Human-computer interaction equipment and control method thereof

Abstract

The invention relates to the technical field of man-machine interaction data, in particular to man-machine interaction equipment and a control method thereof. The method comprises the following steps: acquiring equipment screen operation data, performing time window segmentation, generating a discrete touch event sequence and a dynamic touch feature sequence, clustering and processing the features by utilizing interaction features, forming user operation behavior grouped data, and combining equipment historical interaction data to form a user operation behavior group. Frequent item set mining is carried out to identify a high-frequency operation mode, a user operation preference model is constructed, user operation habits are identified, redundant actions are eliminated, main interaction action data are extracted, interaction intention prediction is carried out on main interaction actions based on historical data of equipment, and dynamic response is carried out on the man-machine interaction equipment according to a prediction result. And effective control is realized. According to the invention, a more personalized and smoother man-machine interaction equipment control method is realized.

Description

Technical Field

[0001] The present invention relates to the field of human-computer interaction data technology, and in particular to a human-computer interaction device and a control method thereof. Background Art

[0002] In modern human-computer interaction systems, the fluency and accuracy of user operations directly affect the user experience of the device. With the popularization of smart devices, how to efficiently identify user intentions and respond in real time has become an important challenge. Existing interaction methods often rely on active user input and lack intelligent prediction and adaptation capabilities, which leads to user fatigue and frustration during use, thus affecting the overall performance of the device. Although some studies have attempted to use machine learning and data analysis techniques to improve the interactive experience, there are still obvious deficiencies in practical applications. Many systems fail to fully consider individual differences among users, resulting in the inability of general models to accurately capture the operating habits and preferences of different users. In addition, when processing dynamic touch data, existing methods often ignore the temporal characteristics and diversity of touch behavior, which limits the accurate prediction and understanding of user intentions. Summary of the invention

[0003] Based on this, it is necessary to provide a human-computer interaction device and a control method thereof to solve at least one of the above technical problems.

[0004] To achieve the above object, a control method for a human-computer interaction device comprises the following steps:

[0005] Step S1: acquiring device screen operation data; performing time window segmentation on the device screen operation data to obtain a discrete touch event sequence; performing touch acceleration calculation on the discrete touch event sequence to generate a dynamic touch feature sequence;

[0006] Step S2: performing interactive feature clustering processing on the dynamic touch feature sequence to obtain an interactive feature clustering result set; grouping the interactive feature clustering result set into operation behavior categories to generate user operation behavior grouping data;

[0007] Step S3: Acquire historical device interaction data; perform frequent item set mining on user operation behavior grouping data based on the historical device interaction data to obtain high-frequency operation patterns; construct an operation preference model for the high-frequency operation patterns to generate a user operation preference model;

[0008] Step S4: using the user operation preference model to identify the user operation habits of the dynamic touch feature sequence to generate user operation habit data; based on the user operation habit data, removing redundant actions from the dynamic touch feature sequence to obtain main interactive action data;

[0009] Step S5: predict the interaction intention of the main interaction action data based on the historical interaction data of the device to obtain predicted interaction intention data; and dynamically respond to the human-computer interaction device based on the predicted interaction intention data to execute the control method of the human-computer interaction device.

[0010] The present invention reduces redundant operations of users when using devices by real-time identification and prediction of user operation intentions, thereby significantly improving the fluency and accuracy of interaction. This intuitive interaction method enables users to use devices more freely, reduces learning curves and operation difficulties, and thus improves overall user satisfaction. By establishing a user operation preference model and performing habit recognition, the system can respond to specific needs and habits of different users in a personalized manner. Each user's interaction history data provides a unique perspective for the system, enabling the device to accurately adjust its interaction mode to meet the personalized needs of users, and enhance the user's sense of belonging and stickiness. Redundant actions are eliminated in the dynamic touch feature sequence, and by identifying common redundant operations of users, false touches and unnecessary operation steps are reduced, which not only improves the efficiency of operation, but also reduces the frustration of users caused by incorrect operations, making the interaction process smoother. Frequent item set mining is performed using the device's historical interaction data to extract high-frequency operation modes, so that the system can optimize the control strategy based on real data analysis results. This data-driven method not only improves the accuracy of control, but also enables the device to make smarter decisions in complex environments. Based on the predicted interaction intention data, the device can perform dynamic interaction responses to achieve real-time feedback on user operations. This rapid adaptability ensures that the device can respond quickly when user intent changes, improves the flexibility and real-time nature of the interaction, and makes the device more intelligent to meet the ever-changing needs of users.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: Acquire a device screen sensor; monitor screen contact actions according to the device screen sensor to obtain device screen operation data;

[0013] Step S12: performing noise filtering on the device screen operation data to obtain filtered screen operation data; performing time window segmentation on the filtered screen operation data to obtain a discrete touch event sequence;

[0014] Step S13: normalizing the coordinates of the discrete touch event sequence to obtain a standardized touch coordinate set; performing time difference processing on the standardized touch coordinate set to generate a touch velocity sequence;

[0015] Step S14: smoothing the touch speed sequence to obtain a smooth speed curve;

[0016] Step S15: extracting features from the smooth velocity curve to obtain a dynamic touch feature sequence.

[0017] The present invention ensures that the acquired data is highly timely and accurate by monitoring the real-time contact actions of the device screen, providing a reliable basis for subsequent analysis. Through noise filtering processing, the signal-to-noise ratio of the screen operation data is significantly improved, the error caused by environmental interference is reduced, and the validity of the data is ensured. The operation data is discretized by time window segmentation, making the analysis of each touch event clearer, which is conducive to identifying operation modes and behavioral characteristics. Through coordinate normalization, the consistency of touch events on different devices or screen sizes is ensured, and the comparability of data is enhanced. Time difference processing effectively extracts touch speed information, providing a quantitative basis for understanding user interaction habits and behaviors. Smoothing processing reduces the noise in the touch speed sequence, provides a more stable speed curve, and enhances the accuracy and reliability of subsequent analysis. The feature extraction step finally generates a dynamic touch feature sequence, which lays a solid foundation for user behavior recognition and interaction intention prediction, and improves the overall intelligence level of human-computer interaction.

[0018] Preferably, step S2 comprises the following steps:

[0019] Step S21: performing feature vectorization processing on the dynamic touch feature sequence to obtain a feature vector set;

[0020] Step S22: Perform initial feature vector clustering on the feature vector set to generate initial clustering results:

[0021] Step S23: performing detailed calculation of the initial clustering result to obtain the optimal clustering number; performing feature vector clustering on the feature vector set based on the optimal clustering number to generate the optimal clustering result;

[0022] Step S24: performing label statistical analysis on the preferred clustering results to obtain main operation behavior category data; grouping the preferred clustering results into operation behavior categories according to the main operation category data to generate user operation behavior grouping data.

[0023] The present invention converts the dynamic touch feature sequence into a numerical representation through feature vectorization processing, which enhances the computability and analysis ability of data processing and provides a basis for subsequent clustering. The initial clustering step effectively identifies potential behavior patterns by clustering the feature vector set, thereby promoting the understanding of user operation behavior. The profile detail calculation provides a quantitative evaluation of the clustering effect, ensuring that the number of selected clusters is optimal, thereby improving the effectiveness and accuracy of the clustering results. The generation of optimal clustering results can more accurately divide the user's operation behavior, lay the foundation for subsequent behavior analysis, and is conducive to exploring the real needs of users. Label statistical analysis conducts an in-depth analysis of the clustering results, helps identify the main operation behavior categories, and enhances the understanding of user operation characteristics and classification accuracy. The generation of user operation behavior grouping data promotes the formulation of personalized services and response strategies, and helps to improve the efficiency of human-computer interaction and user experience.

[0024] Preferably, step S3 comprises the following steps:

[0025] Step S31: Acquire historical interaction data of the device; perform standard formatting on the historical interaction data of the device to obtain standardized historical data;

[0026] Step S32: multi-dimensionally integrate the normalized historical data and the user operation behavior grouping data to generate a comprehensive interaction data set;

[0027] Step S33: performing frequent item set mining on the comprehensive interaction data set to obtain candidate frequent item sets;

[0028] Step S34: Calculate the support of the candidate frequent item sets to generate a support data set; Calculate the confidence of the candidate frequent item sets to generate a confidence data set;

[0029] Step S35: screening the candidate frequent item sets for high-frequency patterns according to the support data set and the confidence data set to obtain high-frequency operation patterns;

[0030] Step S36: Perform association rule modeling on high-frequency operation modes to generate a user operation preference model.

[0031] The present invention ensures the consistency and availability of historical interaction data through standardized formatting, provides a reliable foundation for subsequent data analysis, and enhances the effectiveness of data integration. Multidimensional fusion technology combines historical interaction data with user operation behavior grouping data to form a comprehensive interaction data set, so that the analysis coverage is wider and can more comprehensively reflect user behavior. Frequent item set mining identifies operation patterns that frequently appear in user interactions, helps to extract key behavior features, and provides support for understanding user habits. Support and confidence calculations provide a quantitative basis for the effectiveness of high-frequency patterns, ensuring that the high-frequency operation patterns screened out have practical significance and relevance. High-frequency pattern screening promotes the accurate identification of user operation behaviors, making the subsequent construction of user preference models more in line with actual conditions and improving the reliability of the model. Association rule modeling helps to establish a user operation preference model and form a personalized user portrait, thereby providing targeted optimization suggestions and improvement strategies for human-computer interaction.

[0032] Preferably, step S33 includes the following steps:

[0033] Step S331: dividing the comprehensive interaction data set into time dimensions to obtain independent time series subsets; classifying the independent time series subsets into behavior types to obtain a classified interaction data set;

[0034] Step S332: performing serial number conversion on the classified interaction data set to obtain a numbered sequence data set; performing preliminary segmentation processing on the coded interaction data set to generate a segmented coded data set;

[0035] Step S333: recursively decompose the segmented coding data set into item sets to obtain item set candidate sets; extract pattern features from the item set candidate sets to obtain operation pattern feature item sets;

[0036] Step S334: performing repeated pattern filtering on the operation mode feature item set to obtain a preliminary frequent item set; performing statistical distribution analysis on the preliminary frequent item set to obtain abnormal distribution characteristics;

[0037] Step S335: performing abnormal distribution correction on the preliminary frequent itemsets based on the abnormal distribution characteristics to generate candidate frequent itemsets.

[0038] The present invention provides time series analysis of interactive data through time dimension division, enhances the ability to capture the changing trend of user behavior, and can more accurately reflect the temporal characteristics of user operations. Behavior type classification enables the characteristics of different interactive behaviors to be analyzed independently, which helps to accurately identify user needs, optimize human-computer interaction design, and enhance user experience. Sequence number conversion and segmentation processing simplify the data structure, facilitate subsequent algorithm processing, improve data processing efficiency, and reduce computational complexity. Recursive item set decomposition technology effectively extracts potential operation mode features, so that meaningful behavior patterns can be discovered from complex data, promoting intelligent analysis. Repeated pattern filtering enhances data cleaning capabilities, ensures the accuracy and reliability of analysis results, and reduces the interference of noise data on the results. The correction process of abnormal distribution features improves the adaptability of the model, ensures that the candidate frequent item sets reflect real user behaviors, and enhances the practical value of the model.

[0039] Preferably, step S4 comprises the following steps:

[0040] Step S41: using the user operation preference model to perform feature similarity calculation on the dynamic touch feature sequence to obtain similarity data; extracting similar operations from the dynamic touch feature sequence based on the similarity data to generate user similar operation behavior data;

[0041] Step S42: performing habit pattern recognition on user similar operation behavior data to obtain user operation habit data;

[0042] Step S43: performing invalid interaction behavior analysis on the dynamic touch feature sequence based on the user operation habit data to obtain redundant action recognition data;

[0043] Step S44: Eliminate redundant actions from the dynamic touch feature sequence based on the redundant action recognition data to obtain main interactive action data.

[0044] The present invention improves the ability to identify similar user operation behaviors and enhances the system's adaptability to user habits by performing in-depth analysis of dynamic touch feature sequences. Similar operation extraction ensures that the user's personalized needs are met, enabling the system to more accurately predict and respond to the user's intentions and improve the user's interactive experience. The habit pattern recognition process accurately captures the user's operating habits, supports more in-depth user behavior analysis, and provides a data basis for personalized interaction design. Invalid interaction behavior analysis significantly improves interaction efficiency, reduces unnecessary user operation steps, and optimizes the smoothness of human-computer interaction by identifying and eliminating redundant actions. The implementation of redundant action elimination reduces the waste of system resources, improves the response speed and accuracy of the device, and makes human-computer interaction more efficient and intelligent.

[0045] Preferably, step S43 includes the following steps:

[0046] Step S431: performing repeated event recognition on the user operation habit data to obtain historical user repeated behavior data; performing repeated event recognition on the dynamic touch feature sequence to obtain user repeated behavior data;

[0047] Step S432: extracting the same actions from the historical user repeated behavior data and the user repeated behavior data to generate a set of same operation behaviors;

[0048] Step S433: performing interactive simulation analysis on the same operation behavior set to generate redundant action recognition data.

[0049] The present invention improves the insight into user behavior through repeated event recognition, accurately captures the user's common operation patterns, helps the system understand user habits, and optimizes the interaction design. The extraction of identical actions ensures the effective use of repeated behaviors, enabling the system to intelligently recommend or automatically execute common user operations in appropriate contexts, thereby improving interaction efficiency. Interaction simulation analysis provides the system with more comprehensive user behavior data support, enabling more accurate prediction and optimization in the design phase, thereby improving user experience. Redundant action recognition significantly improves operational fluency by reducing unnecessary interactions, reduces the user's cognitive burden during operation, and enhances the naturalness of human-computer interaction. It promotes the intelligent development of equipment, makes human-computer interaction more personalized and adaptable, and helps improve user satisfaction and loyalty.

[0050] Preferably, step S433 includes the following steps:

[0051] Perform independent action classification on the same operation behavior set to obtain a classified independent behavior set; perform equal segment cutting on the classified independent behavior set to obtain an equal independent behavior set;

[0052] The equally divided independent behavior sets are sorted by timestamp to obtain behavior timestamp data; the equally divided independent behavior sets are labeled by time according to the behavior timestamp data to generate labeled equally divided independent behavior sets;

[0053] A screen operation simulation environment is constructed for the device screen operation data to obtain a simulated interactive environment;

[0054] The labeled equally divided independent behavior set is sequentially stacked to obtain a labeled equally divided independent behavior stack;

[0055] Based on the simulated interactive environment field, interactive operations are simulated for the same set of operation behaviors to generate ideal target interactive response data;

[0056] Repeat the following steps, extracting the top data of the labeled equally divided independent behavior stack each time the cycle is repeated, and obtaining the top data of the stack, until the labeled equally divided independent behavior stack is empty;

[0057] Based on the simulated interactive environment field, interactive operations are simulated on the labeled equally divided independent behavior stack to generate actual target interactive response data;

[0058] Compare the consistency of ideal target interaction response data and actual target interaction response data based on the behavior timestamp data.

[0059] When the ideal target interaction response data is consistent with the actual target interaction response data, the top data of the stack is deemed to be invalid interaction data;

[0060] When the ideal target interaction response data is inconsistent with the actual target interaction response data, the top data of the stack is considered to be the valid interaction data;

[0061] Integrate the identification features of invalid interaction data and valid interaction data to generate redundant action recognition data.

[0062] The present invention improves the fine-grained analysis capability of operational behaviors through independent action classification, so that the system can more accurately identify and understand the specific operations of users, and enhance the personalization of interactions. Equal segment cutting and timestamp sorting ensure the orderliness and consistency of behavioral data, facilitate subsequent analysis and processing, and improve the availability and reliability of data. The construction of a simulated interactive environment field provides a simulation of real scenes, making the evaluation of user operations more realistic and credible, and optimizing the interactive design. Sequential stack storage is more efficient in managing the labeled equal-divided independent behavior sets, which is convenient for gradually extracting and processing data during the simulation process, reducing memory usage and computational complexity. The response consistency comparison between the ideal target and the actual target ensures the accuracy of the interactive operation, can identify invalid interactions in real time, and thus optimize the user experience. The integration of identification features of effective and invalid interactive data provides a basis for improvement, which helps the system to continuously learn and self-optimize, and improve future interactive efficiency and intelligence.

[0063] Preferably, step S5 comprises the following steps:

[0064] Step S51: constructing a prediction model for the historical interaction data of the device to obtain an interaction prediction model;

[0065] Step S52: predicting the interaction intention of the main interaction action data according to the interaction prediction model to obtain predicted interaction intention data;

[0066] Step S53: mapping the predicted interaction intention data to control instructions to generate an interaction instruction sequence; performing electrical signal conversion processing on the interaction instruction sequence to obtain a device control signal;

[0067] Step S54: dynamically respond to and monitor the human-computer interaction device based on the device control signal to execute the control method of the human-computer interaction device.

[0068] The present invention improves the system's ability to predict user behavior and enhances the accuracy of the device's understanding and response to user intentions through the construction of an interactive prediction model, thereby improving the user experience. By predicting the intention of the main interactive action data, the system can adjust the interactive strategy in real time, provide more personalized services, and enhance the intelligent level of user interaction. The generation of interactive instruction sequences is combined with electrical signal conversion processing to ensure the efficient transmission of control signals, reduce response delays, and improve the operational flexibility of the device. The dynamic interactive response and monitoring mechanism ensures real-time feedback of user operations, allowing the device to quickly adapt to changes in user operations, improving the interactivity of the system and user satisfaction. By combining the analysis of historical data with real-time feedback, the continuous improvement of device performance and intelligent development are promoted.

[0069] The present invention also provides a human-computer interaction device, including a controller, for executing the control method of the human-computer interaction device as described above, wherein the human-computer interaction device controller includes:

[0070] The interactive monitoring module is used to obtain device screen operation data; divide the device screen operation data into time windows to obtain a discrete touch event sequence; calculate the touch acceleration of the discrete touch event sequence to generate a dynamic touch feature sequence;

[0071] The category grouping module is used to perform interactive feature clustering processing on the dynamic touch feature sequence to obtain an interactive feature clustering result set; group the interactive feature clustering result set into operation behavior categories to generate user operation behavior grouping data;

[0072] The historical mining module is used to obtain the historical interaction data of the device; based on the historical interaction data of the device, the user operation behavior grouping data is mined for frequent itemsets to obtain high-frequency operation patterns; the operation preference model is constructed for the high-frequency operation patterns to generate the user operation preference model;

[0073] The interactive data screening module is used to identify the user's operation habits for the dynamic touch feature sequence using the user operation preference model to generate user operation habit data; based on the user operation habit data, the dynamic touch feature sequence is subjected to habitual redundant action elimination to obtain the main interactive action data;

[0074] The interactive response module is used to predict the interactive intention of the main interactive action data to obtain the predicted interactive intention data; based on the predicted interactive intention data, the human-computer interactive device is dynamically interactively responded to to execute the control method of the human-computer interactive device.

[0075] The present invention obtains screen operation data in real time through the interactive monitoring module, and converts the operation data into a dynamic touch feature sequence through time window segmentation and touch acceleration calculation, thereby ensuring the timeliness and accuracy of data processing. The category grouping module generates user operation behavior grouping data through interactive feature clustering and operation behavior category grouping, so that the system can deeply understand the user's operation mode and habits. The historical mining module uses frequent item set mining technology to extract high-frequency operation modes from the device's historical interaction data, providing a scientific basis for the construction of the user operation preference model and enhancing the intelligence level of the system. The interactive data screening module identifies user habits through the user operation preference model and eliminates redundant actions to ensure the effectiveness of the main interactive actions, thereby improving the efficiency and accuracy of user operations. The interactive response module realizes dynamic response to user operations based on the predicted interactive intention data, so that the device can adapt to changes in user needs in a timely manner and improve the real-time and intelligent nature of the interactive experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 A schematic flow chart of a control method for a human-computer interaction device;

[0077] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0078] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0079] 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

[0080] The technical method of the present invention is described clearly and completely below 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.

[0081] 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.

[0082] 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.

[0083] To achieve this, please refer to Figures 1 to 3 , a control method for a human-computer interaction device, comprising the following steps:

[0084] Step S1: acquiring device screen operation data; performing time window segmentation on the device screen operation data to obtain a discrete touch event sequence; performing touch acceleration calculation on the discrete touch event sequence to generate a dynamic touch feature sequence;

[0085] Step S2: performing interactive feature clustering processing on the dynamic touch feature sequence to obtain an interactive feature clustering result set; grouping the interactive feature clustering result set into operation behavior categories to generate user operation behavior grouping data;

[0086] Step S3: Acquire historical device interaction data; perform frequent item set mining on user operation behavior grouping data based on the historical device interaction data to obtain high-frequency operation patterns; construct an operation preference model for the high-frequency operation patterns to generate a user operation preference model;

[0087] Step S4: using the user operation preference model to identify the user operation habits of the dynamic touch feature sequence to generate user operation habit data; based on the user operation habit data, removing redundant actions from the dynamic touch feature sequence to obtain main interactive action data;

[0088] Step S5: predict the interaction intention of the main interaction action data based on the historical interaction data of the device to obtain predicted interaction intention data; and dynamically respond to the human-computer interaction device based on the predicted interaction intention data to execute the control method of the human-computer interaction device.

[0089] The present invention reduces redundant operations of users when using devices by real-time identification and prediction of user operation intentions, thereby significantly improving the fluency and accuracy of interaction. This intuitive interaction method enables users to use devices more freely, reduces learning curves and operation difficulties, and thus improves overall user satisfaction. By establishing a user operation preference model and performing habit recognition, the system can respond to specific needs and habits of different users in a personalized manner. Each user's interaction history data provides a unique perspective for the system, enabling the device to accurately adjust its interaction mode to meet the personalized needs of users, and enhance the user's sense of belonging and stickiness. Redundant actions are eliminated in the dynamic touch feature sequence, and by identifying common redundant operations of users, false touches and unnecessary operation steps are reduced, which not only improves the efficiency of operation, but also reduces the frustration of users caused by incorrect operations, making the interaction process smoother. Frequent item set mining is performed using the device's historical interaction data to extract high-frequency operation modes, so that the system can optimize the control strategy based on real data analysis results. This data-driven method not only improves the accuracy of control, but also enables the device to make smarter decisions in complex environments. Based on the predicted interaction intention data, the device can perform dynamic interaction responses to achieve real-time feedback on user operations. This rapid adaptability ensures that the device can respond quickly when user intent changes, improves the flexibility and real-time nature of the interaction, and makes the device more intelligent to meet the ever-changing needs of users.

[0090] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a control method of a human-computer interaction device of the present invention. In this example, the control method of a human-computer interaction device includes the following steps:

[0091] Step S1: acquiring device screen operation data; performing time window segmentation on the device screen operation data to obtain a discrete touch event sequence; performing touch acceleration calculation on the discrete touch event sequence to generate a dynamic touch feature sequence;

[0092] In this embodiment, the process of obtaining device screen operation data uses a touch screen sensor to capture the user's touch actions in real time, including information such as touch position, timestamp, and touch force. By setting a time window, continuous touch event data is segmented to form a discrete touch event sequence. For example, the time window is set to 100 milliseconds, and all touch events within this time are integrated to form a touch event set. Next, the touch acceleration is calculated for the touch event sequence, and the touch speed is obtained by differentiating the touch position in time, and then the acceleration is calculated based on the time difference. The generated dynamic touch feature sequence will contain a timestamp, touch position, and calculated acceleration value, forming a complete data set for subsequent analysis.

[0093] Step S2: performing interactive feature clustering processing on the dynamic touch feature sequence to obtain an interactive feature clustering result set; grouping the interactive feature clustering result set into operation behavior categories to generate user operation behavior grouping data;

[0094] In this embodiment, when interactive feature clustering is performed on the dynamic touch feature sequence, the K-means clustering algorithm is used to classify the touch feature data set. By setting the number of clusters k, the position of each touch point in the feature space is first calculated, and then the clustering results are optimized by an iterative algorithm to obtain the optimal cluster center. After the clustering is completed, the clustering result set is grouped into operation behavior categories, and the same or similar operation behaviors are classified into the same category based on similarity to generate user operation behavior grouping data. For example, if a group of clustering results shows frequent double-click operations, the group of behaviors is marked as "quick operation" and the single-click operation is marked as "basic operation". In this way, a clear classification of different user operation habits can be formed.

[0095] Step S3: Acquire historical device interaction data; perform frequent item set mining on user operation behavior grouping data based on the historical device interaction data to obtain high-frequency operation patterns; construct an operation preference model for the high-frequency operation patterns to generate a user operation preference model;

[0096] In this embodiment, when obtaining the historical interaction data of the device, the user's past operation records, including timestamp, operation type, frequency and other information, are extracted through a data storage system (such as a database). These historical data are normalized for subsequent analysis. Next, the Apriori algorithm is used to perform frequent item set mining on the user operation behavior grouping data, and high-frequency operation patterns are identified by setting a minimum support threshold. For example, if a user frequently performs a combination of "slide" and "click" operations within a certain period of time, this combination is regarded as a high-frequency operation pattern. Subsequently, a user operation preference model is constructed based on the identified high-frequency patterns, and the user's operation tendencies in different situations are recorded so that a more intelligent response can be made in future interactions.

[0097] Step S4: using the user operation preference model to identify the user operation habits of the dynamic touch feature sequence to generate user operation habit data; based on the user operation habit data, removing redundant actions from the dynamic touch feature sequence to obtain main interactive action data;

[0098] In this embodiment, the user operation preference model is used to identify the user operation habits of the dynamic touch feature sequence. First, the user's operation habits are matched with the preference model to identify the user's main operation habits. The user operation habit data is generated by comparing the features contained in the dynamic touch feature sequence with the features in the preference model. For example, when the system detects that the user frequently uses the "slide" and "zoom" operations, these operations will be marked as main habits. Then, based on the user operation habit data, the redundant actions in the dynamic touch feature sequence are eliminated. For example, if some unnecessary double-click actions do not appear in the user's habits, they will be excluded, and finally the main interaction action data is obtained to simplify the subsequent processing process.

[0099] Step S5: predict the interaction intention of the main interaction action data based on the historical interaction data of the device to obtain predicted interaction intention data; and dynamically respond to the human-computer interaction device based on the predicted interaction intention data to execute the control method of the human-computer interaction device.

[0100] In this embodiment, when the interaction intention is predicted for the main interaction action data based on the historical interaction data of the device, the main interaction action data is first input into the prediction model, and the model is trained using a machine learning algorithm (such as a random forest or a support vector machine). During the training process, the model analyzes the patterns in the historical data to predict the user's next intention. The generated predicted interaction intention data will be used to formulate corresponding interaction strategies. Subsequently, the intention data is mapped to the interaction instructions to generate a series of specific interaction instruction sequences. For example, if the user's intention is predicted to be "open the application", the corresponding generated instruction is "start the application". Finally, the system converts these interaction instructions into electrical signals and transmits the signals to the device through the control interface to ensure that the human-computer interaction device can respond dynamically and complete the predetermined control operation.

[0101] Preferably, step S1 comprises the following steps:

[0102] Step S11: Acquire a device screen sensor; monitor screen contact actions according to the device screen sensor to obtain device screen operation data;

[0103] Step S12: performing noise filtering on the device screen operation data to obtain filtered screen operation data; performing time window segmentation on the filtered screen operation data to obtain a discrete touch event sequence;

[0104] Step S13: normalizing the coordinates of the discrete touch event sequence to obtain a standardized touch coordinate set; performing time difference processing on the standardized touch coordinate set to generate a touch velocity sequence;

[0105] Step S14: smoothing the touch speed sequence to obtain a smooth speed curve;

[0106] Step S15: extracting features from the smooth velocity curve to obtain a dynamic touch feature sequence.

[0107] In this embodiment, the acquisition of the device screen sensor is performed through the hardware interface of the access device to ensure that the multi-touch sensor of the screen can monitor the contact action in real time. The monitoring of touch events uses an interrupt method to sample the coordinates, pressure value and touch state (such as pressing, lifting) of the contact point at a high frequency. Whenever a touch event occurs, the sensor generates a corresponding data packet and transmits it to the data processing module to form preliminary device screen operation data. By setting the threshold of the touch event in the data processing module, the user's touch behavior can be effectively captured to ensure that all valid touch operation data, including continuous touch and sliding, are captured. Set a suitable cutoff frequency, input the screen operation data into the filter, and obtain the filtered data set after multiple sampling. Then, the filtered data is time-window segmented, the window length is set to 50 milliseconds, and the data is divided into multiple discrete touch event sequences. Each sequence contains all touch event information within the time period, thereby forming a structured touch event list for subsequent analysis and processing. When the coordinates of the discrete touch event sequence are normalized, the coordinate range of the screen is first determined, for example, set to a pixel value of (0,0) to (1920,1080). Then, the discrete touch events are traversed, and each touch coordinate is divided by the maximum value of the screen to convert the coordinate into a relative coordinate (ranging between 0 and 1). Next, the standardized touch coordinate set is time-differentiated, and a touch velocity sequence is generated by calculating the time difference of adjacent touch events. This velocity sequence provides the movement rate between touch points based on the standardized coordinates and timestamp of each touch event, providing the required data for subsequent analysis. The window size is set to 5, and the average of the current velocity and the two velocity values ​​before and after is taken to calculate a new smoothed velocity value. This process is repeated for the entire velocity sequence to generate a new smoothed velocity curve. In this way, accidental velocity fluctuations can be eliminated, and more stable and reliable velocity data can be obtained to facilitate the accuracy and effectiveness of subsequent feature extraction steps. Determine the type of features to be extracted, including touch acceleration, velocity change rate, and touch time features. Use a custom feature extraction function to process the smoothed velocity curve, calculate the derivative of the touch velocity to obtain the acceleration, and count the maximum and minimum values ​​of the velocity change. Through data reduction technology, the extracted features are stored as a dynamic touch feature sequence. This sequence will contain representative touch feature data, forming the basis for subsequent operations and supporting the device's analysis and response to user operation behaviors.

[0108] Preferably, step S2 comprises the following steps:

[0109] Step S21: performing feature vectorization processing on the dynamic touch feature sequence to obtain a feature vector set;

[0110] Step S22: Perform initial feature vector clustering on the feature vector set to generate initial clustering results:

[0111] Step S23: performing detailed calculation of the initial clustering result to obtain the optimal clustering number; performing feature vector clustering on the feature vector set based on the optimal clustering number to generate the optimal clustering result;

[0112] Step S24: performing label statistical analysis on the preferred clustering results to obtain main operation behavior category data; grouping the preferred clustering results into operation behavior categories according to the main operation category data to generate user operation behavior grouping data.

[0113] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0114] Step S21: performing feature vectorization processing on the dynamic touch feature sequence to obtain a feature vector set;

[0115] In this embodiment, when the dynamic touch feature sequence is processed by feature vectorization, the feature extraction criteria are first determined, including touch speed, acceleration, pressure value, etc. The feature sequence is processed using the NumPy library in the Python programming language, and the eigenvalues ​​of each touch event are integrated into a feature vector. In the specific operation, for each dynamic touch feature sequence, the relevant feature data is extracted and combined into a one-dimensional array to form a feature vector. This operation is repeated until all dynamic touch feature sequences are converted into feature vectors, thereby generating a feature vector set containing multiple feature vectors for subsequent clustering analysis.

[0116] Step S22: Perform initial feature vector clustering on the feature vector set to generate initial clustering results:

[0117] In this embodiment, the K-means clustering algorithm is used when performing initial clustering of feature vectors on a feature vector set. First, the initial value of the number of clusters is set, for example, set to five, and five feature vectors are randomly selected as the initial center points. Next, the distance from each feature vector to the five center points is calculated, and the feature vector is assigned to the cluster corresponding to the center point closest to it. This process is repeated until the position of the cluster center point no longer changes and reaches a convergence state. After clustering is completed, each cluster and the feature vectors it contains are saved to form an initial clustering result, preparing for subsequent cluster refinement analysis.

[0118] Step S23: performing detailed calculation of the initial clustering result to obtain the optimal clustering number; performing feature vector clustering on the feature vector set based on the optimal clustering number to generate the optimal clustering result;

[0119] In this embodiment, when the initial clustering result is calculated by the profile detail, first, for each feature vector, the average distance between it and other feature vectors in the same cluster is calculated, which is recorded as the same-class distance. At the same time, the average distance between the feature vector and the feature vector in the nearest other cluster is calculated, which is recorded as the different-class distance. Then, by comparing the same-class distance and the different-class distance, a numerical value describing the clustering effect of the feature vector is obtained, and the higher the numerical value, the better the clustering effect. After statistical analysis of the clustering effects of all feature vectors, the cluster number with the best clustering effect is selected as the preferred cluster number, and then the feature vector set is clustered again to generate the preferred clustering result.

[0120] Step S24: performing label statistical analysis on the preferred clustering results to obtain main operation behavior category data; grouping the preferred clustering results into operation behavior categories according to the main operation category data to generate user operation behavior grouping data.

[0121] In this embodiment, when the preferred clustering results are subjected to label statistical analysis, the feature vectors in each cluster are first labeled to form the operation behavior labels of each cluster. Then, using statistical analysis tools (such as Pandas library), the frequency of each label appearing in each cluster is calculated to generate the main operation behavior category data. Then, according to the frequency, the feature vectors in the preferred clustering results are grouped according to different operation behavior categories. Finally, user operation behavior grouping data is formed, and the data set contains feature vectors of different user behavior categories, which provides a basis for subsequent data mining and analysis, thereby laying the foundation for behavior identification and control methods.

[0122] Preferably, step S3 comprises the following steps:

[0123] Step S31: Acquire historical interaction data of the device; perform standard formatting on the historical interaction data of the device to obtain standardized historical data;

[0124] Step S32: multi-dimensionally integrate the normalized historical data and the user operation behavior grouping data to generate a comprehensive interaction data set;

[0125] Step S33: performing frequent item set mining on the comprehensive interaction data set to obtain candidate frequent item sets;

[0126] Step S34: Calculate the support of the candidate frequent item sets to generate a support data set; Calculate the confidence of the candidate frequent item sets to generate a confidence data set;

[0127] Step S35: screening the candidate frequent item sets for high-frequency patterns according to the support data set and the confidence data set to obtain high-frequency operation patterns;

[0128] Step S36: Perform association rule modeling on high-frequency operation modes to generate a user operation preference model.

[0129] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0130] Step S31: Acquire historical interaction data of the device; perform standard formatting on the historical interaction data of the device to obtain standardized historical data;

[0131] In this embodiment, in the process of obtaining the historical interaction data of the device, the storage system of the device is first called through the interface to extract all operation records generated by the user in the past interaction process. These operation records include the user's touch events, number of clicks, and device response time. The extracted data is stored in JSON (JavaScript Object Notation) format and needs to be parsed to extract key fields. Then, these data are standardized and formatted, for example, the time format is unified to the ISO 8601 standard, and all operation records are converted into a table form to ensure the consistency of field names, such as "touch time", "operation type", "operation coordinates", etc. After completing this process, the generated standardized historical data can be directly used in subsequent analysis steps.

[0132] Step S32: multi-dimensionally integrate the normalized historical data and the user operation behavior grouping data to generate a comprehensive interaction data set;

[0133] In this embodiment, when the normalized historical data and the user operation behavior grouping data are multi-dimensionally fused, the two data sets are first matched. By using common fields, such as user ID and timestamp, the historical interaction data is combined with the user operation behavior grouping data to form a comprehensive data set containing rich information. Next, the Pandas library is used for data processing, and the merge function is used to perform connection operations based on user ID and time. After data fusion, the fusion results need to be deduplicated to eliminate redundant records and ensure the neatness of the data set. The final generated comprehensive interaction data set contains multi-dimensional information of the user's historical interaction records and current operation behaviors, providing a basis for subsequent frequent item set mining.

[0134] Step S33: performing frequent item set mining on the comprehensive interaction data set to obtain candidate frequent item sets;

[0135] In this embodiment, when mining frequent item sets for a comprehensive interaction dataset, the Apriori algorithm is first selected as a mining tool. A minimum support threshold is set, such as 0.3, to ensure that the mined item sets have a certain frequency in the dataset. The comprehensive interaction dataset is converted into a transaction format and the apriori function is applied for calculation. In this process, the system traverses the dataset, counts the number of times each item set appears, and records the item sets that meet the support threshold. After completion, candidate frequent item sets are obtained, which contain operation combinations that appear frequently in user interactions, providing an important basis for subsequent data analysis.

[0136] Step S34: Calculate the support of the candidate frequent item sets to generate a support data set; Calculate the confidence of the candidate frequent item sets to generate a confidence data set;

[0137] In this embodiment, when calculating the support and confidence of candidate frequent item sets, first define the calculation method of support, that is, the frequency of occurrence of a certain item set in the data set. Use the previously obtained candidate frequent item sets, count the number of occurrences of each item set by traversing the data set, calculate the support of each item set, and generate a support data set. Next, calculate the confidence. For each candidate frequent item set, determine its subsets, and calculate the ratio of the frequency of occurrence of each subset to the frequency of occurrence of the entire item set through conditional probability to obtain a confidence data set. This step ensures that the relevance of each item set is quantified, which is convenient for the subsequent screening of high-frequency patterns.

[0138] Step S35: screening the candidate frequent item sets for high-frequency patterns according to the support data set and the confidence data set to obtain high-frequency operation patterns;

[0139] In this embodiment, when high-frequency patterns are screened for candidate frequent itemsets based on the support data set and the confidence data set, the support and confidence thresholds of the high-frequency patterns are first set, for example, the support is not less than 0.5 and the confidence is not less than 0.6. Then, by traversing the support and confidence data sets, the itemsets that meet these two thresholds are screened out. For each candidate itemset, the system checks whether its support and confidence meet the set standards. If the support and confidence of an itemset are both above the threshold, the itemset will be marked as a high-frequency operation mode. Finally, the obtained high-frequency operation mode will be used to build a user operation preference model.

[0140] Step S36: Perform association rule modeling on high-frequency operation modes to generate a user operation preference model.

[0141] In this embodiment, the classic association rule learning method is used when modeling association rules for high-frequency operation patterns. First, the screened high-frequency operation patterns are used as input, and the effectiveness of each rule is evaluated by calculating the support and confidence of the rules. Next, these patterns are modeled using the association_rules function in the mlxtend library in Python, and the generated rules will include the relationship between conditions and results. When analyzing the rules, pay attention to the level of support and confidence to ensure that the generated user operation preference model can accurately reflect the user's behavior pattern. Ultimately, the generated user operation preference model provides a basis for subsequent device control and optimization, forming an effective feedback mechanism to enhance user experience.

[0142] Preferably, step S33 includes the following steps:

[0143] Step S331: dividing the comprehensive interaction data set into time dimensions to obtain independent time series subsets; classifying the independent time series subsets into behavior types to obtain a classified interaction data set;

[0144] Step S332: performing serial number conversion on the classified interaction data set to obtain a numbered sequence data set; performing preliminary segmentation processing on the coded interaction data set to generate a segmented coded data set;

[0145] Step S333: recursively decompose the segmented coding data set into item sets to obtain item set candidate sets; extract pattern features from the item set candidate sets to obtain operation pattern feature item sets;

[0146] Step S334: performing repeated pattern filtering on the operation mode feature item set to obtain a preliminary frequent item set; performing statistical distribution analysis on the preliminary frequent item set to obtain abnormal distribution characteristics;

[0147] Step S335: performing abnormal distribution correction on the preliminary frequent itemsets based on the abnormal distribution characteristics to generate candidate frequent itemsets.

[0148] In this embodiment, when the comprehensive interaction data set is divided into time dimensions, the length of the time window is first determined, such as one window per hour. The time index function in Python's pandas library is used to divide the data set into multiple independent time series subsets by timestamp. Each subset contains all interaction records within the time period. Next, each independent time series subset is classified into behavior types, and the groupby method is used to group the same type of interaction behaviors according to the predefined labels of the behavior types (such as click, slide, long press, etc.), so as to obtain a classified interaction data set. Finally, the classified interaction data set contains different types of interaction behaviors divided by time periods, which provides a basis for subsequent analysis. When the classified interaction data set is converted into a sequence number, a unique number is first assigned to each type of interaction behavior. Create a dictionary structure to map each behavior type to a corresponding number. For example, click operations are mapped to 1, slide operations are mapped to 2, long press operations are mapped to 3, etc. Use the map function in Python to apply these number mappings to the classified interaction data set to generate a numbered sequence data set. Next, the numbered sequence data set is preliminarily segmented, with every ten interaction events as a segment. The data is segmented and stored in the segmented coding data set using the array slicing function of the numpy library. In this way, the generated segmented coding data set will contain multiple numbered segments, preparing for the subsequent item set decomposition. When recursively decomposing the segmented coding data set, a minimum support threshold is first set to ensure that the obtained item set has sufficient frequency in the data. Apply a recursive algorithm to traverse the segmented coding data set and extract each item set layer by layer. For each segment, the frequency of occurrence of each item set is calculated and compared with the support threshold. Item sets that meet the conditions will be collected into the item set candidate set. Next, the item set candidate set is subjected to pattern feature extraction to extract the key features of each item set, such as the frequency and relevance of the item set, to form the operation mode feature item set. This step ensures that representative operation modes are extracted, laying the foundation for subsequent analysis. When filtering the repeated patterns of the operation mode feature item set, the operation mode feature item set is first traversed to count the number of occurrences of each pattern in the data set. Set a frequency threshold, for example, a pattern must appear at least five times to be considered valid. Through the list derivation in Python, the patterns that do not meet the frequency threshold are removed from the operating pattern feature item set, thereby obtaining a preliminary frequent item set. Subsequently, a statistical distribution analysis is performed on the preliminary frequent item set, the distribution of each pattern in the entire data set is calculated, and the distribution feature data of each pattern is generated. This process involves histogram calculation, which can identify the normal and abnormal distribution characteristics of the pattern in the data set. When performing abnormal distribution correction on the preliminary frequent item set based on the abnormal distribution characteristics, first analyze the statistical distribution analysis results to identify the frequent item sets that appear abnormally in the data set.For example, the frequency of some patterns is significantly higher than that of other patterns, indicating abnormal behavior. Apply the Z-score (standard deviation) method to calculate the Z value of each frequent item set and identify the item sets whose Z values ​​exceed a certain range. These abnormal item sets are corrected and their frequency of occurrence is adjusted, such as normalizing their frequency, to ensure that the distribution of frequent item sets is more balanced. Ultimately, the generated candidate frequent item sets will provide more accurate basic data for the subsequent construction of user operation preference models.

[0149] Preferably, step S4 comprises the following steps:

[0150] Step S41: using the user operation preference model to perform feature similarity calculation on the dynamic touch feature sequence to obtain similarity data; extracting similar operations from the dynamic touch feature sequence based on the similarity data to generate user similar operation behavior data;

[0151] Step S42: performing habit pattern recognition on user similar operation behavior data to obtain user operation habit data;

[0152] Step S43: performing invalid interaction behavior analysis on the dynamic touch feature sequence based on the user operation habit data to obtain redundant action recognition data;

[0153] Step S44: Eliminate redundant actions from the dynamic touch feature sequence based on the redundant action recognition data to obtain main interactive action data.

[0154] In this embodiment, key indicators of each touch feature are extracted, such as touch position, duration and touch strength. These indicators are input as feature vectors, and the similarity between each feature vector is calculated using the cosine similarity algorithm. The cosine_similarity function is applied through the scikit-learn library in Python to generate a similarity data matrix. Next, similar operations are extracted based on the similarity data, a similarity threshold is set, and dynamic touch feature sequences with similarity higher than the threshold are screened out to generate user similar operation behavior data. Ultimately, the data set will contain dynamic touch behavior sequences similar to the user's past operation habits. A clustering algorithm, such as K-means clustering, is used to analyze the patterns in the user's similar operation behavior data. An appropriate number of clusters is set, and three to five categories are initially selected based on the diversity of user behavior. The KMeans model in scikit-learn is used to train the data, and labels are assigned to each cluster to indicate different operation habit patterns. The center of each cluster represents the corresponding habit pattern. By analyzing the clustering results, the user's main operation habits are identified to form a user operation habit data set, which provides a basis for subsequent invalid interaction behavior analysis. When analyzing invalid interaction behaviors of dynamic touch feature sequences based on user operation habit data, the standard for invalid interactions is first set, such as touch time that is too short or touch patterns that are obviously inconsistent with user habits. Using statistical methods, the mean and standard deviation of each touch feature are calculated to identify touch events with a touch time less than the mean minus one standard deviation. These events will be marked as potential invalid interaction behaviors. Next, redundant action recognition data is generated to record the specific information of these invalid touch events, such as touch position and duration, to provide a basis for subsequent elimination steps. When eliminating redundant actions of dynamic touch feature sequences based on redundant action recognition data, the identified redundant actions are first compared with the original dynamic touch feature sequence. Through Python's pandas library, the data filtering function is used to eliminate records in the touch feature sequence that match the redundant actions to form a new data set. This data set will only contain the main interaction actions to ensure that valid user operations are retained. Next, the main interaction action data is generated and stored for subsequent analysis and execution of the control method. This process ensures that the final interaction data is more refined and accurate, and improves the response effect of human-computer interaction.

[0155] Preferably, step S43 includes the following steps:

[0156] Step S431: performing repeated event recognition on the user operation habit data to obtain historical user repeated behavior data; performing repeated event recognition on the dynamic touch feature sequence to obtain user repeated behavior data;

[0157] Step S432: extracting the same actions from the historical user repeated behavior data and the user repeated behavior data to generate a set of same operation behaviors;

[0158] Step S433: performing interactive simulation analysis on the same operation behavior set to generate redundant action recognition data.

[0159] In this embodiment, when performing repeated event recognition on user operation habit data, the user's touch behavior records are first collected, including timestamps, touch positions, and operation types. Adjacent touch events are grouped using time window technology, and a fixed time interval, such as 500 milliseconds, is set to determine whether there are repeated touch events within the time window. If the events overlap within the time window, the event is recorded as a repeated behavior. By analyzing the data set, historical user repeated behavior data is generated. Next, similar repeated event recognition is performed on the dynamic touch feature sequence, and the same time window technology is used to filter out the user's touch behavior under the same conditions to form user repeated behavior data. This process ensures that the user's behavior habits and repeated patterns can be effectively captured. When performing the same action extraction on historical user repeated behavior data and user repeated behavior data, both must first be formatted to ensure that the data type and structure are consistent. By comparing the touch position and operation type of each repeated behavior, the same operation that appears repeatedly in a similar context is identified. For example, a similarity threshold can be set, and if the coordinates of the two touches are within a certain range and the operation type is the same, they are marked as the same action. Use the merge function in the pandas library to combine the conditions of the same action to merge the data set and finally generate the same operation behavior set. This set will contain all the same operations that the user has repeatedly performed in historical and current interactions, providing basic data for subsequent interaction simulation analysis. When performing interaction simulation analysis on the same operation behavior set, you first need to build a simulation environment to facilitate the reproduction of the user's interaction behavior. In this environment, computer simulation technology is applied to generate a virtual device interface to receive each operation sequence in the user's same operation behavior set. By writing a script, the same touch operation is executed in a loop, and various types of feedback generated by the device during the response process are recorded. This process can be implemented using tools such as Unity to simulate the device's response and effect under different operations. Analyze these response data, identify redundant behaviors generated under the same operation, and form redundant action recognition data. This data set will specifically identify which operations failed to respond effectively in the simulation, indicate potential redundant actions, and provide a basis for subsequent interaction optimization.

[0160] Preferably, step S433 includes the following steps:

[0161] Perform independent action classification on the same operation behavior set to obtain a classified independent behavior set; perform equal segment cutting on the classified independent behavior set to obtain an equal independent behavior set;

[0162] The equally divided independent behavior sets are sorted by timestamp to obtain behavior timestamp data; the equally divided independent behavior sets are labeled by time according to the behavior timestamp data to generate labeled equally divided independent behavior sets;

[0163] A screen operation simulation environment is constructed for the device screen operation data to obtain a simulated interactive environment;

[0164] The labeled equally divided independent behavior set is sequentially stacked to obtain a labeled equally divided independent behavior stack;

[0165] Based on the simulated interactive environment field, interactive operations are simulated for the same set of operation behaviors to generate ideal target interactive response data;

[0166] Repeat the following steps, extracting the top data of the labeled equally divided independent behavior stack each time the cycle is repeated, and obtaining the top data of the stack, until the labeled equally divided independent behavior stack is empty;

[0167] Based on the simulated interactive environment field, interactive operations are simulated on the labeled equally divided independent behavior stack to generate actual target interactive response data;

[0168] Compare the consistency of ideal target interaction response data and actual target interaction response data based on the behavior timestamp data.

[0169] When the ideal target interaction response data is consistent with the actual target interaction response data, the top data of the stack is deemed to be invalid interaction data;

[0170] When the ideal target interaction response data is inconsistent with the actual target interaction response data, the top data of the stack is considered to be the valid interaction data;

[0171] Integrate the identification features of invalid interaction data and valid interaction data to generate redundant action recognition data.

[0172] In this embodiment, feature extraction is performed on each behavior in the same operation behavior set to identify the unique features of each behavior, including touch position, operation type, time length, etc. A decision tree-based classification algorithm is used to cluster the behavior data by features, such as using the DecisionTreeClassifier in the scikit-learn library for classification. By setting feature thresholds, the behaviors are divided into different categories to form a classified independent behavior set. This set will contain independent actions of users in different scenarios, providing basic data for subsequent processing. Determine the time length of the cut, such as dividing each behavior into equal segments with a duration of 300 milliseconds. A custom function is used to traverse the classified independent behavior set, and each behavior is segmented according to the set time length, and each segment forms an independent behavior segment. The array slicing function of the numpy library can be used to cut the behavior data. The timestamp information of each behavior segment is extracted and stored in an array. The sort_values ​​method in the pandas library is used to sort the array to ensure that the behaviors are arranged in chronological order. After sorting, the behavior timestamp data is generated, which records the accurate time information of each behavior segment. Next, the equally divided independent behavior set is labeled according to the behavior timestamp data, and a label is assigned to each equally divided segment using the timestamp information to form a labeled equally divided independent behavior set. Select a suitable development tool, such as Unity or Unreal Engine, to build an interactive interface for a simulated device. The interface must contain various interactive elements, such as buttons and sliders, and ensure that their behaviors are the same as those of real devices. Through scripting logic, the user's touch events can be received in the simulated environment, and corresponding operations can be performed, and finally a simulated interactive environment field that can simulate real interactions is built. Initialize an empty stack structure, and use the push operation in the data structure to store the labeled equally divided independent behaviors one by one in the stack. The collections.deque library can be used to implement stack operations to ensure that data is discharged in the order in which they are stored. In this way, a labeled equally divided independent behavior stack is generated to provide data support for subsequent interactive operation simulations. Write a control script to extract and execute the behavior data in the stack one by one. During the loop operation, use the pop function to extract the top data from the labeled equally divided independent behavior stack, and perform corresponding operations in the simulation environment to record the interactive response data of the system. These response data record the device's reaction to each operation in the simulation environment, forming the ideal target interactive response data. When repeating the following steps, extract the top data of the stack of labeled equally divided independent behaviors each loop, use the pop operation to take it out and store it in a temporary variable until the stack is empty. Each time it is extracted, compare the timestamp information of the top data with the timestamp in the interactive environment to ensure that each simulation operation is consistent with the actual time. The loop process ensures that the data in the stack is extracted until it is completely used by writing conditional judgments.When the interactive operation simulation of the labeled equally divided independent behavior stack is performed based on the simulated interactive environment field, the same control script as the above interactive operation simulation is used to operate and execute the behavior data in the stack. During execution, each labeled equally divided independent behavior extracted from the stack is passed to the simulation environment one by one, and the actual target interactive response data generated is recorded. This process ensures that the real response of the system can be obtained under the same operation. When the consistency comparison of the ideal target interactive response data and the actual target interactive response data is performed based on the behavior timestamp data, the timestamps and response contents of the two response data sets are first extracted. Using a custom algorithm, the response contents corresponding to the timestamps are compared one by one, and the consistency and inconsistency are recorded. If the contents of the two response data are completely consistent at the same timestamp, the data at the top of the stack is considered to be invalid interactive data; if they are inconsistent, the data is considered to be valid interactive data. Through this process, the invalid and valid interactive data of the user can be effectively identified, providing a basis for subsequent analysis. When integrating the identification features of invalid interactive data and valid interactive data, the two types of data must first be formatted to ensure that the data structure is consistent. For valid interaction data, its features are extracted, such as response time, touch position, and operation type, etc. For invalid interaction data, its common features need to be analyzed. Through data integration methods, the two types of data are combined to generate redundant action recognition data to form a complete data set to support further analysis.

[0173] Preferably, step S5 comprises the following steps:

[0174] Step S51: constructing a prediction model for the historical interaction data of the device to obtain an interaction prediction model;

[0175] Step S52: predicting the interaction intention of the main interaction action data according to the interaction prediction model to obtain predicted interaction intention data;

[0176] Step S53: mapping the predicted interaction intention data to control instructions to generate an interaction instruction sequence; performing electrical signal conversion processing on the interaction instruction sequence to obtain a device control signal;

[0177] Step S54: dynamically respond to and monitor the human-computer interaction device based on the device control signal to execute the control method of the human-computer interaction device.

[0178] In this embodiment, the generated user interaction data is collected, and the data types include the user's operation behavior, timestamp, device response, etc. These data are stored in the database, and the pandas library is used for data cleaning and preprocessing to ensure the uniform data format. Then, a suitable machine learning algorithm is selected, such as Random Forest or Long Short-Term Memory Network (LSTM), and the model training is performed using scikit-learn or Keras library. The preprocessed data set is divided into a training set and a test set, and the training model is used to enable it to identify the user's interaction mode. Finally, the obtained interaction prediction model can predict future user behavior based on historical data to form an effective interaction prediction model. The main interaction action data is input into the constructed prediction model, and the prediction result is obtained using the model.predict method. The data includes the user's touch behavior and its time series information. The model will analyze the input data, predict the user's interaction intention, and the output result is the predicted interaction intention data. At this time, it is necessary to ensure the accuracy of the prediction, and the performance of the model is monitored and optimized through model evaluation indicators such as accuracy and recall. A control instruction library is established, which contains specific control instructions corresponding to different interaction intentions. Using a dictionary data structure, each predicted intention is mapped to the corresponding control instruction. Then, according to the generated interactive instruction sequence, the electrical signal conversion process is performed to convert the control instruction into an electrical signal that the device can recognize. Using a microcontroller (such as Arduino or Raspberry Pi), the signal output is realized through its GPIO (general input and output) interface to ensure that the signal can be correctly sent to the human-computer interaction device to obtain the device control signal. The control signal is transmitted to the device through a communication interface (such as USB, Bluetooth or Wi-Fi). In the device, a response mechanism is built to ensure that the device can perform corresponding operations according to the received control signal, such as starting, stopping or adjusting the state. Through embedded programming, the state changes of the device are monitored and the response results of the device are fed back in real time. To ensure the effectiveness of the control method, a feedback loop is used to monitor the execution status of the device, and the interaction strategy is adjusted according to the real-time feedback data of the device, thereby realizing an efficient human-computer interaction control method.

[0179] The present invention also provides a human-computer interaction device, including a controller, for executing the control method of the human-computer interaction device as described above, wherein the human-computer interaction device controller includes:

[0180] The interactive monitoring module is used to obtain device screen operation data; divide the device screen operation data into time windows to obtain a discrete touch event sequence; calculate the touch acceleration of the discrete touch event sequence to generate a dynamic touch feature sequence;

[0181] The category grouping module is used to perform interactive feature clustering processing on the dynamic touch feature sequence to obtain an interactive feature clustering result set; group the interactive feature clustering result set into operation behavior categories to generate user operation behavior grouping data;

[0182] The historical mining module is used to obtain the historical interaction data of the device; based on the historical interaction data of the device, the user operation behavior grouping data is mined for frequent itemsets to obtain high-frequency operation patterns; the operation preference model is constructed for the high-frequency operation patterns to generate the user operation preference model;

[0183] The interactive data screening module is used to identify the user's operation habits for the dynamic touch feature sequence using the user operation preference model to generate user operation habit data; based on the user operation habit data, the dynamic touch feature sequence is subjected to habitual redundant action elimination to obtain the main interactive action data;

[0184] The interactive response module is used to predict the interactive intention of the main interactive action data to obtain the predicted interactive intention data; based on the predicted interactive intention data, the human-computer interactive device is dynamically interactively responded to to execute the control method of the human-computer interactive device.

[0185] The present invention obtains screen operation data in real time through the interactive monitoring module, and converts the operation data into a dynamic touch feature sequence through time window segmentation and touch acceleration calculation, thereby ensuring the timeliness and accuracy of data processing. The category grouping module generates user operation behavior grouping data through interactive feature clustering and operation behavior category grouping, so that the system can deeply understand the user's operation mode and habits. The historical mining module uses frequent item set mining technology to extract high-frequency operation modes from the device's historical interaction data, providing a scientific basis for the construction of the user operation preference model and enhancing the intelligence level of the system. The interactive data screening module identifies user habits through the user operation preference model and eliminates redundant actions to ensure the effectiveness of the main interactive actions, thereby improving the efficiency and accuracy of user operations. The interactive response module realizes dynamic response to user operations based on the predicted interactive intention data, so that the device can adapt to changes in user needs in a timely manner and improve the real-time and intelligent nature of the interactive experience.

[0186] 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.

[0187] 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 control method for a human-computer interaction device, characterized in that: The following steps are involved: Step S1: Obtain device screen operation data; Perform time window segmentation on the device screen operation data to obtain a discrete touch event sequence; Calculate touch acceleration for discrete touch event sequences to generate dynamic touch feature sequences; Step S2: performing interactive feature clustering processing on the dynamic touch feature sequence to obtain an interactive feature clustering result set; The interactive feature clustering result set is grouped into operation behavior categories to generate user operation behavior grouping data; Step S3: Obtain device historical interaction data; Based on the historical interaction data of the device, the user operation behavior group data is mined for frequent itemsets to obtain high-frequency operation patterns; the operation preference model is constructed for the high-frequency operation patterns to generate the user operation preference model; Step S4: using the user operation preference model to identify the user operation habits of the dynamic touch feature sequence to generate user operation habit data; based on the user operation habit data, removing redundant actions from the dynamic touch feature sequence to obtain main interactive action data; Step S5: predict the interaction intention of the main interaction action data based on the historical interaction data of the device to obtain predicted interaction intention data; and dynamically respond to the human-computer interaction device based on the predicted interaction intention data to execute the control method of the human-computer interaction device.

2. The control method of the human-computer interaction device according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire a device screen sensor; monitor screen contact actions according to the device screen sensor to obtain device screen operation data; Step S12: performing noise filtering on the device screen operation data to obtain filtered screen operation data; performing time window segmentation on the filtered screen operation data to obtain a discrete touch event sequence; Step S13: normalizing the coordinates of the discrete touch event sequence to obtain a standardized touch coordinate set; performing time difference processing on the standardized touch coordinate set to generate a touch velocity sequence; Step S14: smoothing the touch speed sequence to obtain a smooth speed curve; Step S15: extracting features from the smooth velocity curve to obtain a dynamic touch feature sequence.

3. The control method of the human-computer interaction device according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing feature vectorization processing on the dynamic touch feature sequence to obtain a feature vector set; Step S22: Perform initial feature vector clustering on the feature vector set to generate initial clustering results: Step S23: performing detailed calculation of the initial clustering result to obtain the optimal clustering number; performing feature vector clustering on the feature vector set based on the optimal clustering number to generate the optimal clustering result; Step S24: performing label statistical analysis on the preferred clustering results to obtain main operation behavior category data; grouping the preferred clustering results into operation behavior categories according to the main operation category data to generate user operation behavior grouping data.

4. The control method of the human-computer interaction device according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: Acquire historical interaction data of the device; perform standard formatting on the historical interaction data of the device to obtain standardized historical data; Step S32: multi-dimensionally integrate the normalized historical data and the user operation behavior grouping data to generate a comprehensive interaction data set; Step S33: performing frequent item set mining on the comprehensive interaction data set to obtain candidate frequent item sets; Step S34: Calculate the support of the candidate frequent item sets to generate a support data set; Calculate the confidence of the candidate frequent item sets to generate a confidence data set; Step S35: screening the candidate frequent item sets for high-frequency patterns according to the support data set and the confidence data set to obtain high-frequency operation patterns; Step S36: Perform association rule modeling on high-frequency operation modes to generate a user operation preference model.

5. The control method of the human-computer interaction device according to claim 4, characterized in that: Step S33 includes the following steps: Step S331: dividing the comprehensive interaction data set into time dimensions to obtain independent time series subsets; classifying the independent time series subsets into behavior types to obtain a classified interaction data set; Step S332: performing serial number conversion on the classified interaction data set to obtain a numbered sequence data set; performing preliminary segmentation processing on the coded interaction data set to generate a segmented coded data set; Step S333: recursively decompose the segmented coding data set into item sets to obtain item set candidate sets; extract pattern features from the item set candidate sets to obtain operation pattern feature item sets; Step S334: performing repeated pattern filtering on the operation mode feature item set to obtain a preliminary frequent item set; performing statistical distribution analysis on the preliminary frequent item set to obtain abnormal distribution characteristics; Step S335: performing abnormal distribution correction on the preliminary frequent itemsets based on the abnormal distribution characteristics to generate candidate frequent itemsets.

6. The control method of the human-computer interaction device according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: using the user operation preference model to perform feature similarity calculation on the dynamic touch feature sequence to obtain similarity data; extracting similar operations from the dynamic touch feature sequence based on the similarity data to generate user similar operation behavior data; Step S42: performing habit pattern recognition on user similar operation behavior data to obtain user operation habit data; Step S43: performing invalid interaction behavior analysis on the dynamic touch feature sequence based on the user operation habit data to obtain redundant action recognition data; Step S44: Eliminate redundant actions from the dynamic touch feature sequence based on the redundant action recognition data to obtain main interactive action data.

7. The control method of the human-computer interaction device according to claim 6, characterized in that: Step S43 includes the following steps: Step S431: performing repeated event recognition on the user operation habit data to obtain historical user repeated behavior data; performing repeated event recognition on the dynamic touch feature sequence to obtain user repeated behavior data; Step S432: extracting the same actions from the historical user repeated behavior data and the user repeated behavior data to generate a set of same operation behaviors; Step S433: performing interactive simulation analysis on the same operation behavior set to generate redundant action recognition data.

8. The control method of the human-computer interaction device according to claim 7, characterized in that: Step S433 includes the following steps: Perform independent action classification on the same operation behavior set to obtain a classified independent behavior set; perform equal segment cutting on the classified independent behavior set to obtain an equal independent behavior set; The equally divided independent behavior sets are sorted by timestamp to obtain behavior timestamp data; the equally divided independent behavior sets are labeled by time according to the behavior timestamp data to generate labeled equally divided independent behavior sets; A screen operation simulation environment is constructed for the device screen operation data to obtain a simulated interactive environment; The labeled equally divided independent behavior set is sequentially stacked to obtain a labeled equally divided independent behavior stack; Based on the simulated interactive environment field, interactive operations are simulated for the same set of operation behaviors to generate ideal target interactive response data; Repeat the following steps, extracting the top data of the labeled equally divided independent behavior stack each time the cycle is repeated, and obtaining the top data of the stack, until the labeled equally divided independent behavior stack is empty; Based on the simulated interactive environment field, interactive operations are simulated on the labeled equally divided independent behavior stack to generate actual target interactive response data; Compare the consistency of ideal target interaction response data and actual target interaction response data based on the behavior timestamp data. When the ideal target interaction response data is consistent with the actual target interaction response data, the top data of the stack is deemed to be invalid interaction data; When the ideal target interaction response data is inconsistent with the actual target interaction response data, the top data of the stack is considered to be the valid interaction data; Integrate the identification features of invalid interaction data and valid interaction data to generate redundant action recognition data.

9. The control method of a human-computer interaction device according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: constructing a prediction model for the historical interaction data of the device to obtain an interaction prediction model; Step S52: predicting the interaction intention of the main interaction action data according to the interaction prediction model to obtain predicted interaction intention data; Step S53: mapping the predicted interaction intention data to control instructions to generate an interaction instruction sequence; performing electrical signal conversion processing on the interaction instruction sequence to obtain a device control signal; Step S54: dynamically respond to and monitor the human-computer interaction device based on the device control signal to execute the control method of the human-computer interaction device.

10. A human-computer interaction device, characterized in that: A controller is included, which is used to execute the control method of the human-computer interaction device according to claim 1, and the human-computer interaction device controller includes: The interactive monitoring module is used to obtain device screen operation data; divide the device screen operation data into time windows to obtain a discrete touch event sequence; calculate the touch acceleration of the discrete touch event sequence to generate a dynamic touch feature sequence; The category grouping module is used to perform interactive feature clustering processing on the dynamic touch feature sequence to obtain an interactive feature clustering result set; group the interactive feature clustering result set into operation behavior categories to generate user operation behavior grouping data; The historical mining module is used to obtain the historical interaction data of the device; based on the historical interaction data of the device, the user operation behavior grouping data is mined for frequent itemsets to obtain high-frequency operation patterns; the operation preference model is constructed for the high-frequency operation patterns to generate the user operation preference model; The interactive data screening module is used to identify the user's operation habits for the dynamic touch feature sequence using the user operation preference model to generate user operation habit data; based on the user operation habit data, the dynamic touch feature sequence is subjected to habitual redundant action elimination to obtain the main interactive action data; The interactive response module is used to predict the interactive intention of the main interactive action data to obtain the predicted interactive intention data; based on the predicted interactive intention data, the human-computer interactive device is dynamically interactively responded to to execute the control method of the human-computer interactive device.

Images