Sitting posture assessment method and system

Abstract

The present invention relates to the technical field of health monitoring and intelligent assessment, and in particular to a sitting posture assessment method and system. Firstly, stress wave data, tactile data and thermal imaging data of a user's sitting posture are collected and preprocessed; then, multi-level recursive correction is performed on the preprocessed data based on the user's physiological characteristics to generate final correction data; then, basic features, topological features and their associated features are extracted through multi-stage fusion and dynamic cross-validation, associated feature data are generated, and a health score is performed on the data using a multi-dimensional dynamic scoring mechanism; finally, the health assessment process is optimized through causal chain analysis, and personalized feedback is generated to help the user adjust his sitting posture; the present invention effectively solves the problems of insufficient assessment accuracy and quantification difficulty in the prior art, and provides a more comprehensive and accurate health monitoring method.

Description

Technical Field

[0001] The present invention relates to the technical field of health monitoring and intelligent assessment, and in particular to a sitting posture assessment method and system. Background Art

[0002] Sitting posture assessment is essential to maintain a good body posture and prevent related health problems. Long-term poor sitting posture can lead to a series of health problems, such as spinal curvature, shoulder and neck pain, low back pain, etc. These problems not only affect the individual's quality of life, but may also cause long-term health risks. Therefore, it is of great significance to accurately assess and monitor the user's sitting posture to prevent these problems.

[0003] Existing sitting posture assessment technology (Chinese invention patent, publication number: CN113080941B, name: sitting posture assessment method and related device) mainly relies on obtaining the pressure distribution map of the user when sitting, and combines simple geometric and statistical methods to assess the sitting posture. For example, the existing technology uses a method combining infrared image recognition with a machine learning algorithm to assess the sitting posture, or determines the pressure line of the legs through the pressure distribution map, and assesses the degree of crossing of the user's legs based on the slope of the pressure line. However, these technologies have several defects:

[0004] The method of using infrared images combined with machine learning algorithms is very complex in processing, occupies a large amount of computing resources, and is difficult to promote in daily use; the existing pressure line method quantifies sitting posture by the slope of the pressure line of both legs, but this quantification method is too simple and cannot accurately capture the full picture of sitting posture, especially the lack of comprehensive analysis of multi-dimensional data, making it difficult to reflect actual health risks. Summary of the invention

[0005] In view of the many problems existing in the above-mentioned prior art, the present invention provides a sitting posture assessment method and system. The present invention collects multi-dimensional sitting posture data of the user, processes the data using a multi-level recursive correction algorithm, and performs sitting posture assessment in combination with a multi-dimensional dynamic scoring mechanism; finally, through causal chain analysis and real-time feedback optimization, the user is helped to adjust his sitting posture and prevent health problems; the present invention can provide more accurate and comprehensive sitting posture assessment results than traditional methods.

[0006] A sitting posture assessment method comprises the following steps:

[0007] Collecting a multi-dimensional data set of the user's sitting posture, the multi-dimensional data set including: stress wave data, tactile data and thermal imaging data, performing pre-processing to generate a pre-processed multi-dimensional data set;

[0008] Performing multi-level recursive correction on the preprocessed multi-dimensional data set based on the user's physiological characteristics to generate final corrected stress wave data, final corrected tactile data and final corrected thermal imaging data, and generating final fused data through multi-stage fusion and dynamic cross-validation;

[0009] Extract basic features, topological features and their associated features from the final fused data to generate associated feature data, use a multi-dimensional dynamic scoring mechanism to perform health scoring on the associated feature data, and combine historical data and environmental factors to generate final health score data. Analyze the final health score data and associated feature data through causal chain analysis to generate causal feedback data and optimize parameters in the health assessment process.

[0010] Generate feedback prompts based on the final health score data and causal feedback data, and optimize the user's sitting habits through dynamic adjustment and personalized suggestions.

[0011] Preferably, the preprocessing step includes: performing multi-level filtering and denoising processing on the stress wave data, tactile data and thermal imaging data respectively, performing time alignment processing on the collected multidimensional data sets using a time synchronization algorithm, and performing spatial alignment processing through a spatial correction algorithm to generate a preprocessed multidimensional data set.

[0012] Preferably, the multi-level recursive correction step includes:

[0013] Based on the user's physiological characteristics and historical data, the pre-processed multi-dimensional data set is preliminarily corrected, and the parameters of the correction model are optimized and adjusted through the back propagation algorithm to generate preliminarily corrected stress wave data, preliminarily corrected tactile data and preliminarily corrected thermal imaging data;

[0014] Further correcting the initially corrected stress wave data, the initially corrected tactile data and the initially corrected thermal imaging data, using a multi-level recursive algorithm to adjust the correction weights at each recursive stage, and generating refined corrected stress wave data, refined corrected tactile data and refined corrected thermal imaging data;

[0015] Combined with the user's historical data and dynamic physiological parameters, the refined and corrected stress wave data, the refined and corrected tactile data and the refined and corrected thermal imaging data are finally corrected to generate final corrected stress wave data, final corrected tactile data and final corrected thermal imaging data.

[0016] Preferably, the multi-stage fusion and dynamic cross-validation steps include:

[0017] Preliminarily fusing the final corrected stress wave data, the final corrected tactile data and the final corrected thermal imaging data, and generating preliminary fused data using a fusion algorithm based on weight adjustment;

[0018] On the basis of preliminary fusion data, the causal relationship between data sources is evaluated through the dynamic causal chain analysis method, and the weight of each data source is dynamically adjusted according to the causal relationship to generate intermediate fusion data;

[0019] The topological characteristics of the intermediate fusion data are analyzed, the credibility of different data sources is dynamically evaluated using the Bayesian inference network, and the parameters of the fusion algorithm are adjusted through the cross-validation mechanism to generate the verified final fusion data.

[0020] Preferably, the Bayesian inference network is used to perform credibility assessment on each data source in the intermediate fusion data in the dynamic cross-validation step, and adjust the weight of the data source in real time according to the assessment result.

[0021] Preferably, the step of extracting the basic features, topological features and their associated features includes:

[0022] Using topological data analysis methods, the topological features of the user's sitting posture are extracted from the final fused data, including the connectivity, stability of each part of the sitting posture, and the persistence of the overall structure;

[0023] The extracted topological features are correlated with the basic feature data, the influence of each feature on the sitting posture stability is calculated, and the correlation feature data reflecting the correlation of the sitting posture features is generated.

[0024] Preferably, the multi-dimensional dynamic scoring mechanism includes:

[0025] Based on the mixed Gaussian process model, a preliminary health score is performed on the associated feature data, the contribution of each feature in the health score is calculated, and the preliminary health score data is generated;

[0026] The preliminary health score data is compared with the user's historical health score data, and combined with the current environmental factors, the preliminary health score is dynamically adjusted using a real-time Bayesian optimization algorithm to ultimately generate final health score data that reflects the user's current health status.

[0027] Preferably, the real-time Bayesian optimization algorithm is calculated by the following formula:

[0028] Final health score Calculated by the following formula:

[0029]

[0030] in: For final health score; is the number of features; For the The weight of each feature; For the A preliminary health score for each characteristic; For the The environmental factor scores associated with each characteristic.

[0031] Preferably, the causal chain analysis identifies the causal relationship between basic features, topological features and their associated features and the final health score data through a dynamic causal model, and generates causal feedback data, and optimizes the parameters in the health assessment process by adjusting key parameters in the health assessment model.

[0032] A system for implementing the sitting posture assessment method, comprising:

[0033] A data acquisition module, used for acquiring a multi-dimensional data set of a user's sitting posture, the multi-dimensional data set including stress wave data, tactile data and thermal imaging data, and preprocessing the multi-dimensional data set to generate a preprocessed multi-dimensional data set;

[0034] A correction and fusion module, used for performing multi-level recursive correction on the preprocessed multi-dimensional data set based on the user's physiological characteristics to generate final corrected stress wave data, final corrected tactile data and final corrected thermal imaging data, and generating final fused data through multi-stage fusion and dynamic cross-validation;

[0035] A feature extraction and scoring module is used to extract basic features, topological features and their associated features from the final fused data, generate associated feature data, and use a multi-dimensional dynamic scoring mechanism to perform health scoring on the associated feature data, and generate final health scoring data in combination with historical data and environmental factors;

[0036] A causal analysis and optimization module, for analyzing the final health score data and associated characteristic data through causal chain analysis, generating causal feedback data and optimizing parameters in the health assessment process;

[0037] The feedback and adjustment module is used to generate feedback prompts based on the final health score data and the causal feedback data, and optimize the user's sitting habits through dynamic adjustment and personalized suggestions.

[0038] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0039] The present invention realizes a more comprehensive assessment of the user's sitting posture through multi-dimensional data collection and fusion technology, covering multiple data dimensions such as stress waves, touch and thermal imaging, making the assessment more accurate;

[0040] The present invention uses a multi-level recursive correction technology to achieve the difference processing between different physiological characteristics, thereby improving the personalization and accuracy of sitting posture assessment.

[0041] The present invention realizes the accurate quantification of sitting health risks through a multi-dimensional dynamic scoring mechanism, and dynamically adjusts the scoring in combination with historical data and environmental factors to ensure the real-time and accuracy of the evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the process of the present invention;

[0043] Figure 2 It is a schematic diagram of the multi-level recursive correction process in the present invention;

[0044] Figure 3 This is a schematic diagram of the multi-dimensional dynamic scoring mechanism in the present invention;

[0045] Figure 4 It is a structural block diagram of the system of the present invention. DETAILED DESCRIPTION

[0046] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0047] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0048] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0049] like Figure 1 As shown, a sitting posture assessment method comprises the following steps:

[0050] Collecting a multi-dimensional data set of the user's sitting posture, the multi-dimensional data set including: stress wave data, tactile data and thermal imaging data, performing pre-processing to generate a pre-processed multi-dimensional data set;

[0051] Preferably, the preprocessing step includes: performing multi-level filtering and denoising processing on the stress wave data, tactile data and thermal imaging data respectively, performing time alignment processing on the collected multidimensional data sets using a time synchronization algorithm, and performing spatial alignment processing through a spatial correction algorithm to generate a preprocessed multidimensional data set.

[0052] The multi-dimensional data set includes stress wave data, tactile data and thermal imaging data, which can comprehensively reflect the user's sitting posture characteristics from different angles. Specifically, stress wave data mainly reflects the pressure distribution of the user on the seat, tactile data records the contact intensity and angle changes between the user and the seat surface, and thermal imaging data indirectly evaluates the user's muscle activity and blood circulation by capturing the temperature distribution on the user's body surface.

[0053] The preprocessing step is a key step in normalizing the collected raw data, aiming to improve the accuracy and consistency of the data and ensure the effectiveness of subsequent analysis. First, multi-level filtering and denoising are performed on stress wave data, tactile data, and thermal imaging data. This step processes the data layer by layer through filters of different frequency bands to eliminate interference signals that may be introduced by environmental noise, sensor errors, etc., so as to obtain clearer and more accurate sitting posture feature data.

[0054] The preprocessing step includes using a time synchronization algorithm to time align multidimensional data sets. Since different types of data may be collected at different time points, in order to ensure that they can be effectively analyzed on the same time axis, the data needs to be time aligned. The time synchronization algorithm accurately calculates the time deviation between different data sources and adjusts their timestamps to make all data consistent in the time dimension, thus laying the foundation for subsequent multidimensional fusion analysis.

[0055] Spatial alignment is the last step of preprocessing, and the spatial position consistency of the data is adjusted through the spatial correction algorithm. This step mainly solves the differences in the spatial dimensions of the data collected by different sensors, such as sensor position deviation or spatial errors caused by unstable user posture. The spatial correction algorithm adjusts the spatial coordinates of different data sources based on known spatial reference points or through repeated iterative calculations, so that all data are consistent in the same spatial framework. This not only helps to improve the spatial accuracy of the data, but also makes the analysis of sitting posture characteristics more intuitive and accurate.

[0056] This multi-dimensional and standardized processing method ensures the accuracy of the data in time and space, retains the user's real sitting posture information to the greatest extent, and makes the final health assessment results more scientific and reliable.

[0057] Performing multi-level recursive correction on the preprocessed multi-dimensional data set based on the user's physiological characteristics to generate final corrected stress wave data, final corrected tactile data and final corrected thermal imaging data, and generating final fused data through multi-stage fusion and dynamic cross-validation;

[0058] In the present invention, by performing multi-level recursive correction on the pre-processed multi-dimensional data set and combining it with the user's physiological characteristics, the accuracy of sitting posture assessment can be greatly improved. Specifically, the multi-level recursive correction is based on the user's individual physiological data (such as height, weight, body shape, etc.), and the initially collected stress wave data, tactile data and thermal imaging data are progressively adjusted. The core of this process is to use the back propagation algorithm to gradually optimize the parameters of the correction model, so that each level of correction is closer to the user's actual physiological state, thereby generating more accurate final corrected stress wave data, final corrected tactile data and final corrected thermal imaging data.

[0059] Multi-level recursive correction is not just an enhanced version of a single correction, but a dynamic and progressive process. It gradually approaches the user's individual characteristics in each level of recursion by gradually introducing more detailed physiological parameters at each correction stage and weighting these parameters. For example, in the initial correction stage, the system may only use the user's basic physiological parameters for adjustment, and in the subsequent refined correction stage, more dynamic physiological parameters (such as real-time heart rate, body temperature changes, etc.) will be introduced for more detailed correction. The final correction result not only reflects the user's static physiological characteristics, but can also dynamically adapt to the user's real-time state changes.

[0060] After completing the multi-level recursive correction, the system will generate the final corrected stress wave data, tactile data and thermal imaging data. These data are user sitting posture feature data that have been adjusted with high precision and are highly individualized and accurate. Next, these corrected data will be further processed and integrated through multi-stage fusion and dynamic cross-validation. The principle of multi-stage fusion is to integrate various types of corrected data in time, space and feature dimensions through layer-by-layer fusion steps to form a comprehensive data set with higher consistency and richer information. Dynamic cross-validation introduces a cross-validation mechanism at each fusion stage. Through comparison and verification between different data sources, the weight and credibility of each data source are dynamically adjusted to ensure the accuracy and stability of the final fused data.

[0061] In applications, this fusion and cross-validation process is particularly important. For example, if some inconsistency is found between stress wave data and thermal imaging data during the sitting posture assessment, the system can identify this difference and make adjustments through dynamic cross-validation to optimize the fused data. The fused data finally generated is not just a simple superposition of multi-dimensional information, but high-quality data that has been verified and weighted, which can better reflect the user's actual sitting posture. The reliability and accuracy of this data provides a solid foundation for subsequent health scoring and sitting posture adjustment.

[0062] Through multi-level recursive correction, multi-stage fusion and dynamic cross-validation, the present invention achieves in-depth processing and optimization of user sitting posture data. This process effectively eliminates the deviation and inconsistency in the original data, so that the final generated fusion data has extremely high individual characteristics and accuracy, providing technical support for more accurate sitting posture assessment. This not only enhances the adaptability and robustness of the system, but also improves the scientificity and credibility of the overall assessment, laying the foundation for personalized health management.

[0063] like Figure 2 As shown, preferably, the multi-level recursive correction step includes:

[0064] Based on the user's physiological characteristics and historical data, the pre-processed multi-dimensional data set is preliminarily corrected, and the parameters of the correction model are optimized and adjusted through the back propagation algorithm to generate preliminarily corrected stress wave data, preliminarily corrected tactile data and preliminarily corrected thermal imaging data;

[0065] Further correcting the initially corrected stress wave data, the initially corrected tactile data and the initially corrected thermal imaging data, using a multi-level recursive algorithm to adjust the correction weights at each recursive stage, and generating refined corrected stress wave data, refined corrected tactile data and refined corrected thermal imaging data;

[0066] Combined with the user's historical data and dynamic physiological parameters, the refined and corrected stress wave data, the refined and corrected tactile data and the refined and corrected thermal imaging data are finally corrected to generate final corrected stress wave data, final corrected tactile data and final corrected thermal imaging data.

[0067] In the present invention, the multi-level recursive correction step is the key link to achieve accurate sitting posture assessment. This step first performs a preliminary correction on the preprocessed multidimensional data set based on the user's physiological characteristics and historical data. This process aims to use the known individual physiological characteristics of the user (such as height, weight, body shape) and the trends in historical data to make preliminary adjustments to the data. In this way, the initial deviation caused by external factors such as sensor measurement errors or environmental changes can be eliminated, making the data closer to the user's true physiological state.

[0068] In the preliminary correction, the back propagation algorithm is used to optimize and adjust the parameters of the correction model. The back propagation algorithm is a technology commonly used in neural network training. By calculating the contribution of each parameter to the error, these parameters are adjusted in reverse to minimize the overall error. In the present invention, this algorithm is used to adjust the parameters of the correction model. By continuously optimizing the parameters, the model can more accurately reflect the individual characteristics of the user, and generate preliminary corrected stress wave data, preliminary corrected tactile data and preliminary corrected thermal imaging data.

[0069] Next, the preliminary correction data is further corrected, which is the core part of the multi-level recursive correction. The preliminary correction data is processed using a multi-level recursive algorithm, and the correction weight is adjusted at each recursive stage, so that the correction result of each level is closer to the user's actual sitting posture. The principle of the multi-level recursive algorithm is to perform multiple iterative corrections, and each iteration adjusts the weight and optimizes the parameters according to the results of the previous correction, gradually approaching the optimal correction result. In this way, the generated refined corrected stress wave data, refined corrected tactile data, and refined corrected thermal imaging data have higher accuracy and personalized characteristics.

[0070] The final correction step combines the user's historical data and dynamic physiological parameters to perform the final correction processing on the refined correction data. At this stage, the system introduces the user's real-time physiological state data, such as heart rate, body temperature, etc. These dynamic parameters can reflect the user's true state at the current moment. By integrating historical data and dynamic physiological parameters, the correction model can be further optimized to generate final corrected stress wave data, final corrected tactile data, and final corrected thermal imaging data. These data not only represent the user's static characteristics, but also reflect the user's dynamic state during the assessment, providing highly accurate input for subsequent sitting posture assessment.

[0071] For example, if a user's historical data indicates that they are prone to posture deviation when sitting for a long time, and their current physiological data (such as increased heart rate) indicates that they may be in a state of tension, through the final correction step, the system can more accurately identify this deviation trend and make adjustments in the correction, thereby generating more accurate final correction data. These data can truly reflect the user's current sitting posture and avoid inaccurate assessments caused by short-term physiological fluctuations or historical trends.

[0072] Preferably, the multi-stage fusion and dynamic cross-validation steps include:

[0073] Preliminarily fusing the final corrected stress wave data, the final corrected tactile data and the final corrected thermal imaging data, and generating preliminary fused data using a fusion algorithm based on weight adjustment;

[0074] On the basis of preliminary fusion data, the causal relationship between data sources is evaluated through the dynamic causal chain analysis method, and the weight of each data source is dynamically adjusted according to the causal relationship to generate intermediate fusion data;

[0075] The topological characteristics of the intermediate fusion data are analyzed, the credibility of different data sources is dynamically evaluated using the Bayesian inference network, and the parameters of the fusion algorithm are adjusted through the cross-validation mechanism to generate the verified final fusion data.

[0076] The final corrected stress wave data, final corrected tactile data, and final corrected thermal imaging data are preliminarily fused. This process uses a fusion algorithm based on weight adjustment to integrate the corrected data from different data sources according to the set weights. The core principle of preliminary fusion is to assign appropriate weights to each data type (such as stress wave, tactile, and thermal imaging) for its specific characteristics to ensure that the fused data set can fully reflect the user's sitting posture. For example, if a user's stress wave data is more important in certain sitting postures, the weight of the stress wave data will be set higher during the preliminary fusion, and the generated preliminary fused data will better reflect the user's pressure distribution in that sitting posture.

[0077] Based on the preliminary fusion data, the causal relationship between the data sources is evaluated through the dynamic causal chain analysis method, and the weight of each data source is dynamically adjusted according to the evaluation results. The principle of dynamic causal chain analysis is to identify the potential causal relationship between different data sources, and then adjust the importance of these data in the overall fusion process. For example, tactile data may directly affect the change of stress wave data. This relationship can be identified through causal chain analysis, and the weight of tactile data can be appropriately increased in subsequent fusion, so that the fused data can more accurately reflect the actual sitting state of the user. The intermediate fusion data generated by this process has higher internal consistency and logical relevance.

[0078] The topological feature analysis of the intermediate fusion data is performed, and the credibility of different data sources is dynamically evaluated using the Bayesian inference network. At this stage, the topological feature analysis is used to extract the structural features in the data, especially the connectivity and stability of each part of the sitting posture. The Bayesian inference network is used to dynamically evaluate the credibility of each data source, that is, to judge the reliability and accuracy of each data under different circumstances. This evaluation process can not only detect potential abnormal data, but also adjust the parameters of the fusion algorithm through the cross-validation mechanism to make the final generated data more accurate and stable. Through cross-validation, the system can verify the results of each data source to ensure that the final fusion data not only reflects the comprehensive information of each data source, but also eliminates possible errors or deviations.

[0079] For example, if some inconsistency is found between thermal imaging data and stress wave data during the sitting posture assessment process, the system can evaluate the credibility of each through the Bayesian inference network and dynamically adjust its weight in the final fusion. For example, if the thermal imaging data is assessed as low credibility, the weight of the data in the fusion process will be reduced, thereby reducing its negative impact on the final assessment result. After topological feature analysis and dynamic cross-validation, the final fusion data generated will be more reliable and can more accurately reflect the user's overall sitting posture.

[0080] Preferably, the Bayesian inference network is used to perform credibility assessment on each data source in the intermediate fusion data in the dynamic cross-validation step, and adjust the weight of the data source in real time according to the assessment result.

[0081] In the present invention, stress wave data, tactile data and thermal imaging data after multi-level recursive correction have been fused into intermediate fused data. However, due to different acquisition methods, sensor characteristics and environmental influences, different data sources may have inconsistencies or accuracy differences in some cases. Therefore, pure fused data may contain some uncertainty or error. At this time, it is necessary to evaluate the credibility of each data source through a Bayesian inference network.

[0082] The principle of the Bayesian inference network is to use conditional probability to evaluate the credibility of each data source in the current situation. It determines which data sources are more reliable in the current evaluation scenario by comparing historical data and current fused data. For example, the system may analyze the performance of each data source under similar sitting conditions in the past to determine the mutual consistency between the current tactile data and the stress wave data. If the tactile data has historically shown high reliability, and the current stress wave data deviates significantly from it, the Bayesian inference network may reduce the credibility of the stress wave data, thereby giving the tactile data a higher weight in the final fusion process.

[0083] The effect of this dynamic evaluation process is that the system can adaptively adjust the contribution of each data source in the fusion process to ensure that the final evaluation result is more accurate. By adjusting the weight of the data source in real time, the Bayesian reasoning network can effectively reduce the negative impact caused by the abnormality or inaccuracy of a single data source and enhance the overall quality and stability of the fused data. For example, if at a certain moment, the thermal imaging data causes measurement deviation due to sensor failure, the Bayesian reasoning network can quickly identify this problem by analyzing the correlation between the data, and automatically reduce the weight of the thermal imaging data to prevent it from causing excessive misleading to the final evaluation result.

[0084] In practice, this mechanism ensures that the evaluation system can always provide reliable sitting posture evaluation results when facing changing user states and external environments through real-time calculation and dynamic adjustment. This approach not only improves the robustness of the system, but also makes the system highly adaptive, and can maintain the accuracy and stability of the evaluation in different users and scenarios. Ultimately, the application of the Bayesian reasoning network enables the present invention to flexibly respond to various complex situations, provide users with more accurate sitting posture health recommendations, and ensure the scientificity and feasibility of the evaluation process.

[0085] Extract basic features, topological features and their associated features from the final fused data to generate associated feature data, use a multi-dimensional dynamic scoring mechanism to perform health scoring on the associated feature data, and combine historical data and environmental factors to generate final health score data. Analyze the final health score data and associated feature data through causal chain analysis to generate causal feedback data and optimize parameters in the health assessment process.

[0086] Extract basic features from the final fused data. These basic features include key physical parameters of the user's sitting posture, such as pressure distribution on the chair surface, uniformity of back support, position of the body's center of gravity, etc. These basic features can directly reflect the user's main body load and posture stability when sitting, and are important indicators for judging whether the sitting posture is healthy.

[0087] The extraction of topological features further deepens the understanding of the user's sitting posture. Topological feature analysis uses topological data analysis methods to reveal the connectivity and structural stability between the various parts of the user's sitting posture. For example, topological features can be used to identify whether the user has left-right asymmetry or whether a part of the body remains in an unhealthy bent state for a long time. The introduction of topological features makes the evaluation process not limited to the analysis of linear data, but also includes the understanding of complex sitting posture structures, thereby more comprehensively reflecting the user's actual posture.

[0088] In addition, the extraction of associated features generates higher-level associated feature data by analyzing the relationship between basic features and topological features. For example, the system may analyze the correlation between pressure distribution and the position of the body's center of gravity to determine whether the user has applied uneven pressure on the seat surface due to the center of gravity shift. This type of associated feature data can reveal more complex sitting posture problems and is an important basis for the final health score.

[0089] After the feature data is extracted, the system uses a multi-dimensional dynamic scoring mechanism to perform a health score on the associated feature data. The core of the multi-dimensional dynamic scoring mechanism is that it not only evaluates the independent health status of each feature, but also takes into account the interactions and weights between features. For example, some features (such as the uniformity of back support) may be more important than other features (such as foot support) in a specific sitting posture. The dynamic scoring mechanism will dynamically adjust the weights of each feature according to different sitting conditions and environmental factors to generate a comprehensive health score.

[0090] The calculation of the health score does not only rely on the current feature data, but also combines the user's historical data and current environmental factors. For example, the system will compare the user's past sitting posture data to identify long-term unhealthy trends and consider the impact of the current environment (such as the hardness of the chair, ambient temperature, etc.) on the sitting posture. This comprehensive analysis combining historical and environmental factors can generate more personalized final health score data that reflects the user's actual health status in the current scenario.

[0091] After the final health score data is generated, the system further analyzes it through causal chain analysis. The purpose of causal chain analysis is to identify the impact path of each feature on the health score and find out which features play a key role in the final score. For example, the system may find that changes in back support have the greatest impact on the user's overall health score, which means that improving back support may be an effective way to improve the user's sitting health. Through this causal analysis, the system is able to generate causal feedback data, which can be directly used to optimize parameters in the health assessment process, so that the assessment model can more accurately reflect the user's health status in future assessments.

[0092] For example, if the system identifies that the user's foot support is insufficient due to long-term sitting, resulting in a low overall health score, the system will generate feedback data through causal chain analysis, suggesting that the importance of foot support should be increased in future assessments, and adjust the weight of the scoring mechanism accordingly. This optimization process enables the system to adaptively improve the assessment model and continuously improve the accuracy and reliability of the assessment results.

[0093] By extracting basic features, topological features and their associated features from the final fused data, and using a multi-dimensional dynamic scoring mechanism to perform health scoring, the present invention can conduct a comprehensive and accurate assessment of the user's sitting health status. By combining the comprehensive analysis of historical data and environmental factors, the final health score data generated can truly reflect the user's current status. The causal feedback data generated by causal chain analysis provides the possibility of continuous optimization for future evaluation models, ensuring that the evaluation process continues to evolve and providing users with personalized and scientific health management recommendations.

[0094] Preferably, the step of extracting the basic features, topological features and their associated features includes:

[0095] Using topological data analysis methods, the topological features of the user's sitting posture are extracted from the final fused data, including the connectivity, stability of each part of the sitting posture, and the persistence of the overall structure;

[0096] The extracted topological features are correlated with the basic feature data, the influence of each feature on the sitting posture stability is calculated, and the correlation feature data reflecting the correlation of the sitting posture features is generated.

[0097] The extraction of topological features utilizes the topological data analysis method, which is a core innovation of the present invention. The topological data analysis method reveals the posture characteristics and dynamic changes of users in different sitting postures by identifying and evaluating the connectivity, stability and persistence of the overall structure of each part in the sitting posture.

[0098] Topological features play a particularly important role in sitting posture assessment. It not only focuses on the spatial position of various parts of the user's body, but also analyzes the relationship between these parts. For example, how the user's back and buttocks are supported in a sitting position, and whether the legs and feet are in stable contact with the ground, can all be reflected by topological features. The extraction process of topological features includes identifying key nodes in the user's sitting posture (such as body joints and support points), and then analyzing the connectivity between these nodes. For example, if the user's back contacts the back of the chair less after sitting for a long time, this may mean that the user is leaning forward, resulting in an unhealthy posture. Through topological data analysis, this change can be captured and quantified in real time to generate specific topological feature data.

[0099] Next, the extracted topological features are correlated with the basic feature data. The core of this step is to evaluate the impact of each feature on the overall sitting posture stability through comparison and calculation. Basic features usually include directly measured physical parameters, such as pressure distribution on the seat, the user's center of gravity position, etc. These basic features can intuitively reflect the user's posture state, but may lack a deep understanding of complex sitting posture problems when used alone. By correlating the topological features with these basic features, we can have a more comprehensive understanding of the health of the sitting posture.

[0100] The principle of association analysis is to calculate the contribution of each feature to sitting posture stability. For example, if the system detects a strong correlation between the uniformity of the user's back pressure (basic feature) and the contact area between his back and the chair back (topological feature), this indicates that the back support has a significant impact on the stability of the user's overall sitting posture. Through this association analysis, the system generates association feature data that reflects these relationships. These association feature data not only show the influence of a single feature, but also reveal the interaction between multiple features.

[0101] For example, suppose a user tends to lean forward when sitting, resulting in a gradual decrease in the contact area between the back and the chair back. The basic features can be used to directly observe changes in pressure distribution, while the topological features can be used to analyze the user's posture structure. After correlating these two features, the system may find that the contact area between the back and the chair back is highly positively correlated with the overall sitting posture stability, a finding that can be reflected through the associated feature data. This associated feature data enables the system to more accurately assess the user's sitting posture health and provide targeted adjustment suggestions.

[0102] Through this in-depth feature extraction and association analysis, the present invention is able to capture subtle changes in sitting posture and reveal the impact of these changes on the user's overall health through systematic analysis. This multi-dimensional and multi-level feature analysis method makes sitting posture assessment no longer rely solely on static analysis of a single data, but comprehensively assesses the user's sitting posture health status in a dynamic and interactive way. This not only improves the accuracy of the assessment, but also provides users with more scientific and personalized health management suggestions.

[0103] like Figure 3 As shown, preferably, the multi-dimensional dynamic scoring mechanism includes:

[0104] Based on the mixed Gaussian process model, a preliminary health score is performed on the associated feature data, the contribution of each feature in the health score is calculated, and the preliminary health score data is generated;

[0105] The preliminary health score data is compared with the user's historical health score data, and combined with the current environmental factors, the preliminary health score is dynamically adjusted using a real-time Bayesian optimization algorithm to ultimately generate final health score data that reflects the user's current health status.

[0106] In the sitting posture assessment of the present invention, the multi-dimensional dynamic scoring mechanism is a key assessment tool, which can accurately assess the user's current sitting posture health status through complex mathematical models and algorithms. First, this mechanism is based on the mixed Gaussian process model to perform a preliminary health score on the associated feature data. The mixed Gaussian process model is a powerful nonlinear regression tool that can handle high-dimensional and complex associated data. Through this model, the system can analyze the distribution of each associated feature data and calculate the contribution of each feature to the overall health score.

[0107] In specific applications, the mixed Gaussian process model can capture subtle differences in the user's sitting posture characteristics. For example, the system may analyze the user's back pressure distribution, center of gravity position, and angle changes of various body parts. After preliminary processing, these data will be input into the mixed Gaussian process model. By learning from historical data and processing current data, the model can calculate the influence of each feature on the overall health score in the current scenario, thereby generating preliminary health score data. These preliminary health score data reflect the independent contribution of each feature and provide a basic health status assessment.

[0108] However, relying solely on the preliminary health score data does not fully reflect the user's true health status. In order to obtain a more accurate assessment result, the system compares the preliminary health score data with the user's historical health score data. Through this comparative analysis, the system can identify long-term trends and changes in the user's sitting posture. For example, if the user's sitting posture score gradually decreases over a period of time, the system can identify this trend and give appropriate consideration in the current health assessment.

[0109] In addition to historical data, current environmental factors are also important variables that affect the sitting health score. Environmental factors may include the hardness of the chair, the temperature of the working environment, the user's fatigue level, etc. These factors can significantly affect the user's sitting performance. For example, when fatigued, the user may be more likely to lean forward or bend over, and these changes need to be reflected in the score. To this end, the system uses a real-time Bayesian optimization algorithm to dynamically adjust the preliminary health score. By modeling the relationship between environmental factors and preliminary scores, the Bayesian optimization algorithm can adaptively adjust the scoring results in a dynamic environment, so that the final generated health score is more in line with the user's actual state.

[0110] For example, suppose a user has a low uniformity score for back support when they are tired, but performs well when they are normal. Through a real-time Bayesian optimization algorithm, the system can adjust the weight of this feature according to the current level of fatigue, reduce its negative impact on the overall health score, and generate a more balanced and accurate final health score data. The final health score not only reflects the user's current sitting health status, but also takes into account the user's long-term trends and current environmental factors, ensuring that the score result is highly personalized and accurate.

[0111] The multi-dimensional dynamic scoring mechanism combines the mixed Gaussian process model and the real-time Bayesian optimization algorithm to accurately evaluate the user's sitting health status under multi-dimensional data and dynamic environments. The preliminary health score data provides a basis for subsequent adjustments, and the comparison with historical data and the combination of environmental factors make the final health score more accurate and personalized. This mechanism ensures that the system can reflect the user's health status in real time and provide users with scientific and personalized sitting posture adjustment suggestions, thereby effectively preventing and improving health problems caused by poor sitting posture.

[0112] Preferably, the real-time Bayesian optimization algorithm is calculated by the following formula:

[0113] Final health score Calculated by the following formula:

[0114]

[0115] in: For final health score; is the number of features; For the The weight of each feature; For the A preliminary health score for each characteristic; For the The environmental factor scores associated with each characteristic.

[0116] The real-time Bayesian optimization algorithm generates the final health score through a comprehensive calculation formula. This process can effectively combine the health scores of multiple features with the influence of environmental factors to obtain a score that accurately reflects the user's current sitting health status. The core of the algorithm is to dynamically adjust the weight of each feature so that the final score can adapt to the user's current state and external environmental changes.

[0117] First, the basic structure of the formula is a weighted average calculation that combines the preliminary health score of each characteristic with the relevant environmental factor scores. Specifically, the final health score in the formula It is obtained by weighted summation of all features, where each feature Weight Determines the influence of this feature in the overall score.

[0118] In this formula, Represents the total number of features, which means that the system will consider the health scores of multiple features. The preliminary health score of each feature is determined by the previous analysis process and reflects the health status of the feature in the current situation. For example, if a user's sitting posture causes uneven pressure distribution on the back, it may cause their initial health score to Lower.

[0119] However, relying solely on preliminary health scores may not fully reflect the user's actual health status, as different behaviors and environmental factors may have different effects on these scores. Therefore, the formula also introduces an environmental factor score associated with the feature The environmental factor score takes into account external conditions such as the hardness of the seat, the current ambient temperature, and the user's fatigue level. These conditions may affect the credibility and importance of the feature representation. For example, in a colder environment, the user's muscles may be more tense, resulting in abnormal body temperature distribution in the thermal imaging data. In this case, the environmental factor score This can reflect the impact of this environment on health scores.

[0120] Weights in the formula It is the key to the Bayesian optimization algorithm, which adjusts the weight of each feature in real time to reflect its importance in different situations. If a feature is greatly affected by the environment in the current situation, the algorithm may reduce its weight. , which reduces its contribution to the final health score. On the contrary, if a feature is considered to be a core indicator of the current health status (such as uniformity of back support in a fatigued state), the algorithm will increase its weight so that it accounts for a larger proportion in the final score.

[0121] For example, if the system detects that the user's working environment is in a high temperature, resulting in abnormal thermal imaging data, then the system will dynamically adjust the weight of the thermal imaging data through the real-time Bayesian optimization algorithm. , which may reduce its influence in the overall score to avoid inaccurate environmental data leading to erroneous health scores. At the same time, the system will increase its reliance on other more stable features (such as tactile data and stress wave data) to generate a more accurate final health score. .

[0122] This dynamic adjustment mechanism ensures that the final health score not only reflects the current state of each feature, but also comprehensively considers the impact of environmental factors on these features. In this way, the system can adaptively optimize the scoring process so that the scoring results are closer to the user's actual health status and more personalized. Ultimately, users can get targeted sitting posture adjustment suggestions based on these accurate health score data, thereby effectively preventing or improving health problems caused by poor sitting posture. This dynamic scoring mechanism not only improves the evaluation accuracy of the system, but also provides users with scientific and real-time health management tools to help users maintain good posture and health.

[0123] Preferably, the causal chain analysis identifies the causal relationship between basic features, topological features and their associated features and the final health score data through a dynamic causal model, and generates causal feedback data, and optimizes the parameters in the health assessment process by adjusting key parameters in the health assessment model.

[0124] The core of causal chain analysis lies in the application of dynamic causal models. Dynamic causal models are a tool that can identify causal relationships between variables, and are particularly suitable for analyzing the mutual influence of multiple variables in complex systems. In the present invention, these variables include the user's basic characteristics (such as back pressure distribution, weight distribution, etc.), topological characteristics (such as sitting stability, connectivity of various parts of the body, etc.) and correlation characteristics between these characteristics. Through the causal model, the system can identify which characteristics have a direct or indirect impact on the final health score.

[0125] For example, if the user's back pressure distribution shows unevenness, and this phenomenon is related to the body's center of gravity shift in the topological characteristics, then the causal chain analysis may reveal that the center of gravity shift is the main cause of the uneven pressure. This identification process can not only explain why the current health score is low, but also provide important adjustment basis for subsequent health assessments.

[0126] The causal feedback data generated by causal chain analysis is a kind of inference result based on causal relationship. These feedback data can reflect which features have the greatest impact on the health score in the current situation and the interaction between these features. For example, the system may find that when the user maintains an unhealthy sitting posture for a long time, the relationship between the back support and the leg support plays an important role in the final score. The causal feedback data not only reveals the relationship between these features, but also provides optimization suggestions for the system.

[0127] Through this causal feedback data, the system can dynamically adjust key parameters in the health assessment model. The adjustment process is based on the identified causal relationships, allowing the model to more accurately reflect the user's actual health status. For example, if the causal analysis shows that leg support contributes more to the health score in some cases, the system can adjust the model parameters to increase the weight of the leg support feature in the assessment process, thereby optimizing the final health score result.

[0128] This optimization is not only an adjustment to the current assessment, but also provides a basis for improving future assessments. As more data is accumulated and analyzed, the system's causal chain analysis model will continue to evolve, gradually improving its accuracy in identifying causal relationships and its ability to optimize health assessments. Ultimately, the system will be able to provide more precise and personalized health management recommendations to help users maintain good sitting habits and health status over the long term.

[0129] Through causal chain analysis, the system can identify and understand complex feature relationships, optimize the parameters in the evaluation model, and make the evaluation results closer to the user's actual situation. This process not only improves the intelligence level of the system, but also provides users with scientific and accurate health assessments and suggestions, thereby effectively preventing and improving health problems caused by poor sitting posture. In general, causal chain analysis is one of the key technologies in this invention to ensure the scientificity and accuracy of health assessment results.

[0130] Generate feedback prompts based on the final health score data and causal feedback data, and optimize the user's sitting habits through dynamic adjustment and personalized suggestions.

[0131] The final health score data is generated through a multi-dimensional dynamic scoring mechanism, which integrates the multi-dimensional characteristics of the user's current sitting posture, historical data, and environmental factors, reflecting the user's overall health status at the current moment. The causal feedback data is obtained through causal chain analysis, which reveals the impact path of each feature on the health score and points out which features contribute the most to the current health score or have the most significant impact.

[0132] Based on these two types of data, the system is able to generate specific feedback prompts. The process of generating feedback prompts utilizes the results of health scores and causal analysis to provide targeted and actionable suggestions. For example, if the system detects that the user's current sitting posture score is low, and causal analysis shows that insufficient back support is the main reason, the feedback prompt may suggest that the user adjust the chair back angle or increase lumbar support. This feedback prompt not only directly points out the current problem, but also provides users with practical improvement measures.

[0133] While generating feedback prompts, the system can also optimize the user's sitting habits through a dynamic adjustment mechanism. The core of the dynamic adjustment mechanism is that it can dynamically adjust the content and intensity of suggestions based on the user's real-time data and historical trends. For example, if the system detects that the user's sitting posture has gradually deteriorated over the past period of time, the dynamic adjustment mechanism may increase the frequency of feedback prompts and provide more detailed suggestions. Conversely, if the user's sitting posture gradually improves, the system may reduce the frequency of reminders to avoid excessive intervention.

[0134] Personalized recommendations are provided based on the user's individual characteristics and long-term data analysis. These recommendations not only take into account the user's current status, but also combine the user's historical data, personal preferences, and environmental factors. For example, for a user who works at a desk for a long time, the system may recommend that he stand up regularly or adjust the height of his chair to improve blood circulation. For a user with a history of chronic back pain, the recommendations may focus on adjusting the sitting posture and using specific support equipment to reduce the burden on the back.

[0135] For example, suppose the system detects that a user has a long-term forward leaning problem in sitting, and this problem leads to uneven distribution of back pressure. Through the final health score and causal analysis, the system can identify the root cause of the problem. The system then generates feedback prompts, suggesting that the user adjust the chair back or use a lumbar support cushion to improve the posture. At the same time, the system may provide personalized suggestions based on the user's historical data and current environment (such as desk height, chair type), guiding the user on how to optimize the sitting posture in their specific environment. In addition, the system may dynamically adjust the frequency of reminders and provide more reminders during time periods when users are most prone to posture problems (such as peak work hours).

[0136] like Figure 4 As shown, a system for implementing the sitting posture assessment method comprises:

[0137] The data acquisition module is used to collect a multi-dimensional data set of the user's sitting posture, the multi-dimensional data set includes stress wave data, tactile data and thermal imaging data, and pre-process the multi-dimensional data set to generate a pre-processed multi-dimensional data set; this module is responsible for collecting multi-dimensional data of the user's sitting posture, including stress wave data, tactile data and thermal imaging data. These data reflect the user's physical state from different dimensions. In order to ensure the accuracy and consistency of the data, the module also pre-processes the collected data, removes noise and corrects time and space errors, and generates a more standardized pre-processed multi-dimensional data set.

[0138] The correction and fusion module is used to perform multi-level recursive correction on the preprocessed multi-dimensional data set based on the user's physiological characteristics to generate final corrected stress wave data, final corrected tactile data and final corrected thermal imaging data, and generate final fused data through multi-stage fusion and dynamic cross-validation; this module uses the user's physiological characteristics to perform multi-level recursive correction on the preprocessed multi-dimensional data set. Through this process, the data can be gradually optimized and corrected to generate final corrected stress wave data, tactile data and thermal imaging data. Subsequently, the module also integrates different types of data through multi-stage fusion and dynamic cross-validation to generate a comprehensive and reliable final fused data. This fused data can fully reflect the user's sitting posture characteristics and eliminate the contradictions and deviations between different data sources.

[0139] The feature extraction and scoring module is used to extract basic features, topological features and their associated features from the final fused data, generate associated feature data, and use a multi-dimensional dynamic scoring mechanism to perform health scoring on the associated feature data, and generate the final health scoring data in combination with historical data and environmental factors; in this module, basic features, topological features and their associated features are extracted from the final fused data. These features describe various aspects of the user's sitting posture, including the relationship between body parts and the stability of the overall structure. Then, the system uses a multi-dimensional dynamic scoring mechanism to perform health scoring on these features, and generates the final health scoring data in combination with historical data and current environmental factors. This process ensures the personalization and real-time nature of the scoring, making the evaluation results more in line with the user's actual situation.

[0140] The causal analysis and optimization module is used to analyze the final health score data and related feature data through causal chain analysis, generate causal feedback data and optimize the parameters in the health assessment process; this module identifies the causal relationship between features and the final health score through causal chain analysis. Through this analysis, the system can find out which features have the greatest impact on sitting health and generate causal feedback data. These data are used to optimize the key parameters in the health assessment process, allowing the system to continuously adjust and improve the accuracy of the assessment.

[0141] The feedback and adjustment module is used to generate feedback prompts based on the final health score data and causal feedback data, and optimize the user's sitting habits through dynamic adjustment and personalized suggestions. Based on the final health score data and causal feedback data, the module generates feedback prompts and provides specific sitting adjustment suggestions to the user. The system also optimizes the user's sitting habits in real time through a dynamic adjustment mechanism so that it can adapt to different work and living environments. Personalized suggestions ensure the effectiveness and feasibility of adjustment measures, helping users to continuously improve their sitting posture and prevent potential health problems.

[0142] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.

[0143] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A sitting posture assessment method, characterized in that: The following steps are involved: Collecting a multi-dimensional data set of the user's sitting posture, the multi-dimensional data set includes: stress wave data, tactile data and thermal imaging data, and pre-processing to generate a pre-processed multi-dimensional data set; Performing multi-level recursive correction on the preprocessed multi-dimensional data set based on the user's physiological characteristics to generate final corrected stress wave data, final corrected tactile data and final corrected thermal imaging data, and generating final fused data through multi-stage fusion and dynamic cross-validation; The multi-level recursive correction step comprises: Based on the user's physiological characteristics and historical data, the pre-processed multi-dimensional data set is preliminarily corrected, and the parameters of the correction model are optimized and adjusted through the back propagation algorithm to generate preliminarily corrected stress wave data, preliminarily corrected tactile data and preliminarily corrected thermal imaging data; Further correcting the initially corrected stress wave data, the initially corrected tactile data and the initially corrected thermal imaging data, using a multi-level recursive algorithm to adjust the correction weights at each recursive stage, and generating refined corrected stress wave data, refined corrected tactile data and refined corrected thermal imaging data; Combining the historical data and dynamic physiological parameters of the user, finally correcting the refined and corrected stress wave data, the refined and corrected tactile data, and the refined and corrected thermal imaging data to generate final corrected stress wave data, final corrected tactile data, and final corrected thermal imaging data; The multi-stage fusion and dynamic cross-validation steps include: Preliminarily fusing the final corrected stress wave data, the final corrected tactile data and the final corrected thermal imaging data, and generating preliminary fused data using a fusion algorithm based on weight adjustment; On the basis of preliminary fusion data, the causal relationship between data sources is evaluated through the dynamic causal chain analysis method, and the weight of each data source is dynamically adjusted according to the causal relationship to generate intermediate fusion data; Perform topological feature analysis on the intermediate fusion data, dynamically evaluate the credibility of different data sources using the Bayesian inference network, and adjust the parameters of the fusion algorithm through the cross-validation mechanism to generate the final verified fusion data; Extract basic features, topological features and their associated features from the final fused data to generate associated feature data, use a multi-dimensional dynamic scoring mechanism to perform health scoring on the associated feature data, and combine historical data and environmental factors to generate final health score data. Analyze the final health score data and associated feature data through causal chain analysis to generate causal feedback data and optimize parameters in the health assessment process. The steps of extracting the basic features, topological features and their associated features include: Using topological data analysis methods, the topological features of the user's sitting posture are extracted from the final fused data, including the connectivity, stability of each part of the sitting posture, and the persistence of the overall structure; The extracted topological features are correlated with the basic feature data, the influence of each feature on the sitting posture stability is calculated, and the correlation feature data reflecting the correlation of the sitting posture features is generated; The multi-dimensional dynamic scoring mechanism includes: Based on the mixed Gaussian process model, a preliminary health score is performed on the associated feature data, the contribution of each feature in the health score is calculated, and the preliminary health score data is generated; Compare the preliminary health score data with the user's historical health score data, and use a real-time Bayesian optimization algorithm to dynamically adjust the preliminary health score based on current environmental factors, ultimately generating a final health score data that reflects the user's current health status; Generate feedback prompts based on the final health score data and causal feedback data, and optimize the user's sitting habits through dynamic adjustment and personalized suggestions.

2. The sitting posture assessment method according to claim 1, characterized in that: The preprocessing step includes: performing multi-level filtering and denoising processing on the stress wave data, tactile data and thermal imaging data respectively, performing time alignment processing on the collected multidimensional data sets using a time synchronization algorithm, and performing spatial alignment processing using a spatial correction algorithm to generate a preprocessed multidimensional data set.

3. The sitting posture assessment method according to claim 1, characterized in that: The Bayesian inference network is used to evaluate the credibility of each data source in the intermediate fusion data in the dynamic cross-validation step, and adjust the weight of the data source in real time according to the evaluation result.

4. The sitting posture assessment method according to claim 1, characterized in that: The real-time Bayesian optimization algorithm is calculated by the following formula: Final health score Calculated by the following formula: in: For final health score; is the number of features; For the The weight of each feature; For the A preliminary health score for each characteristic; For the The environmental factor scores associated with each characteristic.

5. The sitting posture assessment method according to claim 1, characterized in that: The causal chain analysis identifies the causal relationship between basic features, topological features and their associated features and final health score data through a dynamic causal model, generates causal feedback data, and optimizes the parameters in the health assessment process by adjusting key parameters in the health assessment model.

6. A system for implementing the sitting posture assessment method according to any one of claims 1 to 5, characterized in that: include: A data acquisition module, used for acquiring a multi-dimensional data set of a user's sitting posture, the multi-dimensional data set including stress wave data, tactile data and thermal imaging data, and preprocessing the multi-dimensional data set to generate a preprocessed multi-dimensional data set; A correction and fusion module, used for performing multi-level recursive correction on the preprocessed multi-dimensional data set based on the user's physiological characteristics to generate final corrected stress wave data, final corrected tactile data and final corrected thermal imaging data, and generating final fused data through multi-stage fusion and dynamic cross-validation; A feature extraction and scoring module is used to extract basic features, topological features and their associated features from the final fused data, generate associated feature data, and use a multi-dimensional dynamic scoring mechanism to perform health scoring on the associated feature data, and generate final health scoring data in combination with historical data and environmental factors; A causal analysis and optimization module, for analyzing the final health score data and associated characteristic data through causal chain analysis, generating causal feedback data and optimizing parameters in the health assessment process; The feedback and adjustment module is used to generate feedback prompts based on the final health score data and the causal feedback data, and optimize the user's sitting habits through dynamic adjustment and personalized suggestions.

Images