A method and system for user input information in an all-in-one smart television

Abstract

The present invention relates to the field of television image communication technology, and in particular to a method and system for user input information in an all-in-one smart television. The method comprises the following steps: collecting user images, performing white balance correction and focusing processing, generating a focused image of a moving object, identifying characters in the focused image and analyzing attention tendency, obtaining character tendency data, analyzing historical interaction images, identifying interaction patterns and constructing a user intention prediction model, and when the tendency data indicates that the user prefers television content, performing intention drills and auxiliary adjustments on the current image based on the prediction model, generating predicted user input information, detecting logical conflicts in the predicted information and performing feedback optimization to output the final user input information. The present invention realizes a more accurate and flexible method for user input information.

Description

Technical Field

[0001] The present invention relates to the technical field of television image communication, and in particular to a method and system for user inputting information in an all-in-one smart television. Background Art

[0002] Users' demands for interactive methods are becoming increasingly diverse, from traditional remote control input to voice control, and then to gesture and face recognition. Users' input methods are gradually enriched. However, when processing these diverse inputs, all-in-one smart TVs face challenges such as complex data sources and inaccurate inputs. Existing user input recognition methods for all-in-one smart TVs often focus on a specific type of input mode, making it difficult to fully capture user behavior. For example, camera-based gesture recognition or facial recognition cannot fully capture users' tendencies or complex interactive intentions, and has low accuracy in low-light environments or dynamic backgrounds. The prediction of input intentions, as a key link in user input information processing, also lacks mature model support, which often makes it difficult for smart TVs to accurately predict the user's next operation and provide truly intelligent services. At the same time, existing input intention models generally lack effective use of historical data, resulting in limited prediction accuracy and flexibility, and the user experience does not achieve the expected intelligent effect. Summary of the invention

[0003] Based on this, it is necessary to provide a method and system for user input information in a smart TV all-in-one to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for user inputting information in a smart TV all-in-one machine comprises the following steps:

[0005] Step S1: capturing a current user image; performing white balance correction on the current user image to obtain a color-corrected image; performing dynamic image focusing on the color-corrected image to generate a moving object focused image;

[0006] Step S2: performing person recognition on the focused image of the moving object to obtain a person image; performing attention tendency analysis on the person image to generate person tendency data;

[0007] Step S3: Acquire historical figure interaction images; perform interaction pattern recognition on the historical figure interaction images to obtain historical interaction pattern data; perform user input intention prediction modeling on the historical interaction pattern data to generate a user input intention prediction model;

[0008] Step S4: when the character tendency data is determined to be character tendency television data, a user input intention drill is performed on the moving object focused image based on the user intention prediction model to generate a predicted user input intention; auxiliary adjustments are made to the predicted user input intention according to the character tendency data and the moving object focused image to generate predicted user input information;

[0009] Step S5: performing logical conflict detection on the predicted user input information to obtain conflict detection data; performing feedback optimization processing on the predicted user input information based on the conflict detection data to generate user input information.

[0010] The present invention acquires a corrected image of true color by collecting the current user image and performing white balance correction, which helps to improve the color accuracy of the image. By focusing the dynamic image, a clear image of the moving object is captured, thereby ensuring the accuracy of subsequent character recognition and attention analysis. By white balance correction and focusing processing, high-quality adjustment of the input image is achieved, so that the smart TV can obtain clear and accurate user image source data, thereby improving the reliability of the recognition effect. Character recognition is performed on the focused image of the moving object to obtain a clear user character image. Then, through attention tendency analysis, tendency data reflecting the user's gaze or attention area is generated based on the analysis algorithm, effectively realizing the capture of the user's attention state, laying a data foundation for predicting the user's interaction intention. The tendency data directly reflects the user's current interest area, thereby providing a basis for subsequent prediction. Through this recognition and analysis process, the smart TV can quickly and accurately focus on the user's interest area, acquire and process historical character interaction images, acquire historical interaction mode data through interaction mode recognition, and further construct a user input intention prediction model. The model is based on the data of historical interaction behavior to form an algorithm framework for predicting the user's operation intention. The input intention prediction model can effectively associate the user's habits in specific scenarios. Operation mode, based on the learning of historical patterns, enables the prediction model to more accurately reflect the user's intention, provides a data basis for the reliability and customization of further operations, and enhances the applicability of the prediction effect. When the character tendency data indicates that the user tends to TV data, the user intention prediction model is used to perform intention drills on the focused image of the moving object to predict the user's input intention, and then the predicted intention is auxiliary adjusted according to the tendency data and the focused image to generate accurate user input information. Dynamic update prediction is achieved through data adjustment to ensure that the prediction results are more in line with the user's immediate needs. The auxiliary adjustment function realizes the precise matching of intention data and real-time image information, enabling smart TVs to generate operation information that is more in line with user needs, ensuring the accuracy and real-time response of prediction information, performing logical conflict detection on the generated predicted user input information, identifying and handling potential operation conflict risks, and adjusting the user input information through feedback optimization to ensure that the generated user input information can smoothly and smoothly guide the TV's response. Through conflict detection and optimization processing, smart TVs can effectively avoid errors caused by repetitive and contradictory operations. Through the adjustment mechanism of logical optimization, the rationality of user input information is ensured, enhancing the fluency of smart TV interactive operations.

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

[0012] Step S11: capturing the current user image; performing color temperature analysis on the current user image to obtain color temperature information of the user image; performing white balance correction calculation on the color temperature information of the user image to generate white balance correction parameters;

[0013] Step S12: performing color correction on the current user image based on the white balance correction parameters to obtain a color-corrected image;

[0014] Step S13: performing motion region detection on the color-corrected image to obtain a motion region image; performing motion trajectory tracking on the motion region image to obtain a motion region trajectory;

[0015] Step S14: dynamically focusing the moving area image according to the moving area trajectory to generate a moving object focused image.

[0016] The present invention acquires the color temperature information of the user image by collecting the current user image and performing color temperature analysis, thereby ensuring that the color characteristics of the image meet the actual lighting conditions of the user's environment. The white balance correction calculation generates white balance correction parameters based on the color temperature information, so that the system can automatically adjust the white balance state of the image according to the color temperature of the light source. The white balance correction parameters ensure the accurate restoration of the colors in the image, reduce the color distortion caused by lighting changes, and provide a basic image with true colors for subsequent image processing steps. The current user image is color corrected based on the white balance correction parameters to obtain a color-corrected image. The color correction ensures the naturalness of the image color, so that the user can obtain an image effect with consistent visual colors under various ambient light sources, thereby improving the image perception. The color-corrected image data is more conducive to subsequent motion detection and trajectory analysis, thereby enhancing the image processing stability of the system and the accuracy of user input recognition. The motion area detection is performed on the color-corrected image to generate a motion image. Moving area image, motion area detection can effectively separate the areas with dynamic changes in the image, so that the system can focus on the motion areas related to user interaction and reduce unnecessary background interference. Motion trajectory tracking further obtains the path information of the motion area to ensure that the system can clearly identify the user's motion trajectory and provide accurate motion data support for subsequent dynamic focusing processing. The motion trajectory data directly reflects the user's movement direction and changes, and improves the system's real-time response capability to user behavior. The moving area image is dynamically focused according to the motion area trajectory to generate a moving object focused image. The dynamic focusing process automatically adjusts the focus position of the image by combining the motion trajectory information, so that the system can lock and track the user's key actions. The focused image effectively eliminates background noise, making the user's actions clearer and more significant in the picture. The moving object focused image provides high-quality visual input for the subsequent interaction process, enhancing the system's recognition accuracy and response accuracy to user input commands.

[0017] Preferably, step S2 comprises the following steps:

[0018] Step S21: performing human outline recognition on the moving object focused image to obtain the moving human outline; performing human body region cutting on the moving object focused image based on the moving human outline to generate a human image;

[0019] Step S22: extract facial feature points from the person image to obtain facial feature data; accurately locate the pupils of the facial feature data to generate eye position data;

[0020] Step S23: performing human gaze simulation on the eye position data to generate simulated annotation data; performing gaze angle calculation on the simulated gaze data to obtain a gaze angle;

[0021] Step S24: performing time-series tracking on the gaze angle to obtain gaze trajectory data; performing attention tendency analysis based on the gaze trajectory data to generate character tendency data.

[0022] The present invention obtains the outline of the moving person by performing person outline recognition on the focused image of the moving object, so that the system can accurately extract the target area related to the user interaction. The person outline recognition ensures that the system stably locates the user's main position in the dynamic scene. The person image obtained by performing human body area cutting on the focused image of the moving object based on the moving person outline removes background interference and only retains the user's human body area, making the subsequent image processing more focused and effective. By extracting facial feature points from the person image, the extracted facial feature data reflects the key point information of the user's face, supporting the system to capture the accuracy and details of facial recognition, and accurately locates the pupil of the facial feature data to generate human eye position data. The pupil position, as the core element of the focus point, can accurately reflect the user's gaze direction, laying a data foundation for gaze angle and attention analysis. The accuracy of the human eye position data directly determines the system's The recognition accuracy of the user's focus point is achieved by simulating human gaze on the eye position data to generate simulated gaze data. The gaze simulation enables the system to restore the user's visual focus area. The simulated gaze data provides a reference for the system to accurately judge the user's current attention direction. The gaze angle is calculated on the simulated gaze data to generate the gaze angle. The gaze angle data quantifies the user's gaze direction and offset, so that the system can clearly track and analyze the user's gaze dynamics. The gaze trajectory data is obtained by time-series tracking of the gaze angle. The gaze trajectory data reflects the changing trajectory of the user's gaze direction over time. Through time-series tracking, continuous monitoring of the user's visual activities is achieved. The character tendency data generated by attention tendency analysis based on the gaze trajectory data further supports the system's in-depth judgment of the user's interests and attention content. The character tendency data provides data support for the system to intelligently respond to user preferences.

[0023] Preferably, step S24 comprises the following steps:

[0024] Performing a timestamp on the gaze angle to obtain time-series marked viewing angle data; performing a jump point detection on the time-series marked viewing angle data to generate viewing angle jump point data;

[0025] The dwell time of the view jump point data is counted to obtain the view retention data; the visual trajectory is reconstructed for the view retention data to generate the gaze trajectory data;

[0026] Performing regional clustering processing on the gaze trajectory data to obtain clustered regional data; performing frequency statistics on the clustered regional data to generate regional frequency data;

[0027] Perform attention pattern recognition on the regional frequency data to generate an attention pattern; perform target region matching on the attention pattern to obtain region matching data;

[0028] Performing continuity evaluation on the regional matching data to obtain continuity data; performing change trend analysis on the continuity data to generate continuity trend features;

[0029] The tendency of the continuity trend characteristics is judged to generate character tendency data.

[0030] The present invention obtains time-series marked perspective data by timestamping the gaze angle, so that the system can record and track the time-varying characteristics of the user's perspective. The time-series marked perspective data ensures the precise positioning of the user's visual behavior and provides detailed data support for subsequent behavior analysis and time dynamic statistics. Jump point detection is performed on the time-series marked perspective data to generate perspective jump point data. By identifying the mutation nodes of the user's perspective, the system can identify significant changes in the user's visual direction, which helps to identify the distraction or re-focusing of the user's attention. The perspective jump point data is subjected to residence time statistics to generate perspective residence data. The user's residence time at different perspectives is counted, which provides a basis for the system to judge the user's interest area. The perspective residence data supports the time intensity analysis of the user's focus area. The perspective residence data is subjected to visual trajectory reconstruction to generate gaze trajectory data. The reconstructed gaze trajectory data restores the complete motion trajectory of the user's perspective in the scene, ensuring that the system can track the user's dynamic visual behavior in real time. The gaze trajectory data is subjected to regional clustering processing to obtain clustering areas. Data, regional clustering simplifies the distribution characteristics of perspective data. Clustered regional data effectively divides the user's perspective focus areas, supports the system's comprehensive judgment of the main focus areas, performs frequency statistics on clustered regional data to generate regional frequency data, frequency data quantifies the user's attention to each area, and clarifies the user's high-frequency focus areas. Attention pattern recognition is performed on regional frequency data to generate attention patterns. Attention patterns reveal the long-term attention behavior characteristics of users in the scene and help the system understand the user's visual preferences. Attention patterns are matched to target areas to obtain regional matching data. Regional matching data supports the system to associate the user's perspective with a specific target area. Continuity evaluation is performed on regional matching data to obtain continuity data. Continuity data reveals the stability characteristics of user interests by measuring the continuity of user attention in the target area. Change trend analysis is performed on continuity data to generate continuity trend characteristics. Tendency judgment is performed on continuity trend characteristics to generate character tendency data. Tendency data summarizes the user's overall interest direction.

[0031] Preferably, step S3 comprises the following steps:

[0032] Step S31: Acquire interactive images of historical figures; perform time-series sorting on the interactive images of historical figures to obtain time-series sequence data;

[0033] Step S32: extracting character behaviors from the time series data to generate a behavior feature sequence; performing pattern clustering analysis on the behavior feature sequence to obtain historical interaction pattern data;

[0034] Step S33: performing feature vectorization processing on the historical interaction pattern data to obtain a pattern feature vector; performing prediction model parameter training on the pattern feature vector to generate prediction model parameters;

[0035] Step S34: performing prediction model integration on the prediction model parameters to generate a user input intention prediction model.

[0036] The present invention obtains historical character interaction images, and performs time-series sorting on the historical character interaction images to obtain time-series data, so that the system can capture the user's historical interaction behavior in time sequence, and provide ordered basic data for subsequent behavior analysis. The time-series data ensures the time correlation of the historical interaction images, thereby improving the accuracy of the system's time-series analysis when predicting user behavior. Character behavior is extracted from the time-series data to generate a behavior feature sequence. By extracting and serializing the user's historical behavior, the behavior features in the user interaction process are effectively extracted. The behavior feature sequence retains the integrity of the user's interaction behavior, and provides a data basis for the system to conduct in-depth user behavior pattern analysis. Pattern clustering analysis is performed on the behavior feature sequence to obtain historical interaction pattern data. Commonalities and regularities in user behavior are identified through pattern clustering, and feature vectorization is performed on the historical interaction pattern data. The pattern feature vector is obtained through processing. Feature vector quantization improves the compatibility of interaction pattern data in model training, so that the system can analyze and predict user behavior more accurately. The pattern feature vector enhances the computability of data and provides high-quality input for parameter training of subsequent prediction models. The pattern feature vector is trained to generate prediction model parameters. Based on the training of prediction model parameters, the system can generate prediction data that is more in line with user habits for the user's personalized interaction history. The prediction model parameters are integrated to generate a user input intention prediction model. A complete prediction model is constructed by integrating model parameters. The user input intention prediction model can accurately predict the user's next input intention after comprehensively considering the user's historical interaction characteristics, thereby improving the system's response speed to user needs and ensuring the real-time response of the smart TV system to user intentions.

[0037] Preferably, step S4 comprises the following steps:

[0038] Step S41: performing a threshold determination on the character tendency data. When the character tendency data is determined to be character tendency TV data, multiple rounds of intention effect simulation are performed on the moving object focus image based on the user intention prediction model to obtain an intention scene data set.

[0039] Step S42: performing reasonable boundary analysis on the intention scenario data set to obtain reasonable boundary data; performing extreme scenario inversion on the intention scenario data set based on the reasonable boundary data to generate a predicted user input intention;

[0040] Step S43: performing ambient light intensity detection on the focused image of the moving object to obtain ambient light intensity data; performing visual comfort calculation on the ambient light intensity data to generate visual pressure data;

[0041] Step S44: predicting the actual demand of the character's preferred TV data based on the visual pressure data to generate predicted character's preferred TV demand; and progressively adjusting the predicted user input intention according to the predicted character's preferred TV demand to generate predicted user input information.

[0042] The present invention performs threshold judgment on character tendency data, and when judging character tendency television data, performs multiple rounds of intention effect simulation on the focused image of the moving object based on the user intention prediction model to obtain the intention scene data set, which can effectively analyze the user's input tendency and needs, ensure the accurate capture of user input information, improve the system's prediction accuracy of user intention through multiple rounds of simulation, perform reasonable boundary analysis on the intention scene data set to obtain reasonable boundary data, perform extreme scene inversion on the intention scene data set based on the reasonable boundary data, generate predicted user input intention, ensure that the generated predicted user input intention meets the possible needs in the real scene, effectively avoid extreme input information through the inversion of extreme scenes, improve the reliability and adaptability of the prediction result, and perform ambient light intensity detection on the focused image of the moving object, Obtaining ambient light intensity data can accurately evaluate the lighting conditions of the user's viewing environment. The detection of ambient light intensity provides objective data support to ensure that subsequent visual comfort calculations can be performed based on accurate environmental parameters. The combination of ambient light intensity and visual comfort provides a basis for generating a TV interaction experience that better meets user needs. Based on visual pressure data, actual demand predictions are made for character-preference TV data to generate predicted character-preference TV needs. In-depth analysis of user viewing needs is provided in combination with ambient light and user visual pressure data. Actual demand predictions accurately capture user input needs in different scenarios. The generated predicted user input intentions are progressively adjusted to better meet user actual needs. The predicted input information is optimized and adjusted, which improves the personalization and accuracy of the TV interaction system.

[0043] Preferably, step S42 includes the following steps:

[0044] Extract scene features from the intention scene data set to obtain scene feature data; cluster and group the scene feature data to generate scene category data;

[0045] Perform distribution range statistics on the intention scenario data set according to the scenario category data to generate a distribution interval set;

[0046] Based on a preset interval threshold, an over-limit detection is performed on the distribution interval set to obtain an over-limit interval set; based on the over-limit interval set, a reasonable boundary screening is performed on the distribution interval set to generate reasonable boundary data;

[0047] Reconstruct the intention scene data set into a reasonable scene according to the reasonable boundary data to generate a reconstructed scene; calculate the difference of the reconstructed scene to obtain the difference data;

[0048] The difference data is analyzed for adjustment direction to generate adjustment direction data; the intention of the intention scene data set is reversely deduced based on the adjustment direction data to obtain the predicted user input intention.

[0049] The present invention obtains scene feature data by extracting scene features from the intention scene data set. By extracting scene features, more comprehensive and accurate scene information can be provided for subsequent analysis, ensuring that the judgment of user interaction intention is highly accurate. The scene category data generated by clustering and grouping provides a clear structure for data set classification. By statistics on the scene distribution range, the distribution and range of the data can be accurately understood, which helps to provide a basis for subsequent over-limit detection and reasonable boundary screening. The distribution interval set provides data support for analyzing the differences between different scenes, and improves the system's understanding and processing capabilities of scene data. Through over-limit detection, outliers or extreme data in scene data can be identified to ensure the rationality of input data and avoid invalid or erroneous intention data from affecting system decisions. The generation of over-limit interval sets provides an important basis for reasonable screening of data. Through reasonable boundary screening, data that does not meet actual needs is eliminated to ensure that the generated intention scene data set is more representative and accurate. The generation of reasonable boundary data guarantees It accurately predicts user input intentions and improves the personalization and response speed of user interaction experience. Through scene reconstruction, the scene data can be effectively adjusted to make it more in line with actual application needs and avoid prediction errors caused by data deviations. The reconstructed scene provides more optimized and accurate scene data for subsequent difference calculations. The difference calculation can quantify the difference between the reconstructed scene and the original scene, providing a basis for adjustment and optimization. The difference data can reveal the specific changes and deviations between different scenes, which helps to further optimize the intention derivation. Through adjustment direction analysis, directional guidance can be provided for subsequent intention reverse derivation to ensure that the generated predicted user input intention is more accurate. The generation of adjustment direction data enables the system to flexibly adjust its response strategy according to changes in different scenes. Through intention reverse derivation, the user behavior characteristics and the adjustment direction of the scene data are combined. The generated predicted user input intention is more in line with actual needs, providing more accurate user interaction feedback and improving the accuracy and personalization of user experience.

[0050] Preferably, step S44 includes the following steps:

[0051] Conduct correlation analysis on visual stress data and character preference TV data to obtain stress-demand correlation data;

[0052] Conduct corresponding impact analysis on pressure-demand correlation data to generate visual pressure impact data;

[0053] Based on the visual pressure influence data, the influence elimination simulation is performed on the character preference TV data to obtain the simulated demand tendency data; the actual demand is estimated on the simulated demand tendency data to generate the predicted character preference TV demand;

[0054] Perform difference comparison processing on the predicted TV demand of the person and the predicted user input intention to obtain demand difference data;

[0055] The predicted user input intention is gradually adjusted according to the demand difference data to generate predicted user input information.

[0056] The present invention obtains pressure-demand correlation data by performing correlation analysis on visual pressure data and character preference TV data. The correlation analysis can reveal the intrinsic relationship between visual pressure and user preference TV demand, so that the system can predict user demand based on real-time visual pressure information, thereby accurately capturing user behavior patterns and improving the refinement level of user interaction. The specific effect of visual pressure on user preference TV demand can be quantified through impact analysis, so that the system can further understand the impact of the external environment on user behavior and provide a more scientific demand prediction model. The deviation caused by visual pressure can be effectively eliminated through impact elimination simulation, ensuring that the demand tendency data is more representative and authentic. The generation of simulated demand tendency data provides a more accurate basis for subsequent demand estimation. The demand estimation combines simulated demand tendency data and user behavior characteristics to provide an accurate prediction model, so that the system can accurately infer the user's TV demand and meet the user's preference in advance. The deviation between demand and intention can be quantified through difference comparison, and the coordination between intention and demand can be further optimized, so that the system can better refine and adjust the user demand. The system can gradually optimize the user input information according to the demand difference through gradual adjustment, so that the predicted user input is more in line with the actual demand and behavior pattern, and the adaptability and accuracy of the system to variable user behavior are improved.

[0057] Preferably, step S5 comprises the following steps:

[0058] Step S51: extracting logical relationships from the predicted user input information to obtain basic logical relationships; verifying constraint conditions on the basic logical relationships to generate constraint verification data;

[0059] Step S52: Conflict-marking the constraint verification data to obtain conflict-marking data; identifying the conflict type of the conflict-marking data to generate conflict detection data;

[0060] Step S53: tracing the conflict detection data to obtain the key factors of the conflict; performing logic adjustment simulation on the predicted user input information based on the key factors of the conflict to generate multiple simulation adjustment data;

[0061] Step S54: performing a second round of logic conflict verification on the plurality of simulation adjustment data to obtain second conflict detection data; performing a difference comparison on the conflict detection data according to the second conflict detection data to generate conflict change data;

[0062] Step S55: screening conflict factors for key conflict factors based on the conflict change data to generate optimized candidate conflict factors; performing feedback optimization processing on the predicted user input information according to the optimized candidate conflict factors to generate user input information.

[0063] The present invention extracts logical relationships from predicted user input information to obtain basic logical relationships, which can deeply explore the potential logical structure of user input, ensure the accuracy and compliance of input information, and provide a clear framework for subsequent constraint verification, so that the system can more accurately analyze and process the logical relationships of user input; the basic logical relationships are verified by constraint conditions to generate constraint verification data, which can ensure that the logical relationships of user input are correctly applied under specific constraint conditions; the constraint verification data are conflict-marked to obtain conflict marking data, and the marked conflict data can effectively identify and indicate possible contradictions or inconsistencies within the system, thereby identifying and processing potential problems in advance, and improving the system's processing accuracy and robustness for user input information; the conflict marking data are identified by conflict type to generate conflict detection data, which can further classify and identify the marked conflicts, so that the system can accurately diagnose different types of conflicts and accurately locate potential system problems; the conflict detection data is traced to obtain the key factors of the conflict, and the tracing process can trace the root cause of the conflict, which helps to discover and identify potential errors or anomalies within the system or in user input information, and ensure that the conflict is resolved. to effectively solve the problem; based on the key conflict factors, the predicted user input information is simulated for logical adjustment to generate multiple simulated adjustment data. The simulated adjustment can adjust the predicted information according to the conflict source to ensure that the prediction result is more in line with the actual demand. The multiple simulated adjustment data are subjected to a second round of logical conflict verification to obtain the second conflict detection data. The second round of verification can re-verify the result after the first adjustment, further identify and solve the residual conflict problem, and ensure the consistency and rationality of the final output result; the conflict detection data is compared with the second conflict detection data to generate conflict change data. The difference comparison can quantify the changes under different conflict states and accurately guide the conflict resolution process. The conflict key factors are screened based on the conflict change data to generate optimized candidate conflict factors. The screening process can focus on the most critical conflict factors, avoid invalid adjustments, and improve the efficiency and accuracy of conflict resolution; the predicted user input information is fed back and optimized according to the optimized candidate conflict factors to generate user input information. The optimization process can adjust the user input information according to the screened key conflict factors, so that the prediction result is more accurate and meets user needs, and the system's adaptability to variable user input is improved.

[0064] The present invention also provides a user input information system in an all-in-one smart TV, which is used to execute the user input information method in an all-in-one smart TV, and the user input information system in an all-in-one smart TV includes:

[0065] The image acquisition module is used to acquire the current user image; perform white balance correction on the current user image to obtain a color-corrected image; perform dynamic image focusing on the color-corrected image to generate a moving object focused image;

[0066] The attention analysis module is used to perform person recognition on the focused image of the moving object to obtain a person image; perform attention tendency analysis on the person image to generate person tendency data;

[0067] A prediction modeling module is used to obtain interactive images of historical figures; perform interaction pattern recognition on the interactive images of historical figures to obtain historical interaction pattern data; perform user input intention prediction modeling on the historical interaction pattern data to generate a user input intention prediction model;

[0068] An information prediction module is used for, when the character tendency data is determined to be character tendency television data, performing a user input intention drill on the moving object focused image based on the user intention prediction model to generate a predicted user input intention; performing auxiliary adjustment on the predicted user input intention according to the character tendency data and the moving object focused image to generate predicted user input information;

[0069] The closed-loop optimization module is used to perform logical conflict detection on the predicted user input information to obtain conflict detection data; and perform feedback optimization processing on the predicted user input information based on the conflict detection data to generate user input information.

[0070] The present invention ensures that the acquired image data has true color and high visual accuracy by collecting and white-balancing the current user image. The color-corrected image after white-balancing correction is used as a basis, and then a dynamic image focusing operation is performed on it to generate a moving object focused image, ensuring that the picture captured by the smart TV system is clearer at the visual level, so that the subsequent analysis process is based on high-quality image data, which can minimize the recognition error caused by poor image quality, especially in dynamic scenes, helping to more accurately capture the user's feature image. The attention analysis module effectively separates the user's character image by performing character recognition on the moving object focused image, ensuring the accurate positioning of the analysis object, and The attention tendency analysis module is used to analyze the attention tendency of the character images and obtain the tendency data of the user's attention area or object. The tendency data directly reflects the user's current interests and concerns, enabling the system to more accurately judge the user's potential needs and interaction intentions. The introduction of the attention tendency analysis module enables the system to capture and analyze the key points of user behavior in real time, improving the naturalness and intelligence of the human-computer interaction process. The prediction modeling module obtains and analyzes historical character interaction images, generates historical interaction pattern data using interaction pattern recognition technology, and further generates a user input intention prediction model based on input intention prediction modeling based on historical data, ensuring that the system can use historical behavior data to predict the user's interaction tendency. The model can predict the user's interaction tendency based on the user's historical operations and interaction patterns. The prediction modeling module builds a high-precision user intention prediction framework through comprehensive analysis of the user's past operation data, providing highly adaptive interactive support for the smart TV system. When the character tendency data determines that the user's attention object is the TV, the information prediction module conducts a user input intention drill on the moving object focus image based on the user intention prediction model to generate a predicted user input intention, and makes auxiliary adjustments to the prediction results based on the character tendency data and the moving object focus image to generate more accurate predicted user input information. The information prediction module ensures that the generated prediction information is more in line with the user's actual needs through the drill and adjustment process. In addition, the system performs comprehensive processing on the character tendency data and the focus image of the moving object during the prediction process, so that the prediction result can adapt to the user's real-time attention changes and improve the prediction accuracy. The closed-loop optimization module performs logical conflict detection on the predicted user input information, and the generated conflict detection data is used to identify potential logical contradictions or misoperation risks in the predicted information. Based on the conflict detection data, the predicted user input information is feedback optimized and processed, and finally highly reliable user input information is generated. The closed-loop optimization module ensures the rationality of the user input information through logical conflict detection and feedback optimization processing, effectively avoids erroneous instructions or inaccurate responses caused by logical conflicts, and further improves the human-computer interaction fluency and intelligence of the smart TV system. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A schematic diagram of a process flow of a method for user inputting information in a smart TV all-in-one machine;

[0072] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0073] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0074] 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

[0075] 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.

[0076] 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.

[0077] 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.

[0078] To achieve this, please refer to Figures 1 to 3 , a method for user inputting information in a smart TV all-in-one machine, comprising the following steps:

[0079] Step S1: capturing a current user image; performing white balance correction on the current user image to obtain a color-corrected image; performing dynamic image focusing on the color-corrected image to generate a moving object focused image;

[0080] Step S2: performing person recognition on the focused image of the moving object to obtain a person image; performing attention tendency analysis on the person image to generate person tendency data;

[0081] Step S3: Acquire historical figure interaction images; perform interaction pattern recognition on the historical figure interaction images to obtain historical interaction pattern data; perform user input intention prediction modeling on the historical interaction pattern data to generate a user input intention prediction model;

[0082] Step S4: when the character tendency data is determined to be character tendency television data, a user input intention drill is performed on the moving object focused image based on the user intention prediction model to generate a predicted user input intention; auxiliary adjustments are made to the predicted user input intention according to the character tendency data and the moving object focused image to generate predicted user input information;

[0083] Step S5: performing logical conflict detection on the predicted user input information to obtain conflict detection data; performing feedback optimization processing on the predicted user input information based on the conflict detection data to generate user input information.

[0084] The present invention acquires a corrected image of true color by collecting the current user image and performing white balance correction, which helps to improve the color accuracy of the image. By focusing the dynamic image, a clear image of the moving object is captured, thereby ensuring the accuracy of subsequent character recognition and attention analysis. By white balance correction and focusing processing, high-quality adjustment of the input image is achieved, so that the smart TV can obtain clear and accurate user image source data, thereby improving the reliability of the recognition effect. Character recognition is performed on the focused image of the moving object to obtain a clear user character image. Then, through attention tendency analysis, tendency data reflecting the user's gaze or attention area is generated based on the analysis algorithm, effectively realizing the capture of the user's attention state, laying a data foundation for predicting the user's interaction intention. The tendency data directly reflects the user's current interest area, thereby providing a basis for subsequent prediction. Through this recognition and analysis process, the smart TV can quickly and accurately focus on the user's interest area, acquire and process historical character interaction images, acquire historical interaction mode data through interaction mode recognition, and further construct a user input intention prediction model. The model is based on the data of historical interaction behavior to form an algorithm framework for predicting the user's operation intention. The input intention prediction model can effectively associate the user's habits in specific scenarios. Operation mode, based on the learning of historical patterns, enables the prediction model to more accurately reflect the user's intention, provides a data basis for the reliability and customization of further operations, and enhances the applicability of the prediction effect. When the character tendency data indicates that the user tends to TV data, the user intention prediction model is used to perform intention drills on the focused image of the moving object to predict the user's input intention, and then the predicted intention is auxiliary adjusted according to the tendency data and the focused image to generate accurate user input information. Dynamic update prediction is achieved through data adjustment to ensure that the prediction results are more in line with the user's immediate needs. The auxiliary adjustment function realizes the precise matching of intention data and real-time image information, enabling smart TVs to generate operation information that is more in line with user needs, ensuring the accuracy and real-time response of prediction information, performing logical conflict detection on the generated predicted user input information, identifying and handling potential operation conflict risks, and adjusting the user input information through feedback optimization to ensure that the generated user input information can smoothly and smoothly guide the TV's response. Through conflict detection and optimization processing, smart TVs can effectively avoid errors caused by repetitive and contradictory operations. Through the adjustment mechanism of logical optimization, the rationality of user input information is ensured, enhancing the fluency of smart TV interactive operations.

[0085] In the embodiment of the present invention, refer to Figure 1 , is a schematic flow chart of the steps of a method for user inputting information in a smart TV all-in-one according to the present invention. In this example, the method for user inputting information in a smart TV all-in-one comprises the following steps:

[0086] Step S1: capturing a current user image; performing white balance correction on the current user image to obtain a color-corrected image; performing dynamic image focusing on the color-corrected image to generate a moving object focused image;

[0087] In this embodiment, the current user image is captured by a built-in high-definition camera. The image capture module is used in the image capture process to ensure that the image resolution reaches 1080p or above, and the acquisition speed can track the user dynamics in real time. White balance correction is performed on the captured image, and the Gray World Algorithm is used to automatically correct the image color temperature. By detecting the deviation of the three tones of red, green and blue in the image, the color temperature of the light source is adjusted so that the image maintains a natural color in any lighting environment. Dynamic image focusing processing is performed, and a convolutional neural network (CNN) model based on deep learning is used to track moving objects in real time, identify fast-moving objects in the image, and focus on them through corresponding algorithms (such as a dynamic blur compensation algorithm) to ensure the clarity and accuracy of the moving object in the image, and finally generate a focused image of the moving object for subsequent analysis and use.

[0088] Step S2: performing person recognition on the focused image of the moving object to obtain a person image; performing attention tendency analysis on the person image to generate person tendency data;

[0089] In this embodiment, character recognition is performed through a deep convolutional neural network (CNN), and pre-trained models such as Faster R-CNN or YOLO (You Only Look Once) are used to detect characters in the image. A target detection algorithm is used to accurately locate and select each character in the image. Based on these character images, a deep learning model based on an attention mechanism is used to analyze the character's attention tendency, analyze the character's gaze direction, body language, and facial expressions, and infer the character's current interest points and attention distribution. Through the sample data in the training set, the model can identify the character's attention to specific TV content or a specific screen area, thereby generating corresponding character tendency data, indicating whether the character tends to pay attention to the TV screen, or other possible interactive behaviors.

[0090] Step S3: Acquire historical figure interaction images; perform interaction pattern recognition on the historical figure interaction images to obtain historical interaction pattern data; perform user input intention prediction modeling on the historical interaction pattern data to generate a user input intention prediction model;

[0091] In this embodiment, historical character interaction images are retrieved from a database. These images include past records of interactions between users and television devices. They are preprocessed using an image processing algorithm to remove background noise and irrelevant elements to ensure that the data is clean and efficient. An interaction pattern recognition algorithm is used, and a time series modeling method based on LSTM (Long Short-Term Memory Network) is adopted to model the character interaction behaviors in historical images, and identify the character's behavior patterns in past interactions, such as viewing time duration and interaction methods (such as remote control buttons, touch screen, etc.). By learning these patterns, the user's typical behaviors in different situations are extracted, thereby generating historical interaction pattern data, which represents the user's preferences and habits in past interactions. Using these historical interaction pattern data, a user input intention prediction model is constructed using a machine learning algorithm (such as a decision tree or a support vector machine). The model predicts the user's future input intentions based on historical interaction behaviors, such as whether the user will choose to switch channels or adjust the volume.

[0092] Step S4: when the character tendency data is determined to be character tendency television data, a user input intention drill is performed on the moving object focused image based on the user intention prediction model to generate a predicted user input intention; auxiliary adjustments are made to the predicted user input intention according to the character tendency data and the moving object focused image to generate predicted user input information;

[0093] In this embodiment, based on the user intention prediction model, the user input intention is rehearsed on the focused image of the moving object, the image is input and combined with the tendency data of the character, and the model is used to predict the user's possible input behavior. By combining deep learning image analysis with behavior prediction, the model generates a predicted user input intention. For example, based on the eye tracking and posture changes of the character, it is predicted that the character may control the TV to switch channels through voice or gestures. Based on the tendency data of the character and the focused image of the moving object, an adaptive adjustment algorithm is used to perform auxiliary adjustments to the predicted user input intention. Combined with data such as emotion analysis, user historical behavior and current environmental changes, the predicted input intention is refined and adjusted to generate the final predicted user input information, which plays a guiding role in the interaction between the user and the TV.

[0094] Step S5: performing logical conflict detection on the predicted user input information to obtain conflict detection data; performing feedback optimization processing on the predicted user input information based on the conflict detection data to generate user input information.

[0095] In this embodiment, a match check is performed between the user input content and the current state of the television device through a set rule engine. For example, when the user input information conflicts with the current channel or volume setting of the television, the system will detect the conflict and record the information. A conflict detection algorithm based on constraint programming (such as SAT Solver) is used to verify the logical relationship between multiple input information to ensure that all input intentions are consistent with the current state of the device. After the logic check, conflict detection data is generated to indicate which input intentions are in conflict. Based on the conflict detection data, a feedback optimization algorithm is used to optimize the predicted user input information. Optimization methods such as genetic algorithms are used to adjust the conflicting parts to ensure that the final user input information can both eliminate conflicts and meet user needs to the greatest extent. After optimization, the final user input information is generated to ensure that the interaction process is smooth and meets user expectations.

[0096] Preferably, step S1 comprises the following steps:

[0097] Step S11: capturing the current user image; performing color temperature analysis on the current user image to obtain color temperature information of the user image; performing white balance correction calculation on the color temperature information of the user image to generate white balance correction parameters;

[0098] Step S12: performing color correction on the current user image based on the white balance correction parameters to obtain a color-corrected image;

[0099] Step S13: performing motion region detection on the color-corrected image to obtain a motion region image; performing motion trajectory tracking on the motion region image to obtain a motion region trajectory;

[0100] Step S14: dynamically focusing the moving area image according to the moving area trajectory to generate a moving object focused image.

[0101] In this embodiment, an image of the current user is captured by a built-in high-definition camera. A dedicated image capture module is used in the image capture process to ensure that the clarity of the captured image reaches the standard of 1920x1080. A sufficient sampling frequency is ensured during image capture to meet the real-time processing requirements. Next, a color temperature analysis is performed on the captured image. A color temperature detection algorithm (such as a gray world algorithm or a maximum RGB algorithm) is used to analyze the impact of the current ambient light source on the image, and the color temperature value of the image is calculated to obtain the color temperature information of the user image. Then, a white balance correction calculation is performed based on the color temperature information. The correction is performed using a white balance algorithm (such as an RGB channel gain algorithm) to obtain white balance correction parameters. The correction parameters include RGB gain values, color temperature compensation values, etc., which are used for subsequent color correction processing. Color correction is performed on the current user image. The color correction algorithm is used to calculate the image. The color deviation of each pixel in the image is calculated, and the RGB channels are adjusted using the white balance correction parameters obtained previously, so as to correct the color temperature problem in the image and ensure that the color of the image is more realistic, especially under different lighting conditions, the color performance of the image is more stable and natural. During processing, all correction operations are based on a certain accuracy range to ensure that color correction does not cause excessive distortion of the image. The generated color-corrected image contains the adjusted color information to meet the image quality requirements of the device. The background subtraction algorithm or optical flow method is used for preliminary extraction of the motion area. In this process, computer vision technology is used to identify the moving objects in the image, and the position and shape of the motion area are determined by comparing the pixel changes between consecutive frames. The detected motion area will be calibrated as a motion area image. Then, the motion trajectory of the motion area image is tracked using Kalman. The Kalman filter algorithm smoothes and predicts the trajectory of moving objects, tracks the boundaries of moving areas in the image, and obtains a complete trajectory of the moving area. The trajectory includes the object's motion path, speed changes, and direction. By analyzing the trajectory of the moving area, the object's speed and direction are determined. A deep learning-based image processing model (such as a convolutional neural network (CNN)) is used for dynamic focusing. The model learns the characteristics of moving objects through training and accurately focuses on the image based on the object's trajectory to ensure that the moving object remains clear in every frame of the image. The motion prediction algorithm is used to predict the object's future position to ensure that the focusing module can adjust the focus in real time and generate a focused image of the moving object. The moving object in the image will present a clearer and more accurate display effect, ensuring that the details of the image are displayed intact.

[0102] Preferably, step S2 comprises the following steps:

[0103] Step S21: performing human outline recognition on the moving object focused image to obtain the moving human outline; performing human body region cutting on the moving object focused image based on the moving human outline to generate a human image;

[0104] Step S22: extract facial feature points from the person image to obtain facial feature data; accurately locate the pupils of the facial feature data to generate eye position data;

[0105] Step S23: performing human gaze simulation on the eye position data to generate simulated annotation data; performing gaze angle calculation on the simulated gaze data to obtain a gaze angle;

[0106] Step S24: performing time-series tracking on the gaze angle to obtain gaze trajectory data; performing attention tendency analysis based on the gaze trajectory data to generate character tendency data.

[0107] As an example of the present invention, see Figure 2 In this example, step S2 includes:

[0108] Step S21: performing human outline recognition on the moving object focused image to obtain the moving human outline; performing human body region cutting on the moving object focused image based on the moving human outline to generate a human image;

[0109] In this embodiment, a convolutional neural network (CNN) is used to segment objects. The trained network model is used to identify the contours of moving objects in the image, with special attention paid to the edges of moving people. The contours of the moving people are accurately outlined using an edge detection algorithm (such as Canny edge detection). Then, by analyzing the contours of the moving people, an algorithm based on region growing or a graph cutting algorithm (such as the GrabCut algorithm) is applied to cut the human body region in the focused image of the moving object, and the background and other non-human parts are removed to generate a person image. The generated person image only retains the person region, and the background information is discarded, in preparation for subsequent facial feature extraction and other processing.

[0110] Step S22: extract facial feature points from the person image to obtain facial feature data; accurately locate the pupils of the facial feature data to generate eye position data;

[0111] In this embodiment, a face detection algorithm based on deep learning (such as Haar Cascade or MTCNN algorithm) is used to locate the face and identify the facial area in the character image. Then, a facial key point detection model (such as a facial feature point detector in the Dlib library) is used to process the face area and extract the key feature points of the face, including the position coordinates of the eyes, nose, mouth, etc., to obtain facial feature data. Subsequently, based on the facial feature data, the pupil is further accurately located and a pupil detection algorithm (such as a pupil location technology based on image grayscale and morphological processing) is used to accurately find the center position of the pupil by analyzing the contrast and circular features of the eye area, thereby generating accurate eye position data. The obtained data includes the coordinate information of the left and right eyes to ensure the accuracy of eye positioning.

[0112] Step S23: performing human gaze simulation on the eye position data to generate simulated annotation data; performing gaze angle calculation on the simulated gaze data to obtain a gaze angle;

[0113] In this embodiment, eye tracking technology (such as an eye tracking algorithm based on computer vision) is used to simulate the gaze direction of the eyes using human eye position data. The simulation process combines the relative position of the human eye in the image and the angular relationship between the gaze object to generate simulated gaze data. The gaze data includes information such as the direction of the eyeball, the position of the gazed object, and the duration of gaze. Then, the gaze angle is calculated for the simulated gaze data. The trigonometric function in geometry is used to calculate the angle between the gazed object and the eye. The gaze angle is calculated based on triangulation of the human eye position and the coordinates of the object in the image to obtain the gaze angle. This angle reflects the relationship between the direction the human eye is pointing and the target object, thereby ensuring the accuracy of the gaze angle calculation.

[0114] Step S24: performing time-series tracking on the gaze angle to obtain gaze trajectory data; performing attention tendency analysis based on the gaze trajectory data to generate character tendency data.

[0115] In this embodiment, by collecting multiple frames of image data, a Kalman filter is used to smooth the gaze angle, and the angle error caused by slight eye shaking in a short period of time is filtered out to ensure the stability of the gaze angle. During the tracking process, the gaze trajectory data is updated frame by frame to obtain the gaze trajectory data. The gaze trajectory data includes the changes in the gaze direction and the movement trajectory of the eyes at multiple time points. Then, attention tendency analysis is performed based on the gaze trajectory data. The gaze trajectory data is analyzed using a deep learning-based pattern recognition algorithm (such as a long short-term memory network LSTM) to extract the user's attention tendency information. The generated character tendency data includes the user's attention to different objects or areas, gaze duration, and gaze pattern, etc., which provides a basis for subsequent user input information prediction and intelligent interaction.

[0116] Preferably, step S24 comprises the following steps:

[0117] Performing a timestamp on the gaze angle to obtain time-series marked viewing angle data; performing a jump point detection on the time-series marked viewing angle data to generate viewing angle jump point data;

[0118] The dwell time of the view jump point data is counted to obtain the view retention data; the visual trajectory is reconstructed for the view retention data to generate the gaze trajectory data;

[0119] Performing regional clustering processing on the gaze trajectory data to obtain clustered regional data; performing frequency statistics on the clustered regional data to generate regional frequency data;

[0120] Perform attention pattern recognition on the regional frequency data to generate an attention pattern; perform target region matching on the attention pattern to obtain region matching data;

[0121] Performing continuity evaluation on the regional matching data to obtain continuity data; performing change trend analysis on the continuity data to generate continuity trend features;

[0122] The tendency of the continuity trend characteristics is judged to generate character tendency data.

[0123] In this embodiment, a real-time image acquisition system is used to continuously acquire the gaze angle of the user's eyes. The user's eye position data is captured by a high frame rate camera, and a timestamp is added to each frame of data. The timestamp data marks the specific acquisition time of each gaze angle data to form time-series marked viewing angle data, thereby ensuring that the viewing angle data at each moment has a clear time identifier, which is convenient for subsequent time series analysis. The accuracy of the timestamp depends on the time synchronization capability of the image acquisition system. The data is then transmitted to the data processing system for further processing, and a jump point detection is performed on the time-series marked viewing angle data. A jump point detection algorithm based on statistics (such as the CUSUM algorithm) is used to analyze the time-series marked viewing angle data. The algorithm compares the viewing angle changes in adjacent time periods. The amplitude of the change is detected, and the time points when the perspective changes dramatically are detected, and these are marked as jump points. Jump points usually indicate that the gaze angle has changed significantly. After detecting these jump points, the perspective jump point data is obtained. The jump point data contains the moment when the gaze angle suddenly changes and the corresponding change in perspective, which provides important reference information for the subsequent stay time statistics. The stay time statistics of the perspective jump point data are performed. The stay time statistics algorithm is used to calculate the user's stay time in the same gaze area for the continuous perspective data between each jump point. Specifically, the time difference between the timestamps is calculated, and the duration of each perspective segment is counted to obtain the perspective residence data. The residence data indicates the length of time the user stays in a certain gaze area. The interval of perspective change is determined based on the jump point. Furthermore, the user's attention duration in each area is accurately recorded by means of timestamp marking, and finally a statistical result is obtained to show the distribution of the duration of stay in each area. The duration statistics are combined with the viewpoint data of each time point, and the user's gaze trajectory is smoothly connected using an interpolation algorithm (such as spline interpolation) to reconstruct the user's visual trajectory. The visual trajectory reconstruction is based on the time series data of each viewpoint, combined with the regional position and gaze angle information, to generate gaze trajectory data. The trajectory data includes the changes in the path during the gaze process, as well as the time and position of each path point, thereby generating a complete gaze path map, showing the user's line of sight movement and regional transfer process. The gaze trajectory data is subjected to regional clustering processing, and a clustering algorithm (such as K -means or DBSCAN algorithm) clusters different areas in the gaze trajectory data and divides the trajectory points into different areas. The clustering algorithm divides the area based on the changes in the view data and timestamps. The clustering process calculates the spatial and temporal distances between the gaze points to obtain the cluster area data. The cluster area data includes the boundary of each user's gaze area, the density of gaze points in the area and its duration, etc., reflecting the hot spots of interest in the user's gaze process. The cluster area data is frequency counted. According to the clustering results, the number of gazes in each cluster area is counted, and the frequency of gaze points in each area is counted. The frequency statistics method counts all the trajectory points in the area and calculates the number of times each area is gazed.The regional frequency data is obtained. The regional frequency data reflects the user's attention level to different regions. Through frequency statistics, the region where the user focuses the most and the intensity of the focus can be identified. Attention pattern recognition is performed on the regional frequency data. Based on the regional data of frequency statistics, the regional frequency data is analyzed using a deep learning algorithm (such as a convolutional neural network (CNN)) to identify the user's attention pattern. In the pattern recognition process, the user's gaze pattern is identified through a trained network model, such as long-term gaze on a specific area, or frequent switching of multiple areas, and finally the user's attention pattern is generated. The attention pattern data contains the rules of the user's gaze behavior, such as gaze time, frequency, and changes in the gaze area. The attention pattern is matched with the target area. The target recognition algorithm (such as a regional matching algorithm based on image recognition) is used to match the user's gaze pattern with the area on the TV interface to identify the specific target area that the user is focusing on. The target area matching algorithm compares the user's gaze visual trajectory with the content on the TV screen and matches it to a specific target area, such as an advertisement, a video playback area, etc., to obtain regional matching data. The regional matching data records the type and position of the user's gaze target area. Conduct continuity evaluation, use time series analysis methods to evaluate the continuity of regional matching data. During the evaluation process, analyze the gaze duration and change trend of each target area, evaluate the gaze time length of different areas, calculate the gaze continuity of each area, and obtain continuity data. The continuity data includes the total gaze time of the user on each target area, the longest gaze time and other information. Perform change trend analysis on the continuity data, use time series analysis algorithms (such as ARIMA model) to analyze the change trend of the continuity data, analyze the change of gaze continuity over time, and obtain continuity trend characteristics. The continuity trend characteristics reflect the changing law of the user's gaze pattern, such as whether the gaze time of a certain area has an increasing or decreasing trend, whether certain areas are continuously paid attention to or ignored, and make a tendency judgment on the continuity trend characteristics. Use decision trees or support vector machines (SVM) and other classification algorithms to analyze the continuity trend characteristics and judge the user's tendency to a certain type of target area. The tendency judgment is based on the continuity trend characteristics, comprehensively evaluate the user's preference for specific areas, and generate the user's tendency data. The tendency data records the user's attention preference for certain specific target areas and provides a basis for predicting user input information. ,

[0124] Preferably, step S3 comprises the following steps:

[0125] Step S31: Acquire interactive images of historical figures; perform time-series sorting on the interactive images of historical figures to obtain time-series sequence data;

[0126] Step S32: extracting character behaviors from the time series data to generate a behavior feature sequence; performing pattern clustering analysis on the behavior feature sequence to obtain historical interaction pattern data;

[0127] Step S33: performing feature vectorization processing on the historical interaction pattern data to obtain a pattern feature vector; performing prediction model parameter training on the pattern feature vector to generate prediction model parameters;

[0128] Step S34: performing prediction model integration on the prediction model parameters to generate a user input intention prediction model.

[0129] As an example of the present invention, see Figure 3 In this example, step S3 includes:

[0130] Step S31: Acquire interactive images of historical figures; perform time-series sorting on the interactive images of historical figures to obtain time-series sequence data;

[0131] In this embodiment, interactive images of historical figures are obtained through a multi-channel video acquisition system. The image content involves figures from multiple historical periods and includes interactive information between them. The system uses high-resolution camera equipment to ensure the accuracy of the details of each frame of the image. The acquired images are saved in chronological order through a video storage device, and the accurate synchronization of data timestamps is ensured to ensure that each frame of the image has an accurate time mark. The interactive images of historical figures are aggregated from multiple sources to obtain image data containing complete interactive scenes of historical figures. The interactive images of historical figures are sorted according to their timestamps to ensure that the images are arranged in the order of occurrence. If there are images without timestamps, image recognition technology (such as a target detection method based on a convolutional neural network (CNN)) is used to analyze the image content, infer its relative time position, and then sort it to finally obtain a time series data, in which the time relationship of each frame of the image is accurately calibrated to construct a time series of interactive images of historical figures.

[0132] Step S32: extracting character behaviors from the time series data to generate a behavior feature sequence; performing pattern clustering analysis on the behavior feature sequence to obtain historical interaction pattern data;

[0133] In this embodiment, an image processing algorithm (such as background modeling and motion detection method) is used to extract the behavior information of historical figures. The image processing system identifies each historical figure in the time series by extracting features such as the figure's outline, posture and motion trajectory in each frame of the image. Then, a deep learning model (such as a long short-term memory network LSTM) is used to analyze the image sequence to extract the figure's behavioral characteristics, such as the figure's action type, interaction mode and expression changes. Finally, a behavior feature sequence is generated, which contains the behavior label and related features of the figure at each time node, and defines the dimensions of the behavior features, including action type, interaction mode, facial expression, etc. After these features are vectorized, a clustering algorithm is used to cluster the behavior feature sequence. The clustering algorithm classifies similar behavior patterns into the same category based on the similarity of the behavior features. During the clustering analysis, multiple historical interaction pattern data are obtained by calculating the distance of each behavior feature data point. The historical interaction pattern data finally output includes typical behavior patterns of historical figures' interactions in different time periods.

[0134] Step S33: performing feature vectorization processing on the historical interaction pattern data to obtain a pattern feature vector; performing prediction model parameter training on the pattern feature vector to generate prediction model parameters;

[0135] In this embodiment, a feature engineering method is used to extract features from each behavior pattern in the historical interaction pattern data, and information such as actions, facial expressions, and interactive behaviors in the behavior pattern are converted into numerical feature vectors. The specific steps include using time domain analysis based on Fourier transform to extract features from the time series in the behavior pattern, converting it into frequency domain features, and further reducing the dimension using the principal component analysis (PCA) method to obtain feature vectors corresponding to each behavior pattern. These feature vectors accurately represent the main features of the historical interaction pattern. The feature vector data is divided into a training set and a test set. The feature vectors in the training set are trained using models such as support vector machines (SVM) or deep neural networks (DNN). The model learns the features of the historical interaction pattern and fits the model parameters. These training parameters are used to predict unknown data. The prediction model parameters finally generated include the learning results of the historical interaction pattern and can effectively predict future interactive behaviors.

[0136] Step S34: performing prediction model integration on the prediction model parameters to generate a user input intention prediction model.

[0137] In this embodiment, the parameters of multiple trained prediction models are integrated, and the output results of multiple prediction models are weighted and fused using an integrated learning method (such as random forest or gradient boosting machine GBDT). The integrated learning algorithm takes the prediction result of each model as input, and generates a comprehensive prediction result by weighted calculation of the outputs of different models. The comprehensive prediction model finally generated can more accurately predict the user's input intention and generate a user input intention prediction model. Combined with the historical interaction mode data and the integration results of the prediction model, the integrated data is further analyzed using a decision tree algorithm to determine the user's input intention under different interaction modes and form a prediction model of the user's input intention. The prediction model can predict the user's interaction needs in the current scenario based on the user's input behavior pattern through deep learning of the historical interaction data. The generated prediction model not only includes the category of the user's input intention, but also can infer the user's possible input content in the future interaction process. The user input intention prediction model finally generated has high accuracy and can respond to user input information in real time.

[0138] Preferably, step S4 comprises the following steps:

[0139] Step S41: performing a threshold determination on the character tendency data. When the character tendency data is determined to be character tendency TV data, multiple rounds of intention effect simulation are performed on the moving object focus image based on the user intention prediction model to obtain an intention scene data set.

[0140] Step S42: performing reasonable boundary analysis on the intention scenario data set to obtain reasonable boundary data; performing extreme scenario inversion on the intention scenario data set based on the reasonable boundary data to generate a predicted user input intention;

[0141] Step S43: performing ambient light intensity detection on the focused image of the moving object to obtain ambient light intensity data; performing visual comfort calculation on the ambient light intensity data to generate visual pressure data;

[0142] Step S44: predicting the actual demand of the character's preferred TV data based on the visual pressure data to generate predicted character's preferred TV demand; and progressively adjusting the predicted user input intention according to the predicted character's preferred TV demand to generate predicted user input information.

[0143] In this embodiment, the acquired tendency data of each character is compared according to a predetermined threshold range, and the tendency threshold range is set to [0, 1]. When the tendency data of the character is greater than 0.7, it is determined to be "character tendency TV data". The threshold determination is implemented by a data analysis tool, and the tendency data of each character is updated and detected in real time using a data processing algorithm to ensure that each determination is adjusted according to the current data state. In the specific implementation process, each input data is compared and evaluated in multiple dimensions through a high-performance computing platform to ensure the accuracy and real-time nature of the data determination. When it is determined to be "character tendency TV data", the next step of the multi-round intention effect simulation process is entered. The focused image of the moving object is obtained through a camera, and a deep learning model (such as a convolutional neural network CNN) is used to identify the moving object and determine the movement trajectory of the object. After that, according to the user's historical behavior and tendency data, the image of the moving object is simulated for multiple rounds based on the prediction model. During the simulation process, the user's tendency is combined to simulate the user's reaction under different intentions. The intention prediction model is reversely inferred to adjust the display effect of the image round by round, and finally a set of multi-round intention effect simulation data is generated to obtain the intention scene data. The dataset includes the user's focusing reaction and behavior performance on moving objects in different situations. The boundary of each group of data in the intention scene dataset is set, and the intention scene dataset is clustered by setting data analysis tools (such as the scikit-learn package in R language or Python). In the clustering process, the range of each group of intention scenes in the dataset is calculated, and the upper and lower limits of each group of data are clarified. The boundary analysis will determine the reasonable boundary data by fitting the historical behavior data of the user's input intention. In the implementation process, the intention scene is visualized and analyzed using computer graphics technology tools (such as OpenCV) to obtain reasonable boundary data. Based on the reasonable boundary data, the intention scene dataset is inverted for extreme scenarios. The intention scene data within the reasonable boundary is processed using the reverse simulation algorithm. Through the extreme value analysis, the user's reaction under different boundary conditions is simulated, and the extreme value of the data within the boundary is calculated. The extreme scene inversion method is used to generate the predicted user input intention. This process will use the Monte Carlo simulation method (Monte Carlo Simulation) performs multiple random sampling of extreme scenes to obtain a set of possible user input intention data, detects the ambient light intensity of the focused image of the moving object, and uses a high-precision light sensor (such as a spectrophotometer) to measure the ambient light intensity in the focused image. The sensor will record the ambient light intensity data in real time and transmit it to the processing unit. The processing unit normalizes the light intensity data through the image processing algorithm to ensure the matching degree of the ambient light intensity data with the image content. The light intensity measurement will be accurate to each image area, especially in dark or strong light environments. The light intensity detection algorithm can refine the changes in the environment to ensure the accuracy of the detection results, and calculate the visual comfort of the ambient light intensity data.According to the light intensity data, the comfort of each image area is evaluated using a visual comfort calculation model. The model is based on the human eye's response to different light intensities and is adjusted in combination with individual differences to calculate the visual pressure data of each area. The model includes an analysis of the light intensity response under long-term viewing. During the data processing process, the weighted average method is used to process the ambient light intensity data to obtain the final visual pressure data. The visual pressure data and the person's preference for TV data are combined, and weighted regression analysis (such as a linear regression model) is used to fit the predicted data to analyze the impact of visual pressure on the person's preference for TV. The weight parameters in the prediction model are gradually adjusted, and the actual demand prediction results are obtained through model training to further optimize the prediction model. The parameters in the model are used to obtain the predicted TV demand data of the persona tendency. The prediction process will be continuously iterated and adjusted through a large amount of historical data to ensure that the prediction results meet the actual needs. The predicted user input intention is gradually adjusted according to the predicted TV demand of the persona tendency. Based on the predicted TV demand data of the persona tendency, the predicted user input intention is gradually optimized using a dynamic adjustment algorithm (such as a recursive neural network RNN). The adjustment process continuously adjusts the parameters of the prediction model through feedback learning of historical behavior data, making the final output user input information more accurate. The final generated predicted user input information can accurately reflect the user's potential needs. The prediction results will be updated in a timely manner to adapt to changes in user behavior patterns.

[0144] Preferably, step S42 includes the following steps:

[0145] Extract scene features from the intention scene data set to obtain scene feature data; cluster and group the scene feature data to generate scene category data;

[0146] Perform distribution range statistics on the intention scenario data set according to the scenario category data to generate a distribution interval set;

[0147] Based on a preset interval threshold, an over-limit detection is performed on the distribution interval set to obtain an over-limit interval set; based on the over-limit interval set, a reasonable boundary screening is performed on the distribution interval set to generate reasonable boundary data;

[0148] Reconstruct the intention scene data set into a reasonable scene according to the reasonable boundary data to generate a reconstructed scene; calculate the difference of the reconstructed scene to obtain the difference data;

[0149] The difference data is analyzed for adjustment direction to generate adjustment direction data; the intention of the intention scene data set is reversely deduced based on the adjustment direction data to obtain the predicted user input intention.

[0150] In this embodiment, the acquired intention scene data is processed by an image processing algorithm, and features such as color, texture, and shape are extracted from each scene using a feature extraction algorithm (such as SIFT, HOG, etc.). During the extraction process, the OpenCV image processing library is used to perform edge detection, corner detection, and color distribution analysis to ensure that the features of each scene can fully reflect the environmental state and image content during user interaction. The feature extraction results will be converted into a high-dimensional data format to form scene feature data, and the scene feature data will be clustered and grouped. The K-means clustering algorithm is used to cluster the extracted scene feature data. The clustering algorithm calculates the distance between each data point and the cluster center, continuously iterates and optimizes, and finally clusters scenes with similar features. The scenes are grouped into one category. In the clustering process, the Euclidean distance metric is used to evaluate the similarity between each scene and the cluster center. By setting a reasonable number of clusters, scene category data is generated to ensure that the data of each category has a high similarity. Range statistics are performed on each type of scene data. In the statistical process, the Pandas library in Python is used to distribute the data. By calculating the maximum and minimum values ​​of the scene features in each category, the distribution interval of each category is obtained. After the statistics are completed, a distribution interval set is generated. The interval set contains the minimum value, maximum value and interval range of each category. The distribution interval set is detected based on the preset interval threshold. The preset threshold (for example, the distribution range exceeds A certain value or lower than a certain value), detect the data concentrated in the distribution interval, implement data traversal detection through Python, and when it is found that the value of a distribution interval exceeds the threshold range, mark the interval as out of limit. The over-limit detection algorithm compares each interval, screens out the data that meets the over-limit condition, generates an over-limit interval set, removes the data concentrated in the over-limit interval from the distribution interval set, and screens the remaining data through the adjustment method. The boundary screening algorithm is used to correct and update the distribution interval in combination with the clustering results. During the screening process, the data is visualized through graphic analysis tools (such as Matplotlib) to ensure that the screened data is within a reasonable boundary range, and finally generates reasonable boundary data. According to the reasonable boundary data According to the intention scene data set, reasonable scenes are reconstructed, and the reconstruction algorithm is used to process the reasonable boundary data. In the reconstruction process, missing or incomplete scene data are filled by interpolation methods (such as Lagrange interpolation) to ensure that all reconstructed scenes meet the reasonable boundary requirements. The reconstruction algorithm will generate reconstructed scenes that meet user needs based on the trends and changes in historical data. The reconstructed scenes are further analyzed using data visualization technology, and the differences of the reconstructed scenes are calculated. The difference algorithms (such as Euclidean distance and cosine similarity) are used to compare the reconstructed scenes with the original scenes. By calculating the differences between the scenes, the difference data is generated. The difference calculation process uses the NumPy library in Python to implement matrix operations.Ensure that the calculation results accurately reflect the magnitude and difference of the scene changes, and finally obtain the difference data reflecting the changes in the user's input intention. Analyze the change trends between different scenes through the difference data, use regression analysis (such as linear regression) to fit the data, and determine the adjustment direction of each scene change. Combine the scene category data in the analysis process to generate adjustment direction data. The adjustment direction data will indicate the priority and direction of the scene data adjustment. According to the adjustment direction data, reversely deduce the intention of the intention scene data set, and use the reverse derivation algorithm (such as the back propagation algorithm) to deduce the intention scene data. Through the known adjustment direction data, gradually reversely infer the user's potential input intention. In the derivation process, use a deep learning model (such as the LSTM long short-term memory network) to process the data to ensure that the derived predicted user input intention can accurately match the user's demand changes, and finally generate predicted user input intention data. ,

[0151] Preferably, step S44 includes the following steps:

[0152] Conduct correlation analysis on visual stress data and character preference TV data to obtain stress-demand correlation data;

[0153] Conduct corresponding impact analysis on pressure-demand correlation data to generate visual pressure impact data;

[0154] Based on the visual pressure impact data, the influence elimination simulation is performed on the character preference TV data to obtain the simulated demand tendency data; the actual demand is estimated on the simulated demand tendency data to generate the predicted character preference TV demand;

[0155] Perform difference comparison processing on the predicted TV demand of the person and the predicted user input intention to obtain demand difference data;

[0156] The predicted user input intention is gradually adjusted according to the demand difference data to generate predicted user input information.

[0157] In this embodiment, visual stress data are collected. These data are obtained by calculating the ambient light intensity and visual comfort. The data are combined with the character preference TV data. The data are obtained by analyzing the user's past viewing preferences and emotional reactions. Then, a multivariate regression analysis model is used to perform correlation analysis on the visual stress data and the character preference TV data. The model calculates the relationship between the variables by the least square method to obtain stress-demand correlation data. The data includes the degree and manner of the influence of visual stress on the user's TV preference. The weight values ​​of different visual stress data are set. The weight values ​​are set according to the intensity of the visual stress and the influence on the user's preference. Then, the product of each item of stress-demand correlation data and the corresponding weight is calculated, and all the data are summed to finally obtain the visual stress influence data. According to the obtained visual stress influence data, the character preference TV data is eliminated and simulated. The elimination simulation is performed by subtracting the visual stress factor with the strongest influence from the character preference TV data. A filtering algorithm (such as Kalman filter) is used in the simulation process. Wave) is used to process the data, and the influence of visual pressure is eliminated by gradually adjusting the character tendency TV data to obtain the simulated demand tendency data. The historical user demand data and the current simulated demand tendency data are combined, and the regression analysis model is used to fit the data to obtain the actual user demand. The regression model adopts linear regression or polynomial regression to ensure the accuracy of the estimated value. The model is trained by inputting sample data during the training process, and the results are verified to finally generate the predicted character tendency TV demand. The predicted character tendency TV demand data and the predicted user input intention data are collected, and the Euclidean distance algorithm is used to calculate the difference between the two sets of data. The Euclidean distance calculates the difference between the demand data and the input intention data to obtain the demand difference data. By analyzing the demand difference data, the adjustment direction and amplitude of the user input intention are determined. Then, an adaptive algorithm (such as a gradient descent algorithm) is used to gradually adjust the predicted user input intention to ensure that each step in the adjustment process can gradually narrow the demand difference.

[0158] Preferably, step S5 comprises the following steps:

[0159] Step S51: extracting logical relationships from the predicted user input information to obtain basic logical relationships; verifying constraint conditions on the basic logical relationships to generate constraint verification data;

[0160] Step S52: Conflict-marking the constraint verification data to obtain conflict-marking data; identifying the conflict type of the conflict-marking data to generate conflict detection data;

[0161] Step S53: tracing the conflict detection data to obtain the key factors of the conflict; performing logic adjustment simulation on the predicted user input information based on the key factors of the conflict to generate multiple simulation adjustment data;

[0162] Step S54: performing a second round of logic conflict verification on the plurality of simulation adjustment data to obtain second conflict detection data; performing a difference comparison on the conflict detection data according to the second conflict detection data to generate conflict change data;

[0163] Step S55: screening conflict factors for key conflict factors based on the conflict change data to generate optimized candidate conflict factors; performing feedback optimization processing on the predicted user input information according to the optimized candidate conflict factors to generate user input information.

[0164] In this embodiment, natural language processing (NLP) is performed on keywords in the user input information, and a dependency syntax analysis tool (such as the SpaCy library) is used to extract the main logical relationships in the information, such as time, conditions, and causal relationships. The extracted logical relationships include user needs, priorities, and associated information, and basic logical relationship data is obtained. These data represent the structured form of the user input information. Then, a rule engine (such as the Drools rule engine) is used to verify the basic logical relationships, check whether each logical relationship meets the conditions and constraints preset by the system, generate constraint verification data, preset a series of conditions, such as the user input must be within a specified time range, the demand must be consistent with the device status, etc. Then, the basic logical relationship data is verified. The rule engine verifies these constraints to ensure the feasibility of the logical relationship in the actual scenario. The generated constraint verification data indicates which logical relationships have passed the verification and which logical relationships have failed to meet the constraint conditions. This process is completed using a condition-based rule reasoning algorithm. Based on the failed items in the constraint verification data, the conflicting parts of the logical relationship are marked using a conflict detection algorithm (such as a decision tree algorithm based on ensemble learning). When marking, it is determined whether different logical conditions contradict each other or lead to unexecutable situations, and the conflict marking data is obtained. This data indicates the conflict points between logical relationships under specific conditions. Then, the conflict identification algorithm is used to process the conflict marking data. The algorithm will identify the type of conflict, such as time conflict, resource conflict, etc. Conflicts, demand conflicts, etc., and generate conflict detection data. Conflict type identification is classified based on the specific attributes of logical relationships. Conflict detection is implemented using fuzzy logic algorithms. Tracing algorithms (such as backtracking algorithms) are used to trace back the root causes of conflicts based on conflict detection data. During the backtracking process, the key factors that initially caused the conflict are found by tracing the sequence of various conflict points. The tracing process involves analyzing the mutual influence of various conditions and generating conflict key factor data through the reasoning process. These factors represent the root causes of logical conflicts. After obtaining the conflict key factors, logical adjustment simulation is performed on the predicted user input information based on these factors. During the simulation, a simulation algorithm (such as Monte Carlo simulation) is used to adjust the user input information through multiple simulations. , generate multiple simulated adjustment data, use the same conflict detection algorithm to perform conflict verification on the simulated adjusted data, check whether there is still a logical conflict after the simulated adjustment, and obtain the second conflict detection data. These data are compared with the first conflict detection data to record the new conflict situation. Then, compare the gap between the second conflict detection data and the first conflict detection data, use the difference comparison analysis method (such as Pearson correlation coefficient) for quantitative evaluation, generate conflict change data, and determine which conflict factors have changed greatly based on the gap analysis in the conflict change data. These factors are considered to be the most critical conflict factors. Then, use the screening algorithm (such as K-means clustering algorithm) to screen the key conflict factors.Conflict factors with large changes and significant impact on user input are extracted to generate optimized candidate conflict factors. Next, feedback optimization processing is performed on the predicted user input information based on the optimized candidate conflict factors. Feedback adjustment algorithms (such as feedback control algorithms) are used to gradually optimize the user input information. Through gradual adjustment, the unreasonable parts of the information are removed to ensure that the final generated user input information better matches the actual needs of the system. During the optimization process, minimization algorithms (such as gradient descent) are used to adjust the input information.

Claims

1. A method for user inputting information in a smart TV all-in-one machine, characterized in that: The following steps are involved: Step S1: capturing a current user image; performing white balance correction on the current user image to obtain a color-corrected image; Perform dynamic image focusing on the color-corrected image to generate a focused image of the moving object; Step S2: performing person recognition on the focused image of the moving object to obtain a person image; performing attention tendency analysis on the person image to generate person tendency data; Step S3: Acquire historical figure interaction images; perform interaction pattern recognition on the historical figure interaction images to obtain historical interaction pattern data; perform user input intention prediction modeling on the historical interaction pattern data to generate a user input intention prediction model; Step S4: when the character tendency data is determined to be character tendency television data, a user input intention drill is performed on the moving object focus image based on the user intention prediction model to generate a predicted user input intention; According to the character tendency data and the moving object focus image, auxiliary adjustment is made to the predicted user input intention to generate predicted user input information; wherein step S4 includes the following steps: Step S41: performing a threshold determination on the character tendency data. When the character tendency data is determined to be character tendency TV data, multiple rounds of intention effect simulation are performed on the moving object focus image based on the user intention prediction model to obtain an intention scene data set. Step S42: performing reasonable boundary analysis on the intention scenario data set to obtain reasonable boundary data; performing extreme scenario inversion on the intention scenario data set based on the reasonable boundary data to generate a predicted user input intention; Step S43: performing ambient light intensity detection on the focused image of the moving object to obtain ambient light intensity data; performing visual comfort calculation on the ambient light intensity data to generate visual pressure data; Step S44: predicting the actual demand of the character's inclination TV data based on the visual pressure data to generate predicted character's inclination TV demand; and progressively adjusting the predicted user input intention according to the predicted character's inclination TV demand to generate predicted user input information; Step S5: performing logical conflict detection on the predicted user input information to obtain conflict detection data; performing feedback optimization processing on the predicted user input information based on the conflict detection data to generate user input information.

2. The method for user inputting information in a smart TV all-in-one according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: capturing the current user image; performing color temperature analysis on the current user image to obtain color temperature information of the user image; performing white balance correction calculation on the color temperature information of the user image to generate white balance correction parameters; Step S12: performing color correction on the current user image based on the white balance correction parameters to obtain a color-corrected image; Step S13: performing motion region detection on the color-corrected image to obtain a motion region image; performing motion trajectory tracking on the motion region image to obtain a motion region trajectory; Step S14: dynamically focusing the moving area image according to the moving area trajectory to generate a moving object focused image.

3. The method for user inputting information in a smart TV all-in-one according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing human outline recognition on the moving object focused image to obtain the moving human outline; performing human body region cutting on the moving object focused image based on the moving human outline to generate a human image; Step S22: extract facial feature points from the person image to obtain facial feature data; accurately locate the pupils of the facial feature data to generate eye position data; Step S23: performing human gaze simulation on the eye position data to generate simulated annotation data; performing gaze angle calculation on the simulated gaze data to obtain a gaze angle; Step S24: performing time-series tracking on the gaze angle to obtain gaze trajectory data; performing attention tendency analysis based on the gaze trajectory data to generate character tendency data.

4. The method for user inputting information in a smart TV all-in-one according to claim 3, characterized in that: Step S24 includes the following steps: Performing a timestamp on the gaze angle to obtain time-series marked viewing angle data; performing a jump point detection on the time-series marked viewing angle data to generate viewing angle jump point data; The dwell time of the view jump point data is counted to obtain the view retention data; the visual trajectory is reconstructed for the view retention data to generate the gaze trajectory data; Performing regional clustering processing on the gaze trajectory data to obtain clustered regional data; performing frequency statistics on the clustered regional data to generate regional frequency data; Perform attention pattern recognition on the regional frequency data to generate an attention pattern; perform target region matching on the attention pattern to obtain region matching data; Performing continuity evaluation on the regional matching data to obtain continuity data; performing change trend analysis on the continuity data to generate continuity trend features; The tendency of the continuity trend characteristics is judged to generate character tendency data.

5. The method for user inputting information in a smart TV all-in-one according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: Acquire interactive images of historical figures; sort the interactive images of historical figures in time sequence to obtain time series data; Step S32: extracting character behaviors from the time series data to generate a behavior feature sequence; performing pattern clustering analysis on the behavior feature sequence to obtain historical interaction pattern data; Step S33: performing feature vectorization processing on the historical interaction pattern data to obtain a pattern feature vector; performing prediction model parameter training on the pattern feature vector to generate prediction model parameters; Step S34: performing prediction model integration on the prediction model parameters to generate a user input intention prediction model.

6. The method for user inputting information in a smart TV all-in-one according to claim 1, characterized in that: Step S42 includes the following steps: Extract scene features from the intention scene data set to obtain scene feature data; cluster and group the scene feature data to generate scene category data; Perform distribution range statistics on the intention scenario data set according to the scenario category data to generate a distribution interval set; Based on a preset interval threshold, an over-limit detection is performed on the distribution interval set to obtain an over-limit interval set; based on the over-limit interval set, a reasonable boundary screening is performed on the distribution interval set to generate reasonable boundary data; Reconstruct the intention scene data set into a reasonable scene according to the reasonable boundary data to generate a reconstructed scene; calculate the difference of the reconstructed scene to obtain the difference data; The difference data is analyzed for adjustment direction to generate adjustment direction data; the intention scene data set is reversely deduced based on the adjustment direction data to obtain the predicted user input intention.

7. The method for user inputting information in a smart TV all-in-one according to claim 1, characterized in that: Step S44 includes the following steps: Conduct correlation analysis on visual stress data and character preference TV data to obtain stress-demand correlation data; Conduct corresponding impact analysis on pressure-demand correlation data to generate visual pressure impact data; Based on the visual pressure impact data, the influence elimination simulation is performed on the character preference TV data to obtain the simulated demand tendency data; the actual demand is estimated on the simulated demand tendency data to generate the predicted character preference TV demand; Perform difference comparison processing on the predicted TV demand of the person and the predicted user input intention to obtain demand difference data; The predicted user input intention is gradually adjusted according to the demand difference data to generate predicted user input information.

8. The method for user inputting information in a smart TV all-in-one according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: extracting logical relationships from the predicted user input information to obtain basic logical relationships; verifying constraint conditions on the basic logical relationships to generate constraint verification data; Step S52: Conflict-marking the constraint verification data to obtain conflict-marking data; identifying the conflict type of the conflict-marking data to generate conflict detection data; Step S53: tracing the conflict detection data to obtain the key factors of the conflict; performing logic adjustment simulation on the predicted user input information based on the key factors of the conflict to generate multiple simulation adjustment data; Step S54: performing a second round of logic conflict verification on the plurality of simulation adjustment data to obtain second conflict detection data; performing a difference comparison on the conflict detection data according to the second conflict detection data to generate conflict change data; Step S55: screening conflict factors for key conflict factors based on the conflict change data to generate optimized candidate conflict factors; performing feedback optimization processing on the predicted user input information according to the optimized candidate conflict factors to generate user input information.

9. A user input information system in a smart TV all-in-one machine, characterized in that: Used to execute the user input information method in the smart TV all-in-one machine according to claim 1, the user input information system in the smart TV all-in-one machine comprises: The image acquisition module is used to acquire the current user image; perform white balance correction on the current user image to obtain a color-corrected image; perform dynamic image focusing on the color-corrected image to generate a moving object focused image; The attention analysis module is used to perform person recognition on the focused image of the moving object to obtain a person image; perform attention tendency analysis on the person image to generate person tendency data; A prediction modeling module is used to obtain interactive images of historical figures; perform interaction pattern recognition on the interactive images of historical figures to obtain historical interaction pattern data; perform user input intention prediction modeling on the historical interaction pattern data to generate a user input intention prediction model; An information prediction module is used for, when the character tendency data is determined to be character tendency television data, performing a user input intention drill on the moving object focused image based on the user intention prediction model to generate a predicted user input intention; performing auxiliary adjustment on the predicted user input intention according to the character tendency data and the moving object focused image to generate predicted user input information; The closed-loop optimization module is used to perform logical conflict detection on the predicted user input information to obtain conflict detection data; and perform feedback optimization processing on the predicted user input information based on the conflict detection data to generate user input information.

Images