operation.
10. A system switching system for a tutoring learning machine, characterized in that, A system switching method for a tutoring learning machine as described in claim 1, including: An emotional fluctuation module, configured to obtain the user's facial image and the user's behavior data; perform dynamic behavior analysis on the user's behavior data to generate dynamic behavior characteristic data; perform emotional fluctuation analysis on the user's facial image according to the dynamic behavior characteristic data to construct a user emotional fluctuation curve; An eye movement trajectory module, configured to perform learning state detection on the dynamic behavior characteristic data through the user's emotional fluctuation curve to generate learning state data; perform eye movement trajectory recognition on the user's facial image according to the learning state data to construct an eye movement trajectory map; An interest tendency module, configured to perform implicit association mining on the user's emotional fluctuation curve using the eye movement trajectory map to generate user behavior association data; perform user interest tendency analysis on the dynamic behavior characteristic data based on the user behavior association data to generate user interest tendency data; perform behavior tendency evolution processing on the user interest tendency data to construct a behavior interest tendency map; A dynamic portrait module, configured to perform personality difference analysis on the dynamic behavior characteristic data using the behavior interest tendency map to generate user personality difference data; construct a user portrait based on the user personality difference data to generate a user portrait; perform real-time dynamic optimization on the user portrait using the behavior interest tendency map to generate a user dynamic portrait; A behavior demand prediction module, which is used to perform emotional trend analysis on the user's emotional fluctuation curve through the user's dynamic portrait to generate user emotional trend data; and perform behavior demand prediction on the dynamic behavior feature data through the user emotional trend data to generate a behavior demand prediction evaluation index. A decision engine module, which is used to compare the behavior demand prediction evaluation index based on a preset behavior demand prediction evaluation threshold index to generate interest preference optimization data; and perform interest switching decision analysis on the interest preference switching data and the interest preference optimization data by using a deep learning algorithm to generate a user interest switching decision engine to execute system switching operations.
Citation Information
Patent Citations
Student comprehensive portrait label management system for online teaching based on deep learning
CN108492224A
Online classroom learning state analysis method
CN115797829A