Intelligent management system and method for interactive data of touch and talk pen

Abstract

The invention relates to the technical field of interactive data management, in particular to an intelligent management system and method for interactive data of a touch and talk pen. The method comprises the following steps: acquiring multi-dimensional acquisition data of the touch and talk pen; performing data preprocessing on the multi-dimensional acquisition data of the touch and talk pen to generate multi-dimensional preprocessing data of the touch and talk pen; performing multi-modal data association on the multi-dimensional preprocessed data of the touch and talk pen to generate multi-dimensional associated data of the touch and talk pen; feature refining extraction is carried out according to the multi-dimensional associated data of the touch and talk pen, and a refined feature data set of the touch and talk pen is generated; therefore, through multi-modal data association analysis and personalized learning model construction, the problems of traditional data isolation, exception recognition lag and extensive interaction management are solved, and the intelligent level of interaction data management of the touch and talk pen and the refinement degree of user experience are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of interactive data management, and in particular to an intelligent management system and method for interactive data of a reading pen. Background Art

[0002] The existing technology for data processing and analysis of point reading pens mostly relies on traditional data collection and analysis modes based on a single mode. For example, most systems only focus on the basic interactive data of the point reading pen, such as trajectory, acceleration and sound, but often ignore the multimodal correlation between the data, resulting in low data utilization efficiency. Secondly, most of the existing point reading pen behavior data analysis methods lack intelligent detection and early warning mechanisms for abnormal behaviors. Traditional threshold setting and rule matching methods are difficult to effectively identify complex interactive behavior patterns, especially under diversified and personalized user needs. Static rules often cannot accurately identify and predict abnormal behaviors of point reading pens. Therefore, the existing technology often has great limitations in identifying complex behavior patterns and providing personalized feedback. Furthermore, although the point reading pen device generates a large amount of multidimensional interactive data, the existing technology does not fully utilize the potential correlation and diversity in these data. Traditional multidimensional data analysis methods lack the ability to comprehensively analyze these multimodal data, resulting in limitations in the analysis results. In addition, in terms of equipment failure and abnormal behavior warnings, existing technologies generally rely on statically set thresholds and preset rules for judgment. This method is prone to misjudgment or omission when faced with changing usage scenarios, and it is difficult to deal with complex and diverse abnormal situations. Summary of the invention

[0003] Based on this, it is necessary to provide a system and method for intelligent management of interactive data of a reading pen to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a method for intelligent management of interactive data of a reading pen is provided, the method comprising the following steps:

[0005] Step S1: acquiring multi-dimensional data collected by the reading pen; performing data preprocessing on the multi-dimensional data collected by the reading pen to generate multi-dimensional preprocessed data of the reading pen;

[0006] Step S2: performing multimodal data association on the multidimensional preprocessed data of the reading pen to generate multidimensional associated data of the reading pen; performing feature refinement extraction based on the multidimensional associated data of the reading pen to generate a refined feature data set of the reading pen;

[0007] Step S3: Acquire the long-term interaction history data of the reading pen; construct a personalized learning behavior model based on the long-term interaction history data of the reading pen and the refined feature data set of the reading pen to generate a personalized model of the reading pen; perform abnormal data detection based on the personalized model of the reading pen to generate abnormal behavior warning data of the reading pen;

[0008] Step S4: Generate a warning report based on the abnormal behavior warning data of the reading pen, thereby completing the intelligent management of the reading pen interaction data.

[0009] The beneficial effect of the present invention is that the multi-dimensional collected data is standardized, which lays a unified data foundation for subsequent analysis and ensures the compatibility and consistency between different data sources. Through multimodal data association technology, data of different modes such as time, space, pressure, audio, etc. are efficiently integrated, and the logical alignment and internal connection mining between data levels are realized. At the same time, the refined features are further extracted to form a refined feature data set that can be efficiently analyzed. This process not only improves the efficiency of data utilization, but also significantly enhances the expressive ability of data, making the subsequent model construction more accurate. The long-term interactive historical data of the point reading pen is introduced, combined with the refined feature data set, to construct a personalized learning behavior model. The model makes full use of the dynamic association between historical behavior features and real-time data, and realizes the accurate identification and prediction of the point reading pen usage mode and abnormal behavior at the data level. The abnormal behavior warning data generated by this model can efficiently capture device anomalies and user behavior anomalies, providing an important basis for subsequent optimization. Finally, the intelligent report generated based on the abnormal behavior warning data not only includes the data analysis results, but also generates targeted optimization suggestions in combination with the data features, providing scientific support for the interactive management of the device and the improvement of user experience. Therefore, the present invention solves the problems of traditional data isolation, delayed abnormal recognition and extensive interaction management through multimodal data association analysis and personalized learning model construction, and improves the intelligence level of interactive data management of the reading pen and the refinement of user experience.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: Acquire multi-dimensional data collected by the reading pen;

[0012] Step S12: Calculate the interaction complexity based on the multi-dimensional data collected by the reading pen to generate the multi-dimensional interaction complexity of the reading pen; remove data redundancy from the multi-dimensional data collected by the reading pen based on the multi-dimensional interaction complexity of the reading pen to generate the reading pen screening data; perform filtering algorithm denoising on the reading pen screening data to generate the reading pen denoising data;

[0013] Step S13: performing pen data standardization processing on the denoised data of the reading pen to generate multi-dimensional pre-processed data of the reading pen, wherein the multi-dimensional pre-processed data of the reading pen includes a vibration signal of the reading pen, a gesture image of the reading pen, a trajectory data of the reading pen and an acceleration data of the reading pen.

[0014] The present invention realizes the comprehensive collection of vibration signals, gesture images, trajectory data and acceleration data by acquisition, providing a multi-modal and multi-level information basis for subsequent data analysis. The interaction complexity calculation is introduced to quantitatively evaluate the interaction of the point reading pen from the dimension of data complexity for the first time. This evaluation mechanism can not only accurately identify the highly redundant parts in the data, but also significantly reduce the complexity of data storage and processing by eliminating redundancy, thereby improving the operating efficiency of the system. At the same time, the filter algorithm is used to denoise the screened data, further improving the signal-to-noise ratio of the data and ensuring the accuracy and reliability of subsequent data analysis. Standardization processing provides a unified measurement scale for multimodal data, eliminating dimensional differences and noise interference between different data sources. The multidimensional pre-processed data generated by this process is optimized in quality and structure, so that vibration signals, gesture images, trajectory data and acceleration data can be efficiently integrated into the subsequent feature analysis and pattern mining process. On the whole, this series of processing methods significantly improves the quality and effectiveness of the point reading pen data, and at the same time lays a solid foundation for building a high-precision and high-robust application model.

[0015] Preferably, step S2 comprises the following steps:

[0016] Step S21: performing multi-modal data association on the multi-dimensional pre-processed data of the reading pen to generate multi-dimensional associated data of the reading pen;

[0017] Step S22: performing collaborative relationship evaluation based on the multi-dimensional associated data of the reading pen to generate collaborative relationship data of the reading pen;

[0018] Step S23: performing feature refinement extraction based on the point reading pen collaborative relationship data and the point reading pen multi-dimensional association data to generate a point reading pen refined feature data set.

[0019] The present invention uses multimodal data association technology to logically align and uniformly model multimodal information such as vibration signals, gesture images, trajectory data and acceleration data, effectively integrating data characteristics of different dimensions such as time, space and interaction mode. This process not only enhances the integrity and consistency of the data, but also provides a more comprehensive data foundation for subsequent analysis. Based on multimodal associated data, collaborative relationship evaluation is carried out to explore the potential interactions and dependencies between different data dimensions. This evaluation mechanism can identify the synergy and correlation characteristics between data, and provides a theoretical basis and practical reference for the deep understanding and comprehensive analysis of multidimensional data. Based on collaborative relationship data and combined with multimodal associated data for feature refinement and extraction, it can not only focus on the efficient capture of global features (such as trajectory flow and gesture mode), but also go deep into fine-grained features (such as microscopic changes in acceleration fluctuations and detailed descriptions of pen tip trajectories), and generate a refined feature data set with high recognition and high expression ability. Through the synergistic effect of fusion, evaluation and extraction, the entire process realizes the deep mining and optimized expression of the interactive data of the point-reading pen, significantly improving the analysis value and application potential of the data.

[0020] Preferably, step S21 includes the following steps:

[0021] Step S211: acquiring geographic location information data and point reading pen recording data; performing keyword analysis on the point reading pen recording data to generate point reading pen recording keyword data; performing geographic tag association on the point reading pen recording keyword data and geographic location information data to generate point reading pen recording-geolocation data;

[0022] Step S212: performing handwriting flow analysis on the gesture image of the reading pen to generate the gesture flow data of the reading pen; performing physical movement trend analysis on the gesture flow data of the reading pen and the vibration signal of the reading pen to generate the physical movement trend data of the reading pen;

[0023] Step S213: Perform multimodal data association on the physical movement trend data of the reading pen and the recording-geographic location data of the reading pen to generate the collaborative relationship data of the reading pen.

[0024] The present invention not only realizes the efficient extraction of language information, but also generates recording keyword data containing key semantic information by performing keyword analysis on the recording data of the point reading pen. The recording keyword data is associated with the geographical location information by geo-tagging to form the recording-geographic location data of the point reading pen, which effectively integrates the semantic information and spatial information and enhances the spatial semantic expression ability of the data. By analyzing the handwriting flow direction of the gesture image of the point reading pen, the macro pattern of the user's dynamic writing during the interaction process is captured, and the physical movement trend analysis is performed in combination with the vibration signal of the point reading pen to generate the physical movement trend data of the point reading pen, thereby revealing the physical characteristics and dynamic laws of the user's interactive behavior. An adaptive feature extraction mechanism is introduced to associate the physical movement trend data with the recording-geographic location data in a multimodal data manner, thereby realizing the deep fusion of language, space and motion data. The mechanism can intelligently identify the synergy and potential connection between multimodal data, and the generated synergistic relationship data is not only highly consistent, but also can reflect the multidimensional characteristics of the user's complex interactive behavior. Through this series of innovative processing methods, the analysis depth and expression ability of the interactive data of the point reading pen are improved, providing solid data support for the deep understanding of interactive behavior and the precise optimization of application scenarios.

[0025] Preferably, step S23 includes the following steps:

[0026] Step S231: extracting trajectory vibration features of the point reading pen trajectory data and the point reading pen acceleration data based on the Fourier transform frequency domain to generate the point reading pen trajectory vibration feature data; performing trajectory density analysis on the point reading pen trajectory vibration feature data, and drawing a dynamic heat map to generate a point reading pen trajectory heat map;

[0027] Step S232: based on the collaborative relationship data of the reading pen, the pressure distribution of the multi-dimensional correlation data of the reading pen is decoded to generate the pressure distribution data of the reading pen; the pause duration of the reading pen trajectory heat map is analyzed to generate the pause duration data of the reading pen;

[0028] Step S233: Based on the pen trajectory heat map, feature refinement and extraction are performed on the pen pause duration data and the pen pressure distribution data, and a causal relationship distribution map is constructed to generate a pen feature causal relationship map; feature set packaging is performed on the pen feature causal relationship map to generate a pen refined feature data set.

[0029] The present invention uses Fourier transform frequency domain to extract trajectory vibration features from the point reading pen trajectory data and acceleration data to generate trajectory vibration feature data. This process captures the high-frequency dynamic change information implicit in the user operation and provides more dimensional accurate data support for trajectory analysis. At the same time, through trajectory density analysis combined with dynamic heat map drawing, the generated point reading pen trajectory heat map intuitively shows the high-frequency operation area and distribution characteristics during the interaction process. Further combined with the point reading pen collaborative relationship data, the multi-dimensional correlation data is decoded for pressure distribution, and the generated pressure distribution data accurately reflects the spatial changes of mechanical characteristics during the user interaction process. In addition, by analyzing the pause duration of the trajectory heat map, the generated pause duration data can effectively reveal the potential patterns in the user's operation rhythm and behavior characteristics. By extracting the features of the trajectory heat map, pause duration data and pressure distribution data, a feature causal relationship distribution map is constructed, and the complex mechanism of the user's operation behavior is deeply analyzed from the causal relationship level. The point reading pen refined feature data set finally generated realizes efficient data organization and storage through feature set packaging, providing high-quality data resources for subsequent analysis and application. This process significantly improves the expressiveness, integration level and analytical value of data, supporting more accurate behavioral pattern mining and system optimization.

[0030] Preferably, step S3 comprises the following steps:

[0031] Step S31: Acquire the long-term interaction history data and real-time interaction data of the reading pen;

[0032] Step S32: constructing a personalized learning behavior model based on the long-term interaction history data of the reading pen and the refined feature data set of the reading pen to generate a personalized model of the reading pen;

[0033] Step S33: Perform data detection on the real-time interactive data of the reading pen based on the personalized model of the reading pen to generate abnormal behavior warning data of the reading pen.

[0034] The present invention provides rich and diverse basic data for data association and model training by acquiring the long-term interaction history data and real-time interaction data of the point reading pen, wherein the long-term history data reflects the long-term behavior pattern of user operation, while the real-time interaction data captures the instant operation dynamics, providing a basis for real-time analysis. The point reading pen is used to refine the feature data set and combine it with the long-term history data to build a personalized learning behavior model. The model achieves a deep understanding of the user's personalized operation habits by mining the behavioral feature rules implicit in the historical data and dynamically adapting it to the real-time data, thereby improving the prediction accuracy and applicability of the model. Based on the constructed personalized model, data detection is performed on the real-time interaction data, and abnormal behavior warning data is generated through the abnormal behavior recognition mechanism. This process can not only capture abnormal operations in user interaction in real time, but also effectively filter false alarm information to ensure the reliability and pertinence of the warning results. Overall, the present invention realizes deep mining and real-time monitoring of interactive behaviors by constructing a personalized model based on the collaboration of historical and real-time data, while improving the accuracy of abnormal behavior recognition, providing data-driven technical support for the intelligent management and personalized services of the point reading pen.

[0035] Preferably, step S32 includes the following steps:

[0036] Step S321: performing a fast cluster analysis on the refined feature data set of the reading pen to generate a feature data cluster of the reading pen;

[0037] Step S322: using the reading pen feature data cluster to define the clustering result behavior rule, and generating the reading pen clustering behavior rule;

[0038] Step S323: constructing a personalized learning behavior model for the long-term interaction history data of the reading pen based on the reading pen clustering behavior rules to generate a reading pen personalized model.

[0039] The present invention uses an unsupervised learning algorithm to perform cluster analysis on the detailed feature data set of the point reading pen, and structures the complex multidimensional data into feature data clusters with internal similarity and external distinction. This process effectively simplifies the dimension and complexity of the data, and at the same time reveals the inherent laws of the point reading pen operation behavior. Through this analysis, the behavior pattern originally implicit in the massive data is made explicit, providing a clear data basis for subsequent analysis. Clustering behavior rules are defined based on the point reading pen feature data cluster, and the data pattern is converted into easy-to-understand behavior rules. This rule can not only summarize the user behavior characteristics, but also convert discrete clustering results into behavior descriptions with practical application value, greatly improving the explanatory power and operability of the point reading pen behavior data. These behavior rules are used to deeply mine the long-term interaction history data of the point reading pen, and through the construction of a personalized learning behavior model, the individualized behavior characteristics of the user are captured, and the operation habits of different users are dynamically adapted. The model can be driven by data and continuously optimize its predictive and adaptive capabilities for user behavior, playing a key role in personalized services and abnormal behavior warnings.

[0040] Preferably, step S4 comprises the following steps:

[0041] Step S41: generating a warning strategy based on the abnormal behavior warning data of the reading pen, and generating a reading pen warning strategy;

[0042] Step S42: performing warning trend analysis according to the reading pen warning strategy to generate reading pen warning trend analysis data;

[0043] Step S43: Generate a report on the early warning trend analysis data of the reading pen, thereby completing the intelligent management of the reading pen interaction data.

[0044] The present invention converts abnormal behaviors extracted from historical interaction data and real-time monitoring into operable response strategies by generating early warning strategies based on abnormal behavior early warning data of the reading pen. This strategy generation process can not only identify potential risk behaviors, but also formulate appropriate response plans based on the laws and trends in the data, thereby improving the automation and accuracy of the management system. According to the early warning strategy of the reading pen, early warning trend analysis is performed, and the risk behaviors and their development trends that will appear in the future are predicted through dynamic monitoring of historical data and real-time data. Early warning trend analysis can effectively identify early signals of potential risks and provide data support for subsequent management decisions to ensure the timeliness and accuracy of early warning responses. By generating reports for early warning trend analysis data, complex early warning information is converted into an easy-to-understand report format, so that managers can quickly obtain key abnormal information and trend changes, and make decisions and adjustments accordingly. This report generation process enhances the operability and visualization effect of the system, so that managers can respond to emergencies more efficiently. Overall, the present invention not only improves the accuracy and response speed of the abnormal behavior early warning of the reading pen through comprehensive analysis and multi-level processing of early warning data, but also improves the intelligent management level of the interactive data of the reading pen, providing strong data support for personalized management in practical applications.

[0045] Preferably, step S41 includes the following steps:

[0046] Step S411: using the preset reading pen warning threshold to perform warning evaluation on the reading pen abnormal behavior warning data, and generate the reading pen warning evaluation data; generating a warning strategy for the reading pen warning evaluation data, and generating a reading pen warning strategy, wherein the reading pen warning strategy includes a reading pen high priority warning strategy and a reading pen low priority warning strategy;

[0047] Step S412: When the warning evaluation data of the reading pen is higher than or equal to the preset warning threshold of the reading pen, the machine code of the reading pen is recorded to generate the machine code data of the reading pen fault; the device fault source is traced and judged on the machine code data of the reading pen fault. If a serious fault is detected, the machine code data of the reading pen fault is silently processed and backed up. If no fault is detected, the machine code data of the reading pen fault is sent to the preset server through the cloud, thereby completing the high-priority warning strategy of the reading pen;

[0048] Step S413: When the reading pen warning evaluation data is lower than the preset reading pen warning threshold, the reading pen warning evaluation data is pushed with optimizable suggestion items, thereby completing the reading pen low priority warning strategy.

[0049] The present invention evaluates the abnormal behavior data of the reading pen by using a preset warning threshold, generates the reading pen warning evaluation data, and further generates warning strategies for high-priority and low-priority problems. This process effectively identifies potential problems of different degrees through an accurate evaluation model, thereby providing a clear basis for subsequent processing and decision-making. The high-priority warning strategy focuses on responding quickly to behaviors that cause serious equipment failures to ensure the stability of the equipment; the low-priority warning strategy focuses on making suggestions for potential optimizable problems, improving equipment performance through gradual optimization, and reducing the probability of failure. Based on the situation that the reading pen warning evaluation data is higher than or equal to the preset threshold, the system records the machine code of the reading pen and generates fault machine code data. Through the equipment fault tracing analysis, the system can effectively determine the nature of the equipment failure and perform different processing according to the severity of the failure. If a serious fault is detected, the system will perform silent processing and back up the data to prevent the fault from expanding; if there is no fault, the fault machine code data will be sent to the cloud for remote monitoring and storage. This processing mechanism can ensure that measures are taken quickly when the equipment fails, while ensuring the security and integrity of the data. When the warning evaluation data is lower than the preset threshold, the system will push optimization suggestions to the reading pen to further improve the performance of the device and continuously monitor the device status. This low-priority warning strategy prevents potential risks and ensures the long-term stable operation of the device by gradually optimizing the data.

[0050] In this specification, a point reading pen interactive data intelligent management system is provided, which is used to execute the above-mentioned point reading pen interactive data intelligent management method, and the point reading pen interactive data intelligent management system includes:

[0051] The multi-dimensional data collection and standardization module is used to obtain the multi-dimensional collection data of the point reading pen; perform data preprocessing on the multi-dimensional collection data of the point reading pen to generate multi-dimensional preprocessing data of the point reading pen;

[0052] The multimodal association and feature refinement module is used to perform multimodal data association on the multidimensional preprocessed data of the reading pen to generate multidimensional associated data of the reading pen; perform feature refinement extraction based on the multidimensional associated data of the reading pen to generate a refined feature data set of the reading pen;

[0053] The personalized learning and behavior detection module is used to obtain the long-term interaction history data of the reading pen; based on the long-term interaction history data of the reading pen and the reading pen's refined feature data set, a personalized learning behavior model is constructed to generate a personalized model of the reading pen; based on the personalized model of the reading pen, abnormal data detection is performed to generate abnormal behavior warning data of the reading pen;

[0054] The intelligent report generation and management module is used to generate early warning reports based on the abnormal behavior warning data of the reading pen, thereby completing the intelligent management of the interactive data of the reading pen.

[0055] The beneficial effect of the present invention is that the multi-dimensional collected data is standardized, which lays a unified data foundation for subsequent analysis and ensures the compatibility and consistency between different data sources. Through multimodal data association technology, data of different modes such as time, space, pressure, audio, etc. are efficiently integrated, and the logical alignment and internal connection mining between data levels are realized. At the same time, the refined features are further extracted to form a refined feature data set that can be efficiently analyzed. This process not only improves the efficiency of data utilization, but also significantly enhances the expressive ability of data, making the subsequent model construction more accurate. The long-term interactive historical data of the point reading pen is introduced, combined with the refined feature data set, to construct a personalized learning behavior model. The model makes full use of the dynamic association between historical behavior features and real-time data, and realizes the accurate identification and prediction of the point reading pen usage mode and abnormal behavior at the data level. The abnormal behavior warning data generated by this model can efficiently capture device anomalies and user behavior anomalies, providing an important basis for subsequent optimization. Finally, the intelligent report generated based on the abnormal behavior warning data not only includes the data analysis results, but also generates targeted optimization suggestions in combination with the data features, providing scientific support for the interactive management of the device and the improvement of user experience. Therefore, the present invention solves the problems of traditional data isolation, delayed abnormal recognition and extensive interaction management through multimodal data association analysis and personalized learning model construction, and improves the intelligence level of interactive data management of the reading pen and the refinement of user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A schematic diagram of the steps of a method for intelligent management of interactive data of a reading pen;

[0057] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0058] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0059] Figure 4 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0060] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0061] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0062] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0063] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. 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.

[0064] To achieve this, please refer to Figures 1 to 4 , a method for intelligent management of interactive data of a point reading pen, the method comprising the following steps:

[0065] Step S1: acquiring multi-dimensional data collected by the reading pen; performing data preprocessing on the multi-dimensional data collected by the reading pen to generate multi-dimensional preprocessed data of the reading pen;

[0066] Step S2: performing multimodal data association on the multidimensional preprocessed data of the reading pen to generate multidimensional associated data of the reading pen; performing feature refinement extraction based on the multidimensional associated data of the reading pen to generate a refined feature data set of the reading pen;

[0067] Step S3: Acquire the long-term interaction history data of the reading pen; construct a personalized learning behavior model based on the long-term interaction history data of the reading pen and the refined feature data set of the reading pen to generate a personalized model of the reading pen; perform abnormal data detection based on the personalized model of the reading pen to generate abnormal behavior warning data of the reading pen;

[0068] Step S4: Generate a warning report based on the abnormal behavior warning data of the reading pen, thereby completing the intelligent management of the reading pen interaction data.

[0069] The beneficial effect of the present invention is that the multi-dimensional collected data is standardized, which lays a unified data foundation for subsequent analysis and ensures the compatibility and consistency between different data sources. Through multimodal data association technology, data of different modes such as time, space, pressure, audio, etc. are efficiently integrated, and the logical alignment and internal connection mining between data levels are realized. At the same time, the refined features are further extracted to form a refined feature data set that can be efficiently analyzed. This process not only improves the efficiency of data utilization, but also significantly enhances the expressive ability of data, making the subsequent model construction more accurate. The long-term interactive historical data of the point reading pen is introduced, combined with the refined feature data set, to construct a personalized learning behavior model. The model makes full use of the dynamic association between historical behavior features and real-time data, and realizes the accurate identification and prediction of the point reading pen usage mode and abnormal behavior at the data level. The abnormal behavior warning data generated by this model can efficiently capture device anomalies and user behavior anomalies, providing an important basis for subsequent optimization. Finally, the intelligent report generated based on the abnormal behavior warning data not only includes the data analysis results, but also generates targeted optimization suggestions in combination with the data features, providing scientific support for the interactive management of the device and the improvement of user experience. Therefore, the present invention solves the problems of traditional data isolation, delayed abnormal recognition and extensive interaction management through multimodal data association analysis and personalized learning model construction, and improves the intelligence level of interactive data management of the reading pen and the refinement of user experience.

[0070] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of a step flow chart of a method for intelligently managing interactive data of a reading pen according to the present invention. In this example, the method for intelligently managing interactive data of a reading pen includes the following steps:

[0071] Step S1: acquiring multi-dimensional data collected by the reading pen; performing data preprocessing on the multi-dimensional data collected by the reading pen to generate multi-dimensional preprocessed data of the reading pen;

[0072] In an embodiment of the present invention, the multi-dimensional acquisition data of the point reading pen is obtained, covering multiple sensor data sources such as vibration signals, gesture images, trajectory data and acceleration data. These data sources collect various information of the point reading pen in real time during use through different sensor modules, providing rich data about the device status and user interaction. In the process of acquiring data, high-precision sensing technology and acquisition algorithms are used to ensure the comprehensiveness and accuracy of the data. Next, the acquired multi-dimensional acquisition data is preprocessed to improve the availability and accuracy of the data. The technical means of data preprocessing include data cleaning, denoising, outlier detection and other links. In the data cleaning stage, a threshold-based method is applied to remove invalid data and erroneous data to ensure data quality. In the denoising process, noise interference is eliminated by filtering algorithms (such as Kalman filtering, low-pass filtering, etc.) to ensure the purity and effectiveness of the signal. In terms of outlier detection, by setting reasonable judgment criteria and algorithms, such as methods based on standard deviation or quantiles, extremely abnormal data is detected and eliminated, further improving the accuracy and reliability of the data. In addition, data preprocessing also includes standardization, through methods such as mean normalization or maximum normalization, so that sensor data of different scales can be analyzed uniformly in the same data space. Through these preprocessing methods, the generated multi-dimensional preprocessed data of the reading pen can effectively provide a high-quality and reliable data foundation for subsequent analysis and model training, thereby laying a solid data support for the subsequent steps of behavior analysis, model training and anomaly detection. Through these data-level technical means, it is ensured that the reading pen device can minimize errors when performing high-precision interactive analysis, and improve the efficiency and stability of the entire system.

[0073] Step S2: performing multimodal data association on the multidimensional preprocessed data of the reading pen to generate multidimensional associated data of the reading pen; performing feature refinement extraction based on the multidimensional associated data of the reading pen to generate a refined feature data set of the reading pen;

[0074] In the embodiment of the present invention, multi-modal data association is performed on the multi-dimensional pre-processed data of the point reading pen. This process is achieved through a multi-modal learning method, which aims to effectively jointly analyze data from different sensors (such as vibration signals, gesture images, trajectory data, acceleration data, etc.). In this link, cross-modal data fusion algorithms are used, such as canonical correlation analysis (CCA) and joint embedding learning. These methods can map data of different modalities into the same shared space to generate multi-dimensional associated data of the point reading pen. In the specific implementation process, by setting similarity metrics, such as Euclidean distance, cosine similarity, etc., the relationship between different modal data is measured, so as to effectively mine the potential interactive features between each modality. Next, feature refinement and extraction are performed based on the multi-dimensional associated data of the point reading pen. This process is achieved through feature engineering methods, including technical means such as feature selection, dimensionality reduction, and encoding. Common technologies include principal component analysis (PCA), independent component analysis (ICA), etc. These methods can extract the most representative features from large-scale, multi-dimensional data. In the process of feature refinement, combined with the interactive scene of the reading pen, further features closely related to user behavior, such as frequency features, time series features, and spatial features, are identified, and the consistency and availability of feature data are improved through feature normalization and standardization. Finally, through feature refinement and extraction, a reading pen refined feature data set is generated. This data set has improved data dimensions, correlation, and expression capabilities, providing rich data support for subsequent personalized learning, behavior analysis, and anomaly detection.

[0075] Step S3: Acquire the long-term interaction history data of the reading pen; construct a personalized learning behavior model based on the long-term interaction history data of the reading pen and the refined feature data set of the reading pen to generate a personalized model of the reading pen; perform abnormal data detection based on the personalized model of the reading pen to generate abnormal behavior warning data of the reading pen;

[0076] In the embodiment of the present invention, the long-term interaction history data of the point reading pen is obtained, and these data include the long-term interaction records between the user and the point reading pen, such as the frequency of use, the interaction duration, the common functions, the operation mode and other information. These data are usually presented in the form of time series data, in which each interaction event contains data of multiple dimensions, such as timestamp, operation type, operation response and the like. In order to extract effective information from this large amount of historical data, a time series data processing method is adopted, including sequence model, sliding window technology and the like, so as to systematically analyze the user behavior within a long span. Next, a personalized learning behavior model is constructed based on the long-term interaction history data of the point reading pen and the refined feature data set of the point reading pen. At this stage, the recurrent neural network (RNN) and long short-term memory network (LSTM) in deep learning can be used to process and learn the long-term dependencies in the time series data. Through the training of historical data, the model can capture the user's behavioral patterns and interaction habits, thereby generating a personalized model of the point reading pen. In addition, in order to enhance the generalization ability of the model, an attention mechanism is introduced to focus on the most important part of the user's behavior through weight adjustment. Finally, abnormal data detection is performed based on the personalized model of the reading pen to generate abnormal behavior warning data of the reading pen. This process uses the online reasoning capability of the model to determine whether the current user behavior deviates from the predicted behavior of the personalized model by monitoring and detecting real-time interactive data. If a significant deviation is detected, abnormal behavior warning data is generated. Usually, such abnormal behavior indicates hardware failure, abnormal user operation, or system failure. In anomaly detection, threshold-based methods can be used, or anomaly detection algorithms such as isolation forest and one-class support vector machine can be used to determine whether the data exceeds the normal range, further improving the accuracy and timeliness of the warning.

[0077] Step S4: Generate a warning report based on the abnormal behavior warning data of the reading pen, thereby completing the intelligent management of the reading pen interaction data.

[0078] In the embodiment of the present invention, key information is first extracted from the abnormal behavior warning data of the reading pen, including data of dimensions such as the frequency, type, time period distribution, and impact range of abnormal behavior. One of the core technical means of data processing is data aggregation and feature extraction. Through the aggregation operation of abnormal data, such as grouping statistics by time period, functional module and other dimensions, the trend of abnormal behavior and its correlation with the user interaction mode can be clearly displayed. In addition, time series analysis techniques, such as autoregressive integrated moving average model (ARIMA) and seasonal trend decomposition model (STL), can also be used to predict the trend of abnormal data to further understand the occurrence of abnormal behavior in the future. In the process of generating reports, data visualization technology is also needed to intuitively display abnormal behavior. Common visualization methods include heat maps, line graphs, bar graphs, etc., especially when analyzing the frequency distribution, time series changes and geographical location information of abnormal behavior, heat maps and time series graphs can effectively highlight key information. These technical means combine data analysis with graphical presentation, so that the report is not only scientific and accurate, but also highly operable and easy to understand, so as to facilitate users, operation and maintenance personnel or decision makers to quickly judge and deal with potential problems. In the process of report generation, natural language generation (NLG) technology can also be combined to automatically generate highly readable text descriptions so that users at different levels can intuitively understand the report content. The text analysis content automatically generated by the algorithm can accurately describe the potential problems behind the data, warning conditions, and corresponding processing suggestions, further improving the intelligence level and practicality of the system.

[0079] Preferably, step S1 comprises the following steps:

[0080] Step S11: Acquire multi-dimensional data collected by the reading pen;

[0081] Step S12: Calculate the interaction complexity based on the multi-dimensional data collected by the reading pen to generate the multi-dimensional interaction complexity of the reading pen; remove data redundancy from the multi-dimensional data collected by the reading pen based on the multi-dimensional interaction complexity of the reading pen to generate the reading pen screening data; perform filtering algorithm denoising on the reading pen screening data to generate the reading pen denoising data;

[0082] Step S13: performing pen data standardization processing on the denoised data of the reading pen to generate multi-dimensional pre-processed data of the reading pen, wherein the multi-dimensional pre-processed data of the reading pen includes a vibration signal of the reading pen, a gesture image of the reading pen, a trajectory data of the reading pen and an acceleration data of the reading pen.

[0083] In an embodiment of the present invention, multi-dimensional data collected by the reading pen is obtained. The data is usually information collected by multiple sensors and input devices, including vibration signals, gesture images, trajectory data and acceleration data. In order to process these data, a series of data preprocessing steps are required to lay the foundation for subsequent analysis. The specific technical means are first embodied in the calculation of the interaction complexity of the multi-dimensional data collected by the reading pen. Interaction complexity is an important indicator to measure the diversity and complexity of the behavior of the user in the interaction process with the reading pen. This calculation method usually relies on the concept of information entropy. By quantifying factors such as the diversity and variability of user behavior, it can evaluate the complexity of the interaction and provide a basis for subsequent data screening. On the basis of generating the multi-dimensional interaction complexity of the reading pen, the next technical means is to eliminate data redundancy. Redundant data will affect the accuracy and efficiency of subsequent processing. Therefore, a method based on correlation analysis, such as the Pearson correlation coefficient or mutual information quantification, is used to identify and eliminate redundant data with high correlation. This process helps to reduce the redundancy of data, improve computing efficiency, and ensure the accuracy of data analysis. Subsequently, the data quality is further improved by denoising through a filtering algorithm. Common denoising techniques include low-pass filters and Kalman filters, which aim to remove data anomalies caused by environmental interference or sensor noise. The filtering process uses time domain or frequency domain analysis to effectively suppress the noise in the data to improve the quality of the signal. The denoised data of the reading pen is standardized to generate multi-dimensional pre-processed data of the reading pen. This process converts each data dimension (such as vibration signal, gesture image, trajectory data, acceleration data) to the same scale, eliminating the dimensional differences between different data dimensions, making subsequent analysis more unified and simple. Common methods of standardization include zero mean unit variance standardization (Z-score standardization) or minimum-maximum normalization (Min-Max normalization), which can effectively ensure the consistency of the data and provide high-quality input data for subsequent feature extraction and model building.

[0084] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0085] Step S21: performing multi-modal data association on the multi-dimensional pre-processed data of the reading pen to generate multi-dimensional associated data of the reading pen;

[0086] Step S22: performing collaborative relationship evaluation based on the multi-dimensional associated data of the reading pen to generate collaborative relationship data of the reading pen;

[0087] Step S23: performing feature refinement extraction based on the point reading pen collaborative relationship data and the point reading pen multi-dimensional association data to generate a point reading pen refined feature data set.

[0088] In an embodiment of the present invention, multi-modal data association is performed on the multi-dimensional pre-processed data of the reading pen to generate multi-dimensional associated data of the reading pen. The technical means of this process are mainly reflected in the implementation of multi-modal data fusion, and an alignment algorithm based on timestamps or spatial coordinates is usually adopted. The core of multi-modal data association technology is to align the data from different sensors (such as the vibration signal, gesture image, trajectory data and acceleration data of the reading pen) in time or space, so that data from different sources can be effectively integrated under a unified framework. Specifically, the dynamic time warping (Dynamic Time Warping, DTW) method can be used to process data with time delay, or deep learning technologies such as convolutional neural networks (CNN) can be used to extract features between each mode and fuse them. This association technology can effectively eliminate the time or space misalignment between data, so that multi-modal data can provide a comprehensive understanding of the behavior of the reading pen in a collaborative way. On the basis of generating the multi-dimensional associated data of the reading pen, the collaborative relationship evaluation that is performed next is to generate the collaborative relationship data of the reading pen by quantifying the interaction and correlation between each data mode. The evaluation of collaborative relationships usually uses methods based on correlation analysis, such as mutual information and Pearson correlation coefficient, to quantify the dependencies between different modalities. In addition, graph neural networks (GNNs) can also be used to construct collaborative relationship graphs between data modalities, and further reveal the potential synergistic effects between modalities through the connection relationship between nodes and edges. This technical means can effectively identify the potential connections between different data modalities, reveal their inherent collaborative mechanisms, and provide a basis for subsequent behavior pattern analysis and prediction. Based on the collaborative relationship data of the reading pen and the multi-dimensional association data of the reading pen, feature refinement and extraction are performed to generate a reading pen refined feature data set. Feature refinement and extraction is carried out by deeply mining and refining the potential information in each data modality. This process uses a variety of technical means, such as principal component analysis (PCA), independent component analysis (ICA) and other dimensionality reduction techniques to extract features with discrimination. In addition, deep learning methods, such as convolutional neural networks (CNN) or long short-term memory networks (LSTM), can also be used to further extract complex features in the data through network structures.

[0089] Preferably, step S21 includes the following steps:

[0090] Step S211: acquiring geographic location information data and point reading pen recording data; performing keyword analysis on the point reading pen recording data to generate point reading pen recording keyword data; performing geographic tag association on the point reading pen recording keyword data and geographic location information data to generate point reading pen recording-geolocation data;

[0091] Step S212: performing handwriting flow analysis on the gesture image of the reading pen to generate the gesture flow data of the reading pen; performing physical movement trend analysis on the gesture flow data of the reading pen and the vibration signal of the reading pen to generate the physical movement trend data of the reading pen;

[0092] Step S213: Perform multimodal data association on the physical movement trend data of the reading pen and the recording-geographic location data of the reading pen to generate the collaborative relationship data of the reading pen.

[0093] In an embodiment of the present invention, the acquired geographic location information data and the recording data of the reading pen are analyzed by keywords to generate the recording keyword data of the reading pen. This process uses the keyword extraction technology in natural language processing (NLP), such as TF-IDF (term frequency-inverse document frequency) or text embedding technology based on deep learning (such as Word2Vec, BERT, etc.) to extract the key information in the recording data of the reading pen. These key information can represent the interactive content of the reading pen and provide an effective basis for subsequent data analysis. Subsequently, the recording keyword data is combined with the geographic location information through geographic tag association. This process mainly relies on geographic information system (GIS) technology, and uses geographic coordinate data to associate keywords and geographic locations, thereby generating the recording-geographic location data of the reading pen. The key technical means of this association is based on the matching of location tags and voice keywords to ensure that the spatial information between different modal data can be effectively connected. The gesture image data of the reading pen generates the gesture flow data of the reading pen through the handwriting flow analysis. This process is usually the target detection and tracking technology in computer vision, and the convolutional neural network (CNN) or regional convolutional neural network (RCNN) in deep learning is used to dynamically analyze the handwriting flow in the gesture image. Flow analysis can help capture the trajectory changes of the reading pen during operation and reveal the details of the operation mode of the reading pen and the user interaction. Next, the gesture flow data and the vibration signal of the reading pen are analyzed for physical motion trend to generate the physical motion trend data of the reading pen. The processing of vibration signals usually uses signal processing technology, such as Fourier transform (FFT) or wavelet transform (Wavelet Transform) to analyze the vibration frequency and amplitude, so as to obtain the physical motion pattern of the reading pen under different operating states. Multimodal data association is performed on the physical motion trend data of the reading pen and the recording-geographic location data of the reading pen to generate the collaborative relationship data of the reading pen. The adaptive feature extraction mechanism mainly relies on the adaptive algorithm in machine learning, which can automatically select the most helpful features for the task and adjust them according to the dynamic changes of the data. For example, a deep neural network (DNN) or a long short-term memory network (LSTM) is used to learn complex features extracted from multimodal data, and multimodal learning techniques such as multi-task learning (MTL) or multi-view learning (MVL) are used to associate the relationship between different modalities.

[0094] Preferably, step S23 includes the following steps:

[0095] Step S231: extracting trajectory vibration features of the point reading pen trajectory data and the point reading pen acceleration data based on the Fourier transform frequency domain to generate the point reading pen trajectory vibration feature data; performing trajectory density analysis on the point reading pen trajectory vibration feature data, and drawing a dynamic heat map to generate a point reading pen trajectory heat map;

[0096] Step S232: based on the collaborative relationship data of the reading pen, the pressure distribution of the multi-dimensional correlation data of the reading pen is decoded to generate the pressure distribution data of the reading pen; the pause duration of the reading pen trajectory heat map is analyzed to generate the pause duration data of the reading pen;

[0097] Step S233: Based on the pen trajectory heat map, feature refinement and extraction are performed on the pen pause duration data and the pen pressure distribution data, and a causal relationship distribution map is constructed to generate a pen feature causal relationship map; feature set packaging is performed on the pen feature causal relationship map to generate a pen refined feature data set.

[0098] In an embodiment of the present invention, Fourier transform frequency domain analysis is used to extract vibration features from the trajectory data and acceleration data of the point reading pen. By converting the time domain signal into the frequency domain signal, Fourier transform can reveal the frequency components of the trajectory and acceleration and their intensity changes during different operations of the point reading pen. This process can help analyze the movement amplitude and frequency characteristics of the point reading pen and identify vibration anomalies or faults. The vibration feature extraction method usually relies on the spectrum analysis of the signal. By identifying the main frequency components of the signal, the motion characteristics of the point reading pen in different operation stages can be obtained. Next, the trajectory density analysis is performed on the vibration feature data of the point reading pen trajectory, and a point reading pen trajectory heat map is generated by drawing a dynamic heat map. The trajectory density analysis is usually based on the spatial position data of the point reading pen, combined with statistical methods such as density estimation (Kernel Density Estimation, KDE) or weighted average method to analyze the activity frequency of the point reading pen in different areas, thereby revealing the hot spots of the point reading pen during the interaction process. The dynamic heat map uses dynamic visualization technology to show the interaction density changes of the reading pen in different time periods and different spatial positions, helping users to intuitively understand the usage behavior patterns of the reading pen. Based on the collaborative relationship data of the reading pen, the pressure distribution of the multi-dimensional correlation data of the reading pen is decoded to generate the pressure distribution data of the reading pen. This process pressure sensing technology and data decoding technology can obtain the data of each sensor in the use of the reading pen in real time, and combine the existing behavior model of the reading pen to decode the pressure data through algorithms (such as Kalman filtering, least squares method, etc.), so as to generate the pressure distribution map of the reading pen. This pressure data provides a quantitative basis for the pressure change of the reading pen in different operating environments, which helps to understand the physical response of the device in various usage scenarios. Next, the pause duration analysis of the reading pen trajectory heat map is performed to generate the pause duration data of the reading pen. The pause duration analysis relies on dynamic time series analysis methods, such as sliding windows, sequence clustering, etc., which can accurately detect the pause state of the reading pen in each time period during the interaction process, thereby quantifying its stay time in a specific area. Based on the pen trajectory heat map, the pause duration data and the pen pressure distribution data are refined and extracted, and a causal relationship distribution map is constructed to generate a pen feature causal relationship map. Feature refinement and extraction usually uses feature engineering techniques, such as principal component analysis (PCA) and linear discriminant analysis (LDA) to reduce the dimension and optimize the data, and extract the most representative features from them. These features help describe the core behavior patterns of the pen. At the same time, the construction of the causal relationship distribution map relies on causal reasoning algorithms (such as Bayesian networks, Granger causality tests, etc.) to determine the causal relationship between different features and further analyze the behavior patterns and interaction logic of the pen. This process, by establishing a sophisticated causal relationship network, not only reveals the behavioral associations of the pen in the interaction process, but also helps predict future operation trends.Finally, the feature set of the causal relationship graph of the reading pen is packaged to generate a refined feature data set of the reading pen.

[0099] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0100] Step S31: Acquire the long-term interaction history data and real-time interaction data of the reading pen;

[0101] Step S32: constructing a personalized learning behavior model based on the long-term interaction history data of the reading pen and the refined feature data set of the reading pen to generate a personalized model of the reading pen;

[0102] Step S33: Perform data detection on the real-time interactive data of the reading pen based on the personalized model of the reading pen to generate abnormal behavior warning data of the reading pen.

[0103] In the embodiment of the present invention, by acquiring the long-term interaction history data of the reading pen and the real-time interaction data of the reading pen, the accumulation of the behavior patterns of the reading pen in multiple interactions is realized at the data level. These data include but are not limited to the user's operation trajectory, interaction frequency, pause duration, gesture recognition data, vibration signal, etc., and the interaction information between the reading pen and the user. Through the comprehensive collection of these data, a comprehensive behavior record of the reading pen in various interactive environments can be obtained. These data sources provide a sufficient basis for subsequent behavior modeling and anomaly detection, especially multi-dimensional data processing in diversified interactive scenarios. A personalized learning behavior model is constructed based on the long-term interaction history data of the reading pen and the refined feature data set of the reading pen. The technical means at this stage rely on machine learning and deep learning methods, such as regression models, classification algorithms, support vector machines (SVM) or neural networks in supervised learning. In this step, the behavior patterns and laws in the long-term interaction history are identified by training algorithms, and combined with the refined feature data set (including the vibration signal, gesture image, trajectory data, etc. of the reading pen), a personalized learning model that can reflect the user's behavior characteristics is gradually constructed. This process not only learns the operating habits of each user, but also can dig out the interaction rules between the reading pen and the user, such as operation frequency, interaction duration, interaction accuracy and other important factors. The establishment of a personalized learning model not only improves the ability to identify individual user behaviors, but also lays the foundation for the subsequent optimization of abnormal behavior detection and early warning systems. Abnormal data detection is performed based on the personalized model of the reading pen to generate abnormal behavior early warning data of the reading pen. The technical means in this step mainly rely on the previously constructed personalized behavior model to dynamically monitor and analyze real-time interaction data. By comparing real-time data with the personalized behavior model, anomaly detection algorithms (such as statistical outlier detection, deep learning-based autoencoders, etc.) are used to identify data points that deviate from the normal behavior pattern. These behaviors that deviate from the normal are abnormal behaviors, which are caused by factors such as equipment failure, operating errors or external interference. Through this anomaly detection mechanism, the system can generate early warning data in real time to remind users or managers of potential problems and provide data support for subsequent corrective measures.

[0104] Preferably, step S32 includes the following steps:

[0105] Step S321: performing a fast cluster analysis on the refined feature data set of the reading pen to generate a feature data cluster of the reading pen;

[0106] Step S322: using the reading pen feature data cluster to define the clustering result behavior rule, and generating the reading pen clustering behavior rule;

[0107] Step S323: constructing a personalized learning behavior model for the long-term interaction history data of the reading pen based on the reading pen clustering behavior rules to generate a reading pen personalized model.

[0108] In an embodiment of the present invention, a fast clustering cluster analysis is performed on the refined feature data set of the reading pen to generate a reading pen feature data cluster. Unsupervised learning algorithms, especially clustering algorithms (such as K-means, hierarchical clustering, DBSCAN, etc.), are used to discover potential structures and patterns from a large amount of unlabeled data. In this step, the clustering algorithm can automatically group the refined feature data of the reading pen (such as gesture trajectory, vibration signal, acceleration data, etc.) and classify samples with similar features into the same cluster. This process does not rely on prior labels and aims to spontaneously mine naturally distributed category information from the data. The clustered data clusters provide a basis for subsequent behavioral rule definition and personalized model construction. By extracting the center point and boundary of each cluster, the potential patterns and feature differences of the data can be clearly identified. The clustering result behavior rule definition is performed using the reading pen feature data cluster to generate the reading pen cluster behavior rules. In this stage, the behavior rules of the reading pen are defined based on the common characteristics of the members in the cluster through the analysis of the clustering results. For example, a certain characteristic data cluster corresponds to the user's operating habits in a specific situation, and another cluster represents the response mode of the device in a certain environment. Through the definition of clustering result behavior rules, the system can establish rules for each type of behavior pattern to characterize its characteristic attributes, such as operation frequency, interaction duration, vibration amplitude, etc., so as to accurately classify and analyze the use behavior of the reading pen. These behavior rules help the system understand the normal behavior range of the device, and then identify abnormal operations or behavioral deviations, which become an important basis for further optimizing the personalized model. Based on the clustering behavior rules of the reading pen, a personalized learning behavior model is constructed for the long-term interaction history data of the reading pen to generate a personalized model of the reading pen. The technical means of this stage mainly rely on the combination of previously defined behavior rules and the long-term interaction history data of the reading pen, and construct a personalized learning behavior model through algorithms (such as decision trees, support vector machines, neural networks, etc. in supervised learning). The construction of the personalized learning model automatically adapts to individual differences through the analysis of user historical behavior data, thereby generating a behavior prediction model for specific users or scenarios.

[0109] As an example of the present invention, refer to Figure 4 As shown, in this example, step S4 includes:

[0110] Step S41: generating a warning strategy based on the abnormal behavior warning data of the reading pen, and generating a reading pen warning strategy;

[0111] Step S42: performing warning trend analysis according to the reading pen warning strategy to generate reading pen warning trend analysis data;

[0112] Step S43: Generate a report on the early warning trend analysis data of the reading pen, thereby completing the intelligent management of the reading pen interaction data.

[0113] In the embodiment of the present invention, the warning strategy is generated based on the abnormal behavior warning data of the reading pen. This process mainly relies on anomaly detection technology and rule-based strategy generation methods. The abnormal behavior warning data of the reading pen is obtained by continuously monitoring real-time data, combining historical interaction data, device behavior patterns and defined behavior rules to identify abnormal patterns or behavior deviations in device operation. Common anomaly detection methods include statistical methods (such as mean deviation, standard deviation, box plot, etc.), machine learning algorithms (such as K-means clustering, isolation forest algorithm, etc.), and autoencoders in deep learning, etc. These methods can identify behaviors that do not conform to predetermined rules or exceed normal ranges from massive data. Based on these detection results, the system will generate corresponding warning strategies. The generation threshold setting, risk assessment and emergency response strategy design of the warning strategy ensure that the device can issue an alarm in time and take corresponding measures when an abnormality is detected, such as alarming, recording fault information or adjusting the operation mode. According to the warning strategy of the reading pen, the warning trend analysis is performed to generate the warning trend analysis data of the reading pen. At this stage, it mainly relies on time series analysis, trend prediction model and data visualization technology. By analyzing the long-term accumulated warning data, the system can identify the frequency, duration, type distribution and pattern changes of abnormal behaviors, so as to predict the warning trends in the future. Common time series analysis methods include autoregressive model (AR), moving average model (MA), ARIMA model, etc., which can model time series data and predict future trends. Combined with machine learning prediction technology (such as deep learning models such as long short-term memory network LSTM), the system can achieve accurate trend prediction based on historical data. In addition, data visualization technology (such as trend curves, heat maps, radar maps, etc.) plays a role in intuitively displaying warning trends in this process, helping decision makers understand system behavior and risk changes. Reports are generated for the warning trend analysis data of the reading pen, thereby completing the intelligent management of the reading pen interactive data. In this process, report generation data mining and natural language generation (NLG) technology. Based on the warning trend analysis data, the system will automatically generate reports containing key information. These reports can not only show the trend of the data, but also provide decision support based on the implementation of the warning strategy. The technical means of report generation include data summary, intelligent reasoning and natural language generation. The system will automatically extract relevant information based on the analysis results to form a text report that is both academic and operational.

[0114] Preferably, step S41 includes the following steps:

[0115] Step S411: using the preset reading pen warning threshold to perform warning evaluation on the reading pen abnormal behavior warning data, and generate the reading pen warning evaluation data; generating a warning strategy for the reading pen warning evaluation data, and generating a reading pen warning strategy, wherein the reading pen warning strategy includes a reading pen high priority warning strategy and a reading pen low priority warning strategy;

[0116] Step S412: When the warning evaluation data of the reading pen is higher than or equal to the preset warning threshold of the reading pen, the machine code of the reading pen is recorded to generate the machine code data of the reading pen fault; the device fault source is traced and judged on the machine code data of the reading pen fault. If a serious fault is detected, the machine code data of the reading pen fault is silently processed and backed up. If no fault is detected, the machine code data of the reading pen fault is sent to the preset server through the cloud, thereby completing the high-priority warning strategy of the reading pen;

[0117] Step S413: When the reading pen warning evaluation data is lower than the preset reading pen warning threshold, the reading pen warning evaluation data is pushed with optimizable suggestion items, thereby completing the reading pen low priority warning strategy.

[0118] In the embodiment of the present invention, the warning data of abnormal behavior of the reading pen is evaluated by using the preset warning threshold of the reading pen, which is realized by the threshold judgment algorithm. The threshold evaluation method is usually based on statistical analysis technology, such as standard deviation method, quantile method, etc., and different levels of warning thresholds are set by in-depth analysis of historical warning data. These thresholds are used to determine whether the abnormal behavior of the reading pen exceeds the normal range. Once the warning data of the reading pen exceeds these set thresholds, the system will generate warning evaluation data and generate warning strategies accordingly. The strategy is divided into high priority and low priority based on the severity of the warning. Usually, high priority warnings are associated with more urgent problems, while low priority warnings focus on less serious abnormal behaviors and take different response measures. The technical means of generating strategies are rule engines and decision tree algorithms, which perform rule matching and strategy selection based on the severity of the data. When the warning evaluation data of the reading pen is higher than or equal to the preset threshold, the system will record the machine code of the reading pen and generate fault machine code data. The fault machine code record uses data storage technology to store the device identifier, error code and other information of each abnormal behavior in the database. When tracing the source of equipment faults, the system will use fault diagnosis algorithms, such as anomaly detection methods based on pattern recognition, or classification algorithms in machine learning (such as support vector machines, random forests, etc.) to classify and diagnose the fault type. If a serious fault is detected, the system will perform silent processing, that is, temporarily freeze and back up the machine code data to prevent the fault information from spreading further or affecting other functions of the device. At the same time, these data will be backed up and stored for subsequent retrieval and analysis. If no fault is detected, the fault machine code data will be sent to the preset server through the cloud platform for further processing or storage. This process relies on cloud communication technology and network protocols (such as HTTP, MQTT, etc.) to achieve remote data transmission. When the warning evaluation data of the reading pen is lower than the preset threshold, the system will generate optimizable suggestions and push them to the user. This process recommends the system and the intelligent push algorithm. The recommendation system is usually based on data mining and machine learning technology, and generates targeted optimization suggestions by analyzing user historical behavior and warning data. These suggestions can include improvements to the operation behavior of the reading pen, optimization tips for device use, etc., aiming to reduce potential low-priority abnormalities in the future. Push technology relies on mechanisms such as message queues and push services to deliver optimization suggestions to users in a timely manner through the cloud platform.

[0119] In this specification, a point reading pen interactive data intelligent management system is provided, which is used to execute the above-mentioned point reading pen interactive data intelligent management method. The point reading pen interactive data intelligent management system:

[0120] The multi-dimensional data collection and standardization module is used to obtain the multi-dimensional collection data of the point reading pen; perform data preprocessing on the multi-dimensional collection data of the point reading pen to generate multi-dimensional preprocessing data of the point reading pen;

[0121] The multimodal association and feature refinement module is used to perform multimodal data association on the multidimensional preprocessed data of the reading pen to generate multidimensional associated data of the reading pen; perform feature refinement extraction based on the multidimensional associated data of the reading pen to generate a refined feature data set of the reading pen;

[0122] The personalized learning and behavior detection module is used to obtain the long-term interaction history data of the reading pen; based on the long-term interaction history data of the reading pen and the reading pen's refined feature data set, a personalized learning behavior model is constructed to generate a personalized model of the reading pen; based on the personalized model of the reading pen, abnormal data detection is performed to generate abnormal behavior warning data of the reading pen;

[0123] The intelligent report generation and management module is used to generate early warning reports based on the abnormal behavior warning data of the reading pen, thereby completing the intelligent management of the interactive data of the reading pen.

[0124] The beneficial effect of the present invention is that the multi-dimensional collected data is standardized, which lays a unified data foundation for subsequent analysis and ensures the compatibility and consistency between different data sources. Through multimodal data association technology, data of different modes such as time, space, pressure, audio, etc. are efficiently integrated, and the logical alignment and internal connection mining between data levels are realized. At the same time, the refined features are further extracted to form a refined feature data set that can be efficiently analyzed. This process not only improves the efficiency of data utilization, but also significantly enhances the expressive ability of data, making the subsequent model construction more accurate. The long-term interactive historical data of the point reading pen is introduced, combined with the refined feature data set, to construct a personalized learning behavior model. The model makes full use of the dynamic association between historical behavior features and real-time data, and realizes the accurate identification and prediction of the point reading pen usage mode and abnormal behavior at the data level. The abnormal behavior warning data generated by this model can efficiently capture device anomalies and user behavior anomalies, providing an important basis for subsequent optimization. Finally, the intelligent report generated based on the abnormal behavior warning data not only includes the data analysis results, but also generates targeted optimization suggestions in combination with the data features, providing scientific support for the interactive management of the device and the improvement of user experience. Therefore, the present invention solves the problems of traditional data isolation, delayed abnormal recognition and extensive interaction management through multimodal data association analysis and personalized learning model construction, and improves the intelligence level of interactive data management of the reading pen and the refinement of user experience.

[0125] 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.

[0126] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may 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 the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for intelligent management of interactive data of a point reading pen, characterized in that: The following steps are involved: Step S1: Acquire multi-dimensional data collected by the reading pen; Performing data preprocessing on the multi-dimensional collected data of the reading pen to generate multi-dimensional preprocessed data of the reading pen; Step S2: performing multi-modal data association on the multi-dimensional pre-processed data of the reading pen to generate multi-dimensional associated data of the reading pen; Perform feature refinement and extraction based on the multi-dimensional correlation data of the reading pen to generate a reading pen refinement feature data set; Step S3: Acquire the long-term interaction history data of the reading pen; construct a personalized learning behavior model based on the long-term interaction history data of the reading pen and the refined feature data set of the reading pen to generate a personalized model of the reading pen; Perform abnormal data detection based on the personalized model of the reading pen to generate abnormal behavior warning data of the reading pen; Step S4: Generate a warning report based on the abnormal behavior warning data of the reading pen, thereby completing the intelligent management of the reading pen interaction data.

2. The method for intelligent management of interactive data of a point reading pen according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire multi-dimensional data collected by the reading pen; Step S12: Calculate the interaction complexity based on the multi-dimensional data collected by the reading pen to generate the multi-dimensional interaction complexity of the reading pen; remove data redundancy from the multi-dimensional data collected by the reading pen based on the multi-dimensional interaction complexity of the reading pen to generate the reading pen screening data; perform filtering algorithm denoising on the reading pen screening data to generate the reading pen denoising data; Step S13: performing pen data standardization processing on the denoised data of the reading pen to generate multi-dimensional pre-processed data of the reading pen, wherein the multi-dimensional pre-processed data of the reading pen includes a vibration signal of the reading pen, a gesture image of the reading pen, a trajectory data of the reading pen and an acceleration data of the reading pen.

3. The method for intelligent management of interactive data of a point reading pen according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing multi-modal data association on the multi-dimensional pre-processed data of the reading pen to generate multi-dimensional associated data of the reading pen; Step S22: performing collaborative relationship evaluation based on the multi-dimensional associated data of the reading pen to generate collaborative relationship data of the reading pen; Step S23: performing feature refinement extraction based on the point reading pen collaborative relationship data and the point reading pen multi-dimensional association data to generate a point reading pen refined feature data set.

4. The method for intelligent management of interactive data of a point reading pen according to claim 1, characterized in that: Step S21 includes the following steps: Step S211: acquiring geographic location information data and point reading pen recording data; performing keyword analysis on the point reading pen recording data to generate point reading pen recording keyword data; performing geographic tag association on the point reading pen recording keyword data and geographic location information data to generate point reading pen recording-geolocation data; Step S212: performing handwriting flow analysis on the gesture image of the reading pen to generate the gesture flow data of the reading pen; performing physical movement trend analysis on the gesture flow data of the reading pen and the vibration signal of the reading pen to generate the physical movement trend data of the reading pen; Step S213: Perform multimodal data association on the physical movement trend data of the reading pen and the recording-geographic location data of the reading pen to generate the collaborative relationship data of the reading pen.

5. The method for intelligent management of interactive data of a point reading pen according to claim 1, characterized in that: Step S23 includes the following steps: Step S231: extracting trajectory vibration features of the point reading pen trajectory data and the point reading pen acceleration data based on the Fourier transform frequency domain to generate the point reading pen trajectory vibration feature data; performing trajectory density analysis on the point reading pen trajectory vibration feature data, and drawing a dynamic heat map to generate a point reading pen trajectory heat map; Step S232: based on the collaborative relationship data of the reading pen, the pressure distribution of the multi-dimensional correlation data of the reading pen is decoded to generate the pressure distribution data of the reading pen; the pause duration of the reading pen trajectory heat map is analyzed to generate the pause duration data of the reading pen; Step S233: Based on the pen trajectory heat map, feature refinement and extraction are performed on the pen pause duration data and the pen pressure distribution data, and a causal relationship distribution map is constructed to generate a pen feature causal relationship map; feature set packaging is performed on the pen feature causal relationship map to generate a pen refined feature data set.

6. The method for intelligent management of interactive data of a point reading pen according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: Acquire the long-term interaction history data and real-time interaction data of the reading pen; Step S32: constructing a personalized learning behavior model based on the long-term interaction history data of the reading pen and the refined feature data set of the reading pen to generate a personalized model of the reading pen; Step S33: Perform data detection on the real-time interactive data of the reading pen based on the personalized model of the reading pen to generate abnormal behavior warning data of the reading pen.

7. The method for intelligent management of interactive data of a point reading pen according to claim 6, characterized in that: Step S32 includes the following steps: Step S321: performing a fast cluster analysis on the refined feature data set of the reading pen to generate a feature data cluster of the reading pen; Step S322: using the reading pen feature data cluster to define the clustering result behavior rule, and generating the reading pen clustering behavior rule; Step S323: constructing a personalized learning behavior model for the long-term interaction history data of the reading pen based on the reading pen clustering behavior rules to generate a reading pen personalized model.

8. The method for intelligent management of interactive data of a point reading pen according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: generating a warning strategy based on the abnormal behavior warning data of the reading pen, and generating a reading pen warning strategy; Step S42: performing warning trend analysis according to the reading pen warning strategy to generate reading pen warning trend analysis data; Step S43: Generate a report on the early warning trend analysis data of the reading pen, thereby completing the intelligent management of the reading pen interaction data.

9. The method for intelligent management of interactive data of a point reading pen according to claim 8, characterized in that: Step S41 includes the following steps: Step S411: using the preset reading pen warning threshold to perform warning evaluation on the reading pen abnormal behavior warning data, and generate the reading pen warning evaluation data; generating a warning strategy for the reading pen warning evaluation data, and generating a reading pen warning strategy, wherein the reading pen warning strategy includes a reading pen high priority warning strategy and a reading pen low priority warning strategy; Step S412: When the warning evaluation data of the reading pen is higher than or equal to the preset warning threshold of the reading pen, the machine code of the reading pen is recorded to generate the machine code data of the reading pen fault; the device fault source is traced and judged on the machine code data of the reading pen fault. If a serious fault is detected, the machine code data of the reading pen fault is silently processed and backed up. If no fault is detected, the machine code data of the reading pen fault is sent to the preset server through the cloud, thereby completing the high-priority warning strategy of the reading pen; Step S413: When the reading pen warning evaluation data is lower than the preset reading pen warning threshold, the reading pen warning evaluation data is pushed with optimizable suggestion items, thereby completing the reading pen low priority warning strategy.

10. An intelligent management system for interactive data of a point reading pen, characterized in that: The method for intelligently managing interactive data of a reading pen according to claim 1 comprises: The multi-dimensional data collection and standardization module is used to obtain the multi-dimensional collection data of the point reading pen; perform data preprocessing on the multi-dimensional collection data of the point reading pen to generate multi-dimensional preprocessing data of the point reading pen; The multimodal association and feature refinement module is used to perform multimodal data association on the multidimensional preprocessed data of the reading pen to generate multidimensional associated data of the reading pen; perform feature refinement extraction based on the multidimensional associated data of the reading pen to generate a refined feature data set of the reading pen; The personalized learning and behavior detection module is used to obtain the long-term interaction history data of the reading pen; based on the long-term interaction history data of the reading pen and the reading pen's refined feature data set, a personalized learning behavior model is constructed to generate a personalized model of the reading pen; based on the personalized model of the reading pen, abnormal data detection is performed to generate abnormal behavior warning data of the reading pen; The intelligent report generation and management module is used to generate early warning reports based on the abnormal behavior warning data of the reading pen, thereby completing the intelligent management of the interactive data of the reading pen.

Images