A data classification method and system based on prior information

Abstract

The present invention discloses a data classification method and system based on prior information. The method belongs to the field of electronic digital processing, and includes the following steps: collecting prior information of children's sitting posture; using sensors to capture the real-time dynamic features of children's sitting posture; performing data preprocessing on the prior information of children's sitting posture and the real-time dynamic features; performing label matching on the preprocessed prior information of children's sitting posture; constructing a classifier based on the preprocessed prior information of children's sitting posture and the sitting posture category label; classifying the preprocessed real-time dynamic features through the classifier, and obtaining the sitting posture category label corresponding to the real-time dynamic features of children's sitting posture as the sitting posture classification result of the real-time dynamic features. The present invention achieves multi-angle analysis of children's sitting posture status by acquiring static pictures, 3D scanning data, and real-time dynamic features in different scenes, and solves the problem of incomplete classification results of children's sitting posture in the prior art.

Description

Technical Field

[0001] The present invention relates to the field of electronic digital processing, and in particular to a data classification method and system based on prior information. Background Art

[0002] The rapid development of machine learning and sensor technology has provided new possibilities for the accurate classification of children's sitting posture data. However, the existing sitting posture classification methods ignore the stress of children's spine, resulting in incomplete results of children's sitting posture classification.

[0003] The existing sitting posture classification system requires children to wear additional wearable sensors, but it will affect comfort and freedom. In addition, under complex environmental conditions, such as lighting changes, occlusion, multi-person interaction, etc., the accuracy of the traditional sitting posture classification system obtained through image processing will be greatly affected.

[0004] For example, the invention patent with announcement number: CN115495609B announces a sitting posture data acquisition system, sitting posture data acquisition method and device, including: the sitting posture data acquisition system includes a control device, and a collection device with a camera that can communicate with the control device; the control device is used to display the collection scene and sitting posture selection interface of the collection device, so that the user can perform selection operations in the collection scene and sitting posture selection interface, and send data collection instructions to the collection device, wherein the collection scene and sitting posture selection interface provide a variety of optional collection scenes and a variety of optional sitting postures; the collection device is used to respond to the data collection instruction, collect the sitting posture data of the collected object, and name the collected sitting posture data according to the collection scene and sitting posture selected in the collection scene and sitting posture selection interface.

[0005] For example, the invention patent with announcement number: CN106445135B discloses a data interaction method and system for posture correction by an intelligent robot, including a visual perception step, in which the current image of the user is captured by the robot vision system to obtain image data classified according to user group attributes, posture attributes and scene attributes; a decision step, in which the image data associated with the user group attributes, posture attributes and scene attributes are combined for analysis to determine the user's posture health and decide whether to remind the user to adjust his posture; and a multimodal output step, in which multimodal output is performed according to the result of the decision.

[0006] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:

[0007] In the prior art, the sitting posture classification system manually determines whether the child's sitting posture is correct by manually observing and describing different characteristics of the sitting posture. However, the manual determination results in incomplete classification of the child's sitting posture. Summary of the invention

[0008] The embodiments of the present application provide a data classification method and system based on prior information, thereby solving the problem of incomplete results of children's sitting posture classification in the prior art and realizing multi-angle analysis of children's sitting posture status.

[0009] An embodiment of the present application provides a data classification method based on prior information, comprising the following steps: collecting prior information on children's sitting posture; capturing real-time dynamic features of children's sitting posture using sensors; performing data preprocessing on the prior information on children's sitting posture and the real-time dynamic features to obtain preprocessed prior information on children's sitting posture and preprocessed real-time dynamic features; performing label matching on the preprocessed prior information on children's sitting posture to obtain a sitting posture category label; constructing a classifier based on the preprocessed prior information on children's sitting posture and the sitting posture category label; classifying the preprocessed real-time dynamic features using the classifier to obtain a sitting posture category label corresponding to the real-time dynamic features of children's sitting posture as a sitting posture classification result of the real-time dynamic features.

[0010] Furthermore, the step of collecting prior information on children's sitting postures includes: collecting static pictures, videos and 3D scanning data of correct sitting postures and bad sitting postures of children of different ages, heights and weights;

[0011] Collect information about children’s sitting posture in different scenarios.

[0012] Furthermore, the step of using sensors to capture the real-time dynamic features of the child's sitting posture includes: the real-time dynamic features of the child's sitting posture include the body's tilt angle, the degree of spinal curvature deformation, and the pressure distribution of each point on the buttocks. The body's tilt angle is calculated using the angular velocity and acceleration measured by the gyroscope and accelerometer; the degree of spinal curvature deformation is measured using a flexible sensor; and the pressure distribution of each point on the buttocks when the child sits on the cushion is measured using a pressure sensor array.

[0013] Furthermore, the data classification method based on prior information is characterized in that: the data preprocessing step includes: cleaning the children's sitting posture prior information to remove duplicate data, incomplete data and normalize it; performing image enhancement, cropping, and scaling operations on the pictures and videos in the children's sitting posture prior information to highlight and extract specified features; converting the text description of the children's sitting posture prior information into numerical data; aligning the real-time dynamic features with the timestamp, and using filters to remove high-frequency noise in the sensor data; and standardizing data of different scales and units to obtain preprocessed children's sitting posture prior information and preprocessed real-time dynamic features.

[0014] Furthermore, the specific steps of the label matching are: obtaining all types of sitting posture category labels;

[0015] The angle between the head and the shoulder, the twisting amplitude of the body, and the twisting correction factor are obtained and input into the body tilt angle calculation formula; the body tilt angle is calculated according to the body tilt angle calculation formula; the elastic modulus of the spine, the cross-sectional inertia moment of the spine, the length of the spine, the actual bending angle of the spine, the physiological curvature of the spine, and the external force on the spine are obtained and input into the spinal bending deformation degree calculation formula; a body tilt angle label is added according to the body tilt angle, wherein the body tilt angle label is divided into a mild tilt label, a moderate tilt label, a severe tilt label, and a non-tilt label; the spinal bending deformation degree is calculated according to the spinal bending deformation degree calculation formula; a spinal bending deformation degree label is added according to the spinal bending deformation degree, wherein the spinal bending deformation degree label includes a normal bending label, a mild deformation label, and a severe deformation label; the pressure distribution of each point on the buttocks is obtained according to the 3D scanning data; the pressure distribution label of each point on the buttocks is added according to the pressure distribution of each point on the buttocks, wherein the pressure distribution label of each point on the buttocks includes a uniform distribution label, a local concentration label, and a severe concentration label.

[0016] Furthermore, the body inclination angle calculation formula is:

[0017]

[0018] In the formula, θ represents the body inclination angle, γ represents the angle between the head and shoulders, and k g represents the gravity correction factor, k β represents the torsion correction factor, and π represents the circumference of a circle.

[0019] Furthermore, the calculation formula for the degree of spinal curvature deformation is:

[0020]

[0021] D S It indicates the degree of spinal curvature deformation, which is used to judge whether the child's sitting posture is correct. E indicates the elastic modulus of the spine, which is used to reflect the stiffness of the spine. I indicates the section inertia moment of the spine, which reflects the influence of the shape of the spine cross section on the curvature. L indicates the length of the spine, and θ s( x ) represents the actual bending angle of the spine at position x, θ ref (x) represents the physiological curvature of the spine at position x, x represents the position coordinate of any point along the length of the spine, and is used to represent any position between the starting point and the end point of the spine. ( x ) represents the external force acting on the spine at position x, and dx represents a very small increment along the length of the spine used to calculate the integral.

[0022] Furthermore, the step of constructing a classifier based on the preprocessed child sitting posture prior information and sitting posture category labels includes: inputting the preprocessed child sitting posture prior information into the classifier; performing label matching with predefined sitting posture category labels based on features in the child sitting posture prior information; and integrating the matched labels and corresponding prior information into a complete data set to obtain a classifier.

[0023] Furthermore, the analysis step of the sitting posture category label corresponding to the real-time dynamic feature of the child's sitting posture is: the preprocessed real-time dynamic feature of the child's sitting posture is used as input data and input into the classifier; the classifier outputs the corresponding sitting posture category label according to the input data.

[0024] Furthermore, the information collection module is used to collect children's sitting posture prior information; the sensor module is used to capture the real-time dynamic features of children's sitting posture using sensors; the preprocessing module is used to perform data preprocessing on the children's sitting posture prior information and the real-time dynamic features to obtain preprocessed children's sitting posture prior information and preprocessed real-time dynamic features; the label matching module is used to perform label matching on the preprocessed children's sitting posture prior information to obtain a sitting posture category label; the classifier module is used to construct a classifier based on the preprocessed children's sitting posture prior information and the sitting posture category label; the classification module is used to classify the preprocessed real-time dynamic features through the classifier to obtain the sitting posture category label corresponding to the children's sitting posture real-time dynamic features.

[0025] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0026] 1. By collecting static pictures and 3D scanning data of correct and bad sitting postures of children of different ages, heights and weights, all-round sitting posture management and multi-angle analysis are achieved, and then multi-angle analysis of children's sitting posture is achieved, effectively solving the problem of incomplete results of children's sitting posture classification in the existing technology.

[0027] 2. The sensor captures and calculates the real-time dynamic characteristics of the child's body inclination angle, the degree of spinal curvature deformation and the pressure distribution at each point on the buttocks, thereby accurately calculating the degree of spinal curvature deformation, and thus realizing real-time monitoring of the health status of the spine.

[0028] 3. By cleaning, normalizing, and extracting key features of children’s sitting posture prior information and real-time dynamic features, accurate classification of various sitting postures is achieved, and the classifier can better identify and distinguish different types of sitting postures and provide accurate classification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1A flow chart of a data classification method based on prior information is provided for an embodiment of the present application;

[0030] Figure 2 A data diagram of the degree of spinal curvature deformation provided in the embodiment of the present application;

[0031] Figure 3 A structural diagram of a data classification system based on prior information provided in an embodiment of the present application. DETAILED DESCRIPTION

[0032] The embodiments of the present application solve the problem of incomplete results of children's sitting posture classification in the prior art by providing a data classification method and system based on prior information. By utilizing flexible sensors, gyroscopes, accelerometers and pressure sensor arrays to obtain static images, 3D scanning data, and real-time dynamic features in different scenes, multi-angle analysis of children's sitting posture is achieved.

[0033] The technical solution in the embodiments of the present application is to solve the problem of incomplete results of the above-mentioned classification of children's sitting postures, and the overall idea is as follows: by collecting children's sitting posture prior information; using sensors to capture the real-time dynamic features of children's sitting postures; performing data preprocessing on the children's sitting posture prior information and the real-time dynamic features to obtain preprocessed children's sitting posture prior information and preprocessed real-time dynamic features; performing label matching on the preprocessed children's sitting posture prior information to obtain sitting posture category labels; constructing a classifier based on the preprocessed children's sitting posture prior information and the sitting posture category labels; classifying the preprocessed real-time dynamic features through the classifier to obtain the sitting posture category labels corresponding to the real-time dynamic features of the children's sitting posture, which are used as the sitting posture classification results of the real-time dynamic features, thereby achieving multi-angle analysis of the children's sitting posture state.

[0034] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0035] like Figure 1 As shown, it is a flow chart of a data classification method based on prior information provided by an embodiment of the present application, the method is applied to a data classification system based on prior information, the method comprising the following steps: collecting prior information of children's sitting posture; capturing real-time dynamic features of children's sitting posture by using sensors; performing data preprocessing on the prior information and real-time dynamic features of children's sitting posture to obtain preprocessed prior information and preprocessed real-time dynamic features of children's sitting posture; performing label matching on the preprocessed prior information of children's sitting posture to obtain sitting posture category labels; constructing a classifier based on the preprocessed prior information of children's sitting posture and the sitting posture category labels; classifying the preprocessed real-time dynamic features by the classifier to obtain sitting posture category labels corresponding to the real-time dynamic features of children's sitting posture as sitting posture classification results of the real-time dynamic features.

[0036] In this embodiment, the prior information related to the child's sitting posture includes the child's age, gender, height, weight, bone development status, behavioral habits, and previous medical history. The purpose of preprocessing the collected prior information is to eliminate noise and outliers in the data, improve the reliability and consistency of the data, and construct a feature set related to the child's sitting posture. The feature set is a key feature set used to describe and analyze the child's sitting posture, helping the classifier to accurately identify and classify different types of sitting postures. The feature set includes body shape-based features such as height and body mass index, bone development-based features such as spinal curvature and pelvic inclination, and behavioral habit-based features such as sitting duration, sitting frequency, and activity level. The sensor data is used to capture the real-time dynamic features of the child's sitting posture, including the child's sitting angle and posture stability. It is conducive to systematic and automated sitting posture monitoring and reduces manual intervention. Improve classification accuracy and ensure scientific management of children's sitting posture. Through this method, children's bad sitting posture can be effectively identified and corrected, and spinal diseases and other related health problems can be prevented.

[0037] Furthermore, the step of collecting prior information on children's sitting posture includes: collecting static images, videos and 3D scanning data of correct sitting posture and bad sitting posture of children of different ages, heights and weights; collecting children's sitting posture information in different scenarios.

[0038] In this embodiment, the sitting posture information of children in different scenarios is collected, and the different scenarios include: school classrooms, home desks, public transportation (such as school buses), and outdoor places (such as parks). The sitting posture information includes: the posture angle of the sitting posture, the relative positions of various parts of the body, and the alignment of the head, shoulders, and spine. It is conducive to providing diversified scene data and ensuring the generalization ability of the sitting posture classification model. It covers a variety of situations in children’s daily life to ensure the comprehensiveness of sitting posture monitoring. Through the sitting posture data in different scenarios, we can fully understand the sitting posture of children in various environments, and help formulate more scientific and effective sitting posture correction measures.

[0039] Furthermore, the step of using sensors to capture the real-time dynamic features of the child's sitting posture includes: the real-time dynamic features of the child's sitting posture include the body's tilt angle, the degree of spinal curvature deformation, and the pressure distribution of each point on the buttocks. The body's tilt angle is calculated using the angular velocity and acceleration measured by the gyroscope and accelerometer; the degree of spinal curvature deformation is measured using a flexible sensor; and the pressure distribution of each point on the buttocks when the child sits on the cushion is measured using a pressure sensor array.

[0040] In this embodiment, by real-time dynamic monitoring of children's sitting posture, bad sitting posture can be discovered and corrected in time. Accurately measure data in multiple dimensions to improve the accuracy of sitting posture analysis. By capturing real-time dynamic features, children's sitting posture can be monitored at any time to prevent and reduce health problems caused by bad sitting posture.

[0041] Furthermore, the steps of data preprocessing include: cleaning the children's sitting posture prior information to remove duplicate data, incomplete data and normalize it; performing image enhancement, cropping, and scaling operations on the pictures and videos in the children's sitting posture prior information to highlight and extract specified features; converting the text description of the children's sitting posture prior information into numerical data; aligning the real-time dynamic features with the timestamp, and using filters to remove high-frequency noise in the sensor data; and standardizing data of different scales and units to obtain the preprocessed children's sitting posture prior information and the preprocessed real-time dynamic features.

[0042] In this embodiment, the specified features include: the position of key points of the body (such as the head, shoulders, waist, and hips), the sitting angle and spinal curvature, and the relative position of the head and shoulders. The role of preprocessing is to make the preprocessed data more standardized and consistent, and to improve data quality. After extracting the specified features, the data is more representative and has more analytical value. Through data preprocessing and feature extraction, the accuracy and stability of subsequent classifier training and real-time monitoring are guaranteed. The specific method for converting the text description of the prior information of the child's sitting posture into numerical data is: number the text description in the prior information starting from 0.

[0043] Furthermore, the specific steps of label matching are as follows: obtaining all types of sitting posture category labels; obtaining the angle between the head and shoulders, the twisting amplitude of the body, and the twisting correction factor and inputting them into the body tilt angle calculation formula; calculating the body tilt angle according to the body tilt angle calculation formula; obtaining the elastic modulus of the spine, the cross-sectional inertia moment of the spine, the length of the spine, the actual bending angle of the spine, the physiological curvature of the spine, and the external force on the spine and inputting them into the spinal bending deformation degree calculation formula; adding a body tilt angle label according to the body tilt angle, wherein the body tilt angle label is divided into a mild tilt label, a moderate tilt label, a severe tilt label, and a non-tilt label; calculating the spinal bending deformation degree according to the spinal bending deformation degree calculation formula; adding a spinal bending deformation degree label according to the spinal bending deformation degree, wherein the spinal bending deformation degree label includes a normal bending label, a mild deformation label, and a severe deformation label; obtaining the pressure distribution of each point on the buttocks according to the 3D scanning data; adding a pressure distribution label of each point on the buttocks according to the pressure distribution of each point on the buttocks, wherein the pressure distribution label of each point on the buttocks includes a uniform distribution label, a local concentration label, and a severe concentration label.

[0044] In this embodiment, by obtaining comprehensive sitting posture category labels, the system can more accurately classify different types of sitting postures. Specifically, the comprehensive acquisition of sitting posture category labels ensures that the system can cover various types of sitting postures during training, and improves the generalization ability of the classifier. By calculating the body inclination angle and the degree of spinal curvature deformation and adding corresponding labels, the system can analyze the child's sitting posture in detail and accurately identify mild, moderate, severe and normal inclination and deformation. In addition, adding labels according to the pressure distribution of each point on the buttocks can effectively monitor the pressure distribution of the child's sitting posture, thereby helping to identify unevenly distributed, locally concentrated and severely concentrated pressure conditions, which is crucial for preventing injuries to the spine and other parts of the body. Through refined label matching, the system can more accurately identify and classify different types of sitting postures. This is of great significance for subsequent classifier training and real-time monitoring. Accurate label matching not only improves the accuracy of classification, but also provides a reliable basis for subsequent correction. For the monitoring and correction of children's sitting posture, being able to timely identify and correct bad sitting postures is the key to preventing long-term health problems. Through label matching, the system can provide timely feedback on sitting posture, help children maintain correct sitting posture, prevent scoliosis, hunchback and other problems, and thus improve children's quality of life and learning efficiency. In addition, by refining the sitting posture labels, the system can provide personalized sitting posture correction suggestions to help children develop good sitting habits. Refined label matching is conducive to improving classification accuracy. By comprehensively considering a variety of sitting posture characteristics, the comprehensiveness of the classification is ensured. Through precise label matching, different sitting posture types can be more accurately identified and classified, providing a basis for subsequent correction.

[0045] Furthermore, the body tilt angle calculation formula is:

[0046]

[0047] In the formula, θ represents the body inclination angle, γ represents the angle between the head and shoulders, and k g represents the gravity correction factor, k β represents the torsion correction factor, and π represents the circumference of a circle.

[0048] In this embodiment, k β =1-c*|h|; c represents the constant that controls the twisting effect, which can be obtained by consulting expert experimental data, and h represents the twisting amplitude of the body. For the conversion between angle and radian, the body tilt angle θ can be used to judge whether the child's sitting posture is correct. The body tilt angle θ can also be obtained by computer vision technology; k g It can be obtained through Newton's first law and can also be obtained through experimental calculation; k βIt can be obtained by testing in the actual system or by experimental simulation; γ can be obtained by sensor measurement or by image recognition through machine vision technology; h can be obtained by ruler measurement or by background difference method; the body tilt angle can be accurately quantified by formula calculation. It provides accurate quantitative data for sitting posture classification and analysis, and enhances the scientific nature of the analysis. The body tilt angle formula is obtained by analyzing and deforming the atan2(y,x) formula. atan2 ( The function of the y,x) function is to return the arc tangent of the x-coordinate and y-coordinate of the specified point.

[0049] Furthermore, the calculation formula for the degree of spinal curvature deformation is:

[0050]

[0051] D S It indicates the degree of spinal curvature deformation, which is used to judge whether the child's sitting posture is correct. E indicates the elastic modulus of the spine, which is used to reflect the stiffness of the spine. I indicates the section inertia moment of the spine, which reflects the influence of the shape of the spine cross section on the curvature. L indicates the length of the spine, and θ s( x ) represents the actual bending angle of the spine at position x, θ ref( x ) represents the physiological curvature of the spine at position x, where x represents the position coordinate of any point along the length of the spine, and is used to represent any position between the starting point and the end point of the spine. ( x) represents the external force acting on the spine at position x, and dx represents a very small increment along the length of the spine used to calculate the integral.

[0052] In this embodiment, the starting point of the spine is the child's neck, and the end point is the child's coccyx. S The bending angle and deformation degree of the spine can also be estimated by using X-rays to obtain spinal images; E can be obtained by consulting relevant biomechanical and medical literature to obtain the elastic modulus values ​​of the spine under different conditions, or by experimentally measuring the stress-strain relationship curve of the spine under different stress conditions and calculating the elastic modulus; I can be obtained by using 3D scanning data of the spine to establish a 3D model of the spine and calculate the section inertia moment at different positions; L can be obtained by directly measuring the length of the child's spine using a measuring tool, or by using 3D scanning technology to obtain spine data and calculate the total length of the spine; θ s The actual bending angles at different positions of the spine can be obtained by installing flexible sensors at different positions and measuring the actual bending angles at different positions through the deformation of the flexible sensors. The actual bending angles at different positions of the spine can also be determined by scanning the child with a 3D scanner and analyzing the 3D model.ref It can be obtained by consulting medical information, or by using the normal spinal curvature angle in clinical data; x can be obtained by dividing the spine into several segments, such as measuring every certain distance (such as every 1 cm), or by using high-resolution sensors (such as 3D scanners or high-density pressure sensor arrays) to continuously record actual data at various positions on the spine; F(x) can be obtained by installing a pressure sensor array on the child's seat, measuring the pressure distribution on the seat when the child is sitting, and calculating the external force at different positions. It can also be obtained by using a dynamic pressure sensor to directly measure the pressure at different positions of the spine when the child is sitting. The degree of curvature deformation of the spine can be accurately quantified through formula calculation. Provide accurate quantitative data for sitting posture classification and analysis, enhance the scientific nature of the analysis, and improve the standardization and repeatability of data processing. As shown in Table 1, the degree of spinal curvature deformation data table:

[0053] Table 1 Data table of spinal curvature deformation degree

[0054]

[0055]

[0056] like Figure 2 The data table of spinal curvature deformation degree is composed of Table 1 and Figure 2 It can be analyzed that the elastic modulus and the section inertia moment I jointly determine the degree of bending deformation. The length of the spine L affects the total accumulation of bending deformation. The longer the length, the greater the impact of deformation. The difference between the actual bending angle and the physiological curvature indicates the actual deviation of the spine. The greater the difference, the greater the degree of deformation. External force F ( x) The external force acting on the spine directly affects the deformation of the spine. The greater the external force, the greater the deformation. Analysis shows that by adjusting the elastic modulus and section moment of inertia of the spine, as well as controlling the external force, the degree of curvature deformation of the spine can be effectively managed and reduced.

[0057] Furthermore, the step of constructing a classifier based on the preprocessed child sitting posture prior information and sitting posture category labels includes: inputting the preprocessed child sitting posture prior information into the classifier; performing label matching with predefined sitting posture category labels based on features in the child sitting posture prior information; and integrating the matched labels with the corresponding prior information into a complete data set to obtain a classifier.

[0058] In this embodiment, data is input into the classifier training framework: by inputting the preprocessed children's sitting posture prior information into the classifier training framework, the classifier can be systematically trained so that it can learn and identify different sitting posture types from a large amount of data. This systematic training can improve the learning ability and accuracy of the classifier. Data integration: integrating the matched labels and the corresponding prior information into a complete data set can provide comprehensive data support for the classifier. It helps to improve the generalization ability of the classifier so that it can show good classification effects when facing different sitting posture types. Through the above steps, the system can build an efficient and accurate classifier to provide a basis for subsequent real-time classification. Specifically: Improve the accuracy of the classifier: through systematic training and accurate label matching, the classifier can accurately identify and classify different sitting posture types. This helps to improve the overall classification effect and reliability of the system. Enhance the learning ability of the system: through the training of a large amount of data, the classifier can continuously learn and improve, so that it can also make accurate classifications when facing new sitting posture types. Provide scientific basis: through the analysis and classification results of the classifier, a reliable basis can be provided for the scientific management and correction of children's sitting posture. This helps prevent and correct children's bad sitting postures and promote their healthy growth. Improve the automation of the system: By building an efficient classifier, the system can achieve automatic classification and reduce manual intervention and operation. This not only improves the work efficiency of the system, but also ensures the objectivity and consistency of the classification results. Systematically training the classifier ensures the efficiency and accuracy of the classifier. By integrating multiple prior information, the generalization ability of the classifier is improved. Provide an accurate and reliable classifier that can effectively identify and classify children's various sitting postures, providing technical support for subsequent real-time monitoring and correction.

[0059] Furthermore, the analysis steps of the sitting posture category label corresponding to the real-time dynamic features of the child's sitting posture are as follows: the preprocessed real-time dynamic features of the child's sitting posture are input into the classifier as input data; the classifier outputs the corresponding sitting posture category label according to the input data.

[0060] In this embodiment, real-time data input: the pre-processed real-time dynamic features are used as input data and input into the trained classifier model, which can ensure that the system can obtain and process the children's sitting posture data in real time. This real-time performance can timely discover and correct bad sitting postures and prevent long-term health problems. Classification calculation: the classifier calculates the most matching sitting posture category label based on the input data, and can quickly and accurately identify the current sitting posture type. Accurate classification judgments can be made in a short time to ensure the timeliness and accuracy of feedback. Result output: the classifier outputs the corresponding sitting posture category label, which is the sitting posture classification result of the real-time dynamic feature, and can provide instant feedback for the children's sitting posture management. This instant feedback can help children and guardians understand and adjust their sitting postures in a timely manner and promote the formation of good sitting habits. Real-time dynamic classification can timely identify and feedback the children's sitting posture. Ensure the accuracy and real-time performance of the classification results. Through real-time classification and feedback, children's bad sitting postures can be corrected in time, long-term health problems can be prevented, and the quality of life and learning efficiency of children can be improved.

[0061] like Figure 3 As shown, it is a structural schematic diagram of a data classification system based on prior information provided in an embodiment of the present application. A data classification system based on prior information provided in an embodiment of the present application includes: an information collection module, a sensor module, a preprocessing module, a label matching module, a classifier module, and a classification module; wherein the information collection module is used to collect prior information on children's sitting posture; the sensor module is used to capture the real-time dynamic features of children's sitting posture using sensors; the preprocessing module is used to perform data preprocessing on the prior information on children's sitting posture and the real-time dynamic features to obtain the preprocessed prior information on children's sitting posture and the preprocessed real-time dynamic features; the label matching module is used to perform label matching on the preprocessed prior information on children's sitting posture to obtain a sitting posture category label; the classifier module is used to construct a classifier based on the preprocessed prior information on children's sitting posture and the sitting posture category label; the classification module is used to classify the preprocessed real-time dynamic features through the classifier to obtain the sitting posture category label corresponding to the real-time dynamic features of the children's sitting posture.

[0062] In summary, this embodiment captures the real-time dynamic characteristics of the child's body inclination angle, the degree of spinal curvature deformation and the pressure distribution at each point on the buttocks by integrating multiple data sources and technical means, thereby accurately calculating the degree of spinal curvature deformation and effectively solving the problem of incomplete results of children's sitting posture classification in the prior art.

[0063] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0064] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0065] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0066] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0067] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0068] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A data classification method based on prior information, characterized in that: The following steps are involved: Collect prior information about children’s sitting posture; Use sensors to capture the real-time dynamic characteristics of children's sitting posture; Performing data preprocessing on the child's sitting posture prior information and the real-time dynamic features to obtain preprocessed child's sitting posture prior information and preprocessed real-time dynamic features; Perform label matching on the preprocessed children’s sitting posture prior information to obtain the sitting posture category label; Construct a classifier based on the preprocessed children’s sitting posture prior information and sitting posture category labels; The preprocessed real-time dynamic features are classified by a classifier to obtain the sitting posture category label corresponding to the real-time dynamic features of the child's sitting posture as the sitting posture classification result of the real-time dynamic features; The specific steps of the tag matching are: Get all types of sitting posture category labels; Obtain the angle between the head and shoulders, the twisting amplitude of the body, and the twisting correction factor and input them into the body tilt angle calculation formula; The body inclination angle is calculated according to the body inclination angle calculation formula; The elastic modulus of the spine, the moment of inertia of the spine section, the length of the spine, the actual bending angle of the spine, the physiological curvature of the spine and the external force on the spine are obtained and input into the calculation formula of the degree of bending deformation of the spine; Add body tilt angle labels according to the body tilt angle, where the body tilt angle labels are divided into slight tilt labels, moderate tilt labels, severe tilt labels and no tilt labels; The degree of spinal curvature deformation is calculated according to the spinal curvature deformation degree calculation formula; Adding spinal curvature deformation degree labels according to the degree of spinal curvature deformation, wherein the spinal curvature deformation degree labels include normal curvature labels, mild deformation labels and severe deformation labels; The pressure distribution of each point on the buttocks is obtained based on the 3D scanning data; According to the pressure distribution of each point on the buttocks, a pressure distribution label of each point on the buttocks is added, wherein the pressure distribution label of each point on the buttocks includes a uniform distribution label, a local concentration label and a severe concentration label; The body inclination angle calculation formula is: In the formula, θ represents the body inclination angle, γ represents the angle between the head and shoulders, and k g represents the gravity correction factor, k β represents the torsion correction factor, and π represents the circumference of a circle; The calculation formula for the degree of spinal curvature deformation is: D S It indicates the degree of spinal curvature deformation, which is used to judge whether the child's sitting posture is correct. E indicates the elastic modulus of the spine, which is used to reflect the stiffness of the spine. I indicates the section inertia moment of the spine, which reflects the influence of the shape of the spine cross section on the curvature. L indicates the length of the spine, and θ s( x ) represents the actual bending angle of the spine at position x, θ ref (x) represents the physiological curvature of the spine at position x, x represents the position coordinate of any point along the length of the spine, which is used to represent any position between the starting point and the end point of the spine, F(x) represents the external force acting on the spine at position x, and dx represents a very small increment along the length of the spine, which is used to calculate the integral.

2. A data classification method based on prior information as claimed in claim 1, characterized in that: The step of collecting the prior information of the child's sitting posture comprises: Collect static images, videos and 3D scan data of correct and bad sitting postures of children of different ages, heights and weights; Collect information about children’s sitting posture in different scenarios.

3. A data classification method based on prior information as claimed in claim 1, characterized in that: The step of using a sensor to capture the real-time dynamic characteristics of the child's sitting posture comprises: The real-time dynamic characteristics of children's sitting posture include the body's tilt angle, spinal curvature deformation, and pressure distribution at each point on the buttocks; The body's tilt angle is calculated using the angular velocity and acceleration measured by the gyroscope and accelerometer; Use flexible sensors to measure the degree of spinal curvature deformation; The pressure sensor array is used to measure the pressure distribution at various points on the buttocks of a child when he sits on a seat cushion.

4. A data classification method based on prior information as claimed in claim 1, characterized in that: The data preprocessing step includes: Data cleaning was performed on the children’s sitting posture prior information to remove duplicate data and incomplete data and perform normalization processing; Perform image enhancement, cropping, and scaling operations on the pictures and videos of children’s sitting posture prior information to highlight and extract specified features; Convert the text description of the children's sitting posture prior information into numerical data; Align real-time dynamic features to timestamps and use filters to remove high-frequency noise from sensor data; The data of different scales and units are standardized to obtain the preprocessed children's sitting posture prior information and the preprocessed real-time dynamic features.

5. The data classification method based on prior information as claimed in claim 1, characterized in that: The step of constructing a classifier based on the preprocessed child sitting posture prior information and sitting posture category labels comprises: Input the preprocessed children’s sitting posture prior information into the classifier; According to the features in the prior information of children’s sitting posture, label matching is performed with the predefined sitting posture category labels; The matched labels and the corresponding prior information are integrated into a complete dataset to obtain a classifier.

6. A data classification method based on prior information as claimed in claim 1, characterized in that: The analysis steps of the sitting posture category label corresponding to the real-time dynamic feature of the child's sitting posture are as follows: The preprocessed real-time dynamic features of children’s sitting posture are used as input data and input into the classifier; The classifier outputs the corresponding sitting posture category label based on the input data.

7. A data classification system based on prior information, used to execute a data classification method based on prior information according to any one of claims 1 to 6, characterized in that: include: Information collection module, sensor module, preprocessing module, label matching module, classifier module, classification module; Wherein, the information collection module is used to collect prior information of children's sitting posture; The sensor module is used to capture the real-time dynamic characteristics of the child's sitting posture using a sensor; The preprocessing module is used to perform data preprocessing on the child's sitting posture prior information and the real-time dynamic features to obtain the preprocessed child's sitting posture prior information and the preprocessed real-time dynamic features; The label matching module is used to perform label matching on the preprocessed child sitting posture prior information to obtain a sitting posture category label; The classifier module is used to construct a classifier based on the preprocessed child sitting posture prior information and sitting posture category labels; The classification module is used to classify the preprocessed real-time dynamic features through a classifier to obtain a sitting posture category label corresponding to the real-time dynamic features of the child's sitting posture.

Images