Health state assessment method and device, electronic equipment and storage medium

Abstract

The invention belongs to the technical field of health state assessment, and discloses a health state assessment method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining the physiological feature data of a user in a preset period, constructing the physiological feature data into tensor data containing an individual dimension, a feature dimension and a time dimension, complementing missing data in the tensor data by using a frequency domain transformation method and a singular value decomposition method to obtain complemented physiological feature data, inputting the complemented physiological feature data into a preset health state evaluation model, and calculating to obtain a health state evaluation result of the user in a preset period; by inputting the complemented physiological feature data obtained by complementing with the frequency domain transformation method and the singular value decomposition method into the preset health state assessment model, the health state assessment result of the user in the preset period is calculated, and the accuracy of health state assessment is improved.

Description

Health status assessment methods, devices, electronic equipment and storage media Technical Field

[0001] This application relates to the technical field of health status assessment, and more specifically, to a health status assessment method, apparatus, electronic device, and storage medium. Background Technology

[0002] With increasing public awareness of health, health status assessment technologies are playing an increasingly important role in disease prevention and guiding healthy lifestyles. However, in practical applications, especially in non-clinical scenarios, users often struggle to ensure continuous daily measurement of physiological characteristics, leading to significant gaps in the acquired data. For example, students may be unable to measure their physiological characteristics daily due to course schedules, personal habits, or simply forget to wear the device, resulting in missing data and a break in the data chain.

[0003] Current technologies typically employ simple interpolation methods, such as linear interpolation and mean interpolation, to handle this type of missing data. While these methods are easy to operate, their limitation lies in failing to fully consider the holistic and periodic nature of human physiological characteristics. Human physiological activities exhibit complex periodic patterns; for example, physiological indicators such as heart rate, blood pressure, and body temperature show specific fluctuation patterns within a day or a physiological cycle. Simple interpolation methods often ignore these inherent physiological rhythms and individual differences, leading to significant deviations between the imputed data and the true physiological state, thus affecting the accuracy of subsequent health status assessments.

[0004] Furthermore, physiological characteristic data is typically multi-dimensional and multimodal, including various physiological indicators such as heart rate, blood pressure, and sleep quality, and these indicators exhibit complex non-linear relationships. Simple interpolation methods struggle to capture these high-order complex relationships and cannot effectively utilize the intrinsic connections between different physiological characteristics to more accurately fill in missing data. Moreover, in the field of traditional Chinese medicine, data obtained through the "observation, auscultation, inquiry, and palpation" methods also exhibit complex non-linear relationships; simple interpolation methods similarly fail to capture the intrinsic connections between these data to accurately fill in missing data. Therefore, how to effectively complete discontinuously collected physiological characteristic data while taking into account the holistic and periodic nature of human physiology is a significant technical challenge currently facing the field of health status assessment.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] The purpose of this application is to provide a health status assessment method, device, electronic device, and storage medium. By inputting the completed physiological characteristic data obtained by using frequency domain transformation and singular value decomposition methods into a preset health status assessment model, the user's health status assessment result within a preset period is calculated. This solves the problem that existing physiological characteristic data completion methods fail to fully consider the integrity and periodicity of human physiological characteristic data, resulting in a large deviation between the completed data and the actual physiological state, thus affecting the accuracy of health status assessment. This method can effectively complete the missing physiological characteristic data and improve the accuracy of health status assessment.

[0007] In a first aspect, this application provides a health status assessment method, comprising: acquiring physiological characteristic data of a user within a preset period; constructing the physiological characteristic data into tensor data containing individual dimensions, feature dimensions, and time dimensions; using frequency domain transformation and singular value decomposition methods to complete the missing data in the tensor data to obtain completed physiological characteristic data; and inputting the completed physiological characteristic data into a preset health status assessment model to calculate the health status assessment result of the user within the preset period.

[0008] The health status assessment method provided in this application can assess health status by inputting the completed physiological characteristic data obtained by using frequency domain transformation and singular value decomposition methods into a preset health status assessment model, and calculating the user's health status assessment result within a preset period. This solves the problem that existing physiological characteristic data completion methods fail to fully consider the integrity and periodicity of human physiological characteristic data, resulting in a large deviation between the completed data and the actual physiological state, thus affecting the accuracy of health status assessment. This method can effectively complete the missing physiological characteristic data and improve the accuracy of health status assessment.

[0009] Optionally, the missing data in the tensor data is completed using frequency domain transformation and singular value decomposition methods to obtain completed physiological feature data. This includes: performing a frequency domain transformation on the tensor data based on the time dimension to represent the original time-domain signal of the tensor data in the frequency domain, obtaining a frequency domain tensor; performing data repair on the frequency domain tensor using the singular value decomposition method to complete the missing data, obtaining a completed frequency domain tensor; and performing an inverse frequency domain transformation on the completed frequency domain tensor based on the time dimension to obtain the completed physiological feature data.

[0010] The health status assessment method provided in this application can assess health status by converting time-domain data into frequency-domain data for processing. It effectively utilizes the advantages of frequency domain transformation in processing periodic data, better captures the periodic patterns of physiological characteristic data, and thus improves the accuracy of missing data completion.

[0011] Optionally, the frequency domain tensor is repaired using a singular value decomposition method to fill in missing data and obtain a repaired frequency domain tensor. This includes: extracting a complex value matrix for each frequency point from the frequency domain tensor; performing singular value decomposition on the complex value matrix to reconstruct and update the singular value elements in the complex value matrix to obtain a repaired frequency domain tensor; traversing the complex value matrices for all frequency points and summing all the repaired frequency domain tensors to obtain a repaired frequency domain tensor.

[0012] Optionally, singular value decomposition (SVD) is performed on the complex numerical matrix to reconstruct and update the singular value elements in the complex numerical matrix, thereby obtaining a repaired frequency domain tensor. This includes: performing SVD on the complex numerical matrix to obtain a SVD-decomposed complex numerical matrix; updating the corresponding singular value elements based on the relationship between each singular value element in the SVD-decomposed complex numerical matrix and a preset singular threshold to obtain updated singular value elements; and updating the SVD-decomposed complex numerical matrix using the updated singular value elements to obtain a repaired frequency domain tensor.

[0013] Optionally, based on the relationship between each singular value element in the complex numerical matrix after singular value decomposition and a preset singular threshold, the corresponding singular value elements are updated to obtain updated singular value elements. This includes: sequentially determining whether each singular value element in the complex numerical matrix after singular value decomposition is greater than the preset singular threshold; if so, calculating the difference between the corresponding singular value element and the preset singular threshold, and updating the corresponding singular value element to the difference to obtain updated singular value elements; if not, replacing the corresponding singular value element with zero to obtain updated singular value elements.

[0014] Optionally, traversing the complex value matrix of all frequency points and summing all the repaired frequency domain tensors to obtain the completed frequency domain tensor includes: traversing the complex value matrix of all frequency points to obtain the repaired frequency domain tensor corresponding to all frequency points; comparing the complex value matrix of all frequency points with the corresponding repaired frequency domain tensor to determine whether all singular value elements in the repaired frequency domain tensor have been updated; if so, summing all the repaired frequency domain tensors to obtain the completed frequency domain tensor; if not, comparing the size relationship between the unupdated singular value elements and a preset singular threshold to update the unupdated singular value elements, and then summing all the repaired frequency domain tensors to obtain the completed frequency domain tensor.

[0015] Optionally, inputting the completed physiological characteristic data into a preset health status assessment model to calculate the user's health status assessment result within a preset period includes: inputting the completed physiological characteristic data into the preset health status assessment model to obtain the user's preliminary health status assessment result within the preset period; calculating the corresponding information entropy based on the probability distribution of each assessment result in the preliminary health status assessment result; determining whether the information entropy is greater than a preset uncertainty threshold; if not, determining that the preliminary health status assessment result is reliable and determining the preliminary health status assessment result as the health status assessment result; if yes, calculating the difference between each assessment result to extract associated physiological characteristic questions from a preset question bank corresponding to assessment results with differences less than the preset assessment threshold, having the user answer the associated physiological characteristic questions, updating the corresponding physiological characteristic data based on the user's answer, and inputting the updated physiological characteristic data into the preset health status assessment model to repeatedly execute the health status assessment steps until the information entropy corresponding to the updated preliminary health status assessment result is less than or equal to the preset uncertainty threshold, and then determining the updated preliminary health status assessment result as the health status assessment result.

[0016] The health status assessment method provided in this application can assess health status. By introducing uncertainty assessment and interactive data update mechanism, when the uncertainty of the assessment results is high, the data can be further improved through user feedback, thereby improving the reliability and accuracy of health status assessment.

[0017] Secondly, this application provides a health status assessment device, comprising: an acquisition module for acquiring physiological characteristic data of a user within a preset period; a construction module for constructing the physiological characteristic data into tensor data containing individual dimensions, feature dimensions, and time dimensions; a completion module for completing missing data in the tensor data using frequency domain transformation and singular value decomposition methods to obtain completed physiological characteristic data; and a calculation module for inputting the completed physiological characteristic data into a preset health status assessment model to calculate the health status assessment result of the user within the preset period.

[0018] This health status assessment device calculates the user's health status assessment results within a preset period by inputting the completed physiological characteristic data obtained using frequency domain transformation and singular value decomposition methods into a preset health status assessment model. This solves the problem that existing physiological characteristic data completion methods fail to fully consider the integrity and periodicity of human physiological characteristic data, resulting in significant deviations between the completed data and the actual physiological state, thus affecting the accuracy of health status assessment. The device can effectively complete missing physiological characteristic data, thereby improving the accuracy of health status assessment.

[0019] Thirdly, this application provides an electronic device including a processor and a memory, the memory storing a computer program executable by the processor, wherein when the processor executes the computer program, it performs the steps in the health status assessment method described above.

[0020] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the health status assessment method described above.

[0021] Beneficial effects: The health status assessment method, device, electronic device, and storage medium provided in this application, by inputting the completed physiological characteristic data obtained by using frequency domain transformation and singular value decomposition methods into a preset health status assessment model, calculates the user's health status assessment results within a preset period. This solves the problem that existing physiological characteristic data completion methods fail to fully consider the integrity and periodicity of human physiological characteristic data, resulting in a large deviation between the completed data and the actual physiological state, thus affecting the accuracy of health status assessment. It can effectively complete the missing physiological characteristic data and improve the accuracy of health status assessment. Attached Figure Description

[0022] Figure 1 is a flowchart of the health status assessment method provided in the embodiments of this application.

[0023] Figure 2 is a schematic diagram of the health status assessment device provided in the embodiment of this application.

[0024] Figure 3 is a schematic diagram of the structure of the electronic device provided in the embodiment of this application.

[0025] Labeling Explanation: 1. Acquisition Module; 2. Construction Module; 3. Completion Module; 4. Calculation Module; 301. Processor; 302. Memory; 303. Communication Bus. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0027] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0028] Please refer to Figure 1, which illustrates a health status assessment method in some embodiments of this application. The method is used to assess health status and includes the following steps: Step S101, acquiring physiological characteristic data of a user within a preset period; Step S102, constructing tensor data containing individual, feature, and time dimensions from the physiological characteristic data; Step S103, using frequency domain transformation and singular value decomposition methods to complete missing data in the tensor data, obtaining completed physiological characteristic data; Step S104, inputting the completed physiological characteristic data into a preset health status assessment model to calculate the user's health status assessment result within the preset period.

[0029] This health status assessment method inputs the completed physiological characteristic data obtained by using frequency domain transformation and singular value decomposition methods into a preset health status assessment model to calculate the user's health status assessment results within a preset period. This solves the problem that existing physiological characteristic data completion methods fail to fully consider the integrity and periodicity of human physiological characteristic data, resulting in a large deviation between the completed data and the actual physiological state, thus affecting the accuracy of health status assessment. It can effectively complete the missing physiological characteristic data and improve the accuracy of health status assessment.

[0030] Specifically, in step S101, physiological characteristic data of the user within a preset period is acquired. This physiological characteristic data refers to various biological indicators that reflect the human body's health status. Specifically, it can include data from the four diagnostic methods of Traditional Chinese Medicine, encompassing multi-source heterogeneous data such as images, waveforms, sounds, and text, including tongue images, pulse data, questionnaires, and voice data. This data can be collected through various methods such as image capture, medical-grade pulse sensors, questionnaires, and voice input. Furthermore, physiological characteristic data can also include data such as heart rate, blood pressure, body temperature, blood oxygen saturation, sleep duration, and steps. This data can be collected through various methods such as wearable devices, smart home devices, or medical-grade sensors. Further, physiological characteristic data can also include non-invasive surface image data and periodic physiological fluctuation signal data, such as CT scans, magnetic resonance imaging, surface electromyography, electrocardiograms, and electroencephalograms. This data can be obtained through external professional equipment (such as specialized physical examinations at hospitals). When acquiring physiological characteristic data, at least one of the above data types should be acquired as physiological characteristic data. The more data acquired, the more accurate the final health status assessment result. The preset period refers to the time range used to assess a user's health status, such as one day, one week, one month, or one quarter, which can be set according to actual needs.

[0031] Specifically, in step S102, the physiological characteristic data is constructed into tensor data containing individual, feature, and time dimensions, which can better preserve the multidimensional correlation between data. The tensor data is a multidimensional array used to represent the physiological characteristic data; the individual dimension represents different users or individuals; the feature dimension represents different physiological characteristic indicators, such as heart rate and blood pressure; and the time dimension represents different time points within a preset period.

[0032] For example, suppose there are N users, each with M physiological characteristics (such as heart rate, blood pressure, and body temperature), and T time points within a preset period. This data can be organized into an N×M×T three-dimensional tensor. The first dimension represents the individual, the second represents the physiological characteristics, and the third represents time. This tensor structure can effectively represent the physiological state of different individuals and different characteristics at different time points, laying the foundation for subsequent missing data completion and health status assessment.

[0033] Specifically, in step S103, the missing data in the tensor data is completed using the frequency domain transformation method and the singular value decomposition method to obtain the completed physiological feature data. This includes: performing a frequency domain transformation on the tensor data based on the time dimension to represent the original time domain signal of the tensor data in the frequency domain, thereby obtaining a frequency domain tensor; performing data repair on the frequency domain tensor using the singular value decomposition method to complete the missing data, thereby obtaining the completed frequency domain tensor; and performing an inverse frequency domain transformation on the completed frequency domain tensor based on the time dimension to obtain the completed physiological feature data.

[0034] In step S103, physiological characteristic data typically exhibit periodicity or quasi-periodicity, and frequency domain transformation can effectively reveal these periodic components and their corresponding frequency information. Therefore, by performing frequency domain transformation in the time dimension, such as using the Fourier transform method, the time series data of each individual in each characteristic dimension can be converted to the frequency domain, thereby obtaining a frequency domain tensor. The purpose of this step is to transform the local missing data problem in the time domain signal into a global or local anomaly in the frequency domain, facilitating subsequent data repair.

[0035] In the frequency domain, the properties of singular value decomposition (SVD) can be used to fill in missing data. SVD is a powerful matrix factorization technique capable of revealing the main structure and underlying patterns in data. In the frequency domain, physiological characteristic data typically exhibits stronger low-rank properties, meaning it can be approximated with fewer singular values. When data is missing, iterative optimization or low-rank approximation can be used to infer and complete the missing frequency domain components using existing data points. This allows for efficient and accurate recovery of missing physiological characteristic data by leveraging the inherent structure of the frequency domain data.

[0036] By employing inverse frequency domain transformation, such as the inverse Fourier transform, the repaired and completed frequency domain tensor can be converted back into a time domain signal, yielding completed physiological characteristic data. This step provides complete and continuous physiological characteristic data in the time dimension, thus providing reliable input for subsequent health status assessment.

[0037] Specifically, in step S103, the frequency domain tensor is repaired by singular value decomposition to fill in the missing data and obtain the repaired frequency domain tensor. This includes: extracting the complex value matrix of each frequency point from the frequency domain tensor; performing singular value decomposition on the complex value matrix to reconstruct and update the singular value elements in the complex value matrix to obtain the repaired frequency domain tensor; traversing the complex value matrix of all frequency points and summarizing all repaired frequency domain tensors to obtain the repaired frequency domain tensor.

[0038] In step S103, the frequency domain tensor obtained after frequency domain transformation is decomposed according to its frequency dimension to obtain a two-dimensional complex value matrix corresponding to each specific frequency point (i.e., the complex value matrix of each frequency point). This decomposes the high-dimensional frequency domain tensor into a lower-dimensional matrix that is easier to process, so that independent singular value decomposition processing can be performed on the characteristics of each frequency point in the future.

[0039] Specifically, in step S103, singular value decomposition is performed on the complex numerical matrix to reconstruct and update the singular value elements in the complex numerical matrix, thereby obtaining the repaired frequency domain tensor. This includes: performing singular value decomposition on the complex numerical matrix to obtain a complex numerical matrix after singular value decomposition; updating the corresponding singular value elements based on the relationship between each singular value element in the complex numerical matrix after singular value decomposition and a preset singular threshold to obtain updated singular value elements; and updating the complex numerical matrix after singular value decomposition using the updated singular value elements to obtain the repaired frequency domain tensor.

[0040] In step S103, the complex numerical matrix is ​​decomposed into the product of three matrices, i.e., S = UMV. T In this matrix, S is the complex numerical matrix after singular value decomposition, M is a diagonal matrix whose diagonal elements are the singular values ​​(i.e., singular value elements), U is a multi-order orthogonal matrix, V is a 3rd-order orthogonal matrix, and the superscript T is the transpose symbol. The singular value elements in the diagonal matrix M reflect the importance or energy distribution of the original data in different directions.

[0041] Specifically, in step S103, based on the relationship between each singular value element in the complex numerical matrix after singular value decomposition and a preset singular threshold, the corresponding singular value elements are updated to obtain the updated singular value elements. This includes: sequentially determining whether each singular value element in the complex numerical matrix after singular value decomposition is greater than the preset singular threshold; if so, calculating the difference between the corresponding singular value element and the preset singular threshold, and updating the corresponding singular value element to the difference to obtain the updated singular value elements; if not, replacing the corresponding singular value element with zero to obtain the updated singular value elements.

[0042] In step S103, when updating the singular value elements, a preset singularity threshold is introduced to distinguish different components in the data. First, each singular value element in the complex numerical matrix after singular value decomposition is traversed, and each element is sequentially checked against the preset singularity threshold. If the check is positive, it indicates that the singular value element represents important information in the data. In this case, the difference between the singular value element and the preset singularity threshold is calculated, and the singular value element is updated to the difference, thereby preserving important information and performing a certain degree of noise reduction. If the check is negative, meaning the singular value element is less than or equal to the preset singularity threshold, it is considered that the singular value element may represent noise or unimportant information in the data. In this case, the corresponding singular value element is directly replaced with zero, thereby directly removing noise or unimportant information.

[0043] The updated singular value elements are added to the complex value matrix after singular value decomposition to denoise and complete the original data in the frequency domain, reconstructing the repaired complex value matrix and obtaining the repaired frequency domain tensor. Specifically, in step S103, the complex value matrices of all frequency points are traversed, and all repaired frequency domain tensors are summarized to obtain the completed frequency domain tensor. This includes: traversing the complex value matrices of all frequency points to obtain the repaired frequency domain tensors corresponding to all frequency points; comparing the complex value matrices of all frequency points with the corresponding repaired frequency domain tensors to determine whether all singular value elements in the repaired frequency domain tensors have been updated; if so, summarizing all repaired frequency domain tensors to obtain the completed frequency domain tensor; if not, comparing the size relationship between the unupdated singular value elements and the preset singular threshold, updating the unupdated singular value elements, and summarizing all repaired frequency domain tensors to obtain the completed frequency domain tensor.

[0044] In step S103, the complex value matrix corresponding to each frequency point is processed one by one, and the frequency domain tensor after repair for each frequency point is generated according to the above singular value decomposition and update rules.

[0045] The complex value matrices of all frequency points are compared with the corresponding repaired frequency domain tensors to determine whether all singular value elements in the repaired frequency domain tensors have been updated. Once it is confirmed that all singular value elements have been updated, the repaired frequency domain tensors corresponding to all frequency points are integrated to form the final completed frequency domain tensor. If any unupdated singular value elements are found, the update logic is executed again, that is, its relationship with a preset singularity threshold is judged and adjusted (e.g., replaced with zero or the difference is calculated) until all singular value elements are processed. Then, the results are summarized to form the final completed frequency domain tensor.

[0046] Before summarizing the repaired frequency domain tensor, a comparison and judgment mechanism for the update status of all singular value elements is introduced to ensure that each singular value element is optimized strictly according to the preset singular threshold. This ensures that all missing data in the frequency domain tensor is fully and accurately filled in before the inverse frequency domain transformation, significantly improving the integrity and reliability of data repair.

[0047] Specifically, in step S104, the completed physiological characteristic data is input into a preset health status assessment model to calculate the user's health status assessment result within a preset period. This includes: inputting the completed physiological characteristic data into the preset health status assessment model to obtain the user's preliminary health status assessment result within the preset period; calculating the corresponding information entropy based on the probability distribution of each assessment result in the preliminary health status assessment result; determining whether the information entropy is greater than a preset uncertainty threshold; if not, determining that the preliminary health status assessment result is reliable and determining the preliminary health status assessment result as the health status assessment result; if so, calculating the difference between each assessment result to extract the associated physiological characteristic questions corresponding to assessment results with differences less than the preset assessment threshold from a preset question bank, having the user answer the associated physiological characteristic questions, updating the corresponding physiological characteristic data based on the user's answer, and inputting the updated physiological characteristic data into the preset health status assessment model to repeat the health status assessment steps until the information entropy corresponding to the updated preliminary health status assessment result is less than or equal to the preset uncertainty threshold, and then determining the updated preliminary health status assessment result as the health status assessment result.

[0048] In step S104, the completed physiological characteristic data is input into a preset health status assessment model to obtain the user's preliminary health status assessment results within a preset period. The preliminary health status assessment results refer to the initial assessment output (or assessment output) given by the health status assessment model based on the completed physiological characteristic data, along with corresponding intervention measures. Specifically, this includes the prediction confidence or probability distribution (i.e., the probability distribution of assessment results) for different syndromes (i.e., each assessment result, such as liver fire excess, spleen yang deficiency, kidney yang deficiency, etc.). For example, the preliminary health status assessment results might show a probability of 0.45 for "spleen yang deficiency" and 0.42 for "kidney yang deficiency," with intervention measures including traditional Chinese medicine such as astragalus and angelica.

[0049] The health status assessment model is a pre-trained hypergraph neural network model (or graph neural network model). The hypergraph neural network model constructs a hypergraph structure through nodes and hyperedges. Nodes represent discrete entities, including physiological characteristic data, syndromes, and interventions; the data in a node is represented in vector form. Hyperedges represent higher-order associations between entities. A hyperedge can connect any number of nodes, i.e., it can associate multiple entities, thus forming a complete "syndrome (physiological characteristic data) - symptom (syndrome) - medicine (intervention)" relationship chain. For example, a hyperedge associating "pale tongue - weak pulse - aversion to cold - spleen yang deficiency - astragalus" indicates that the physiological characteristic data are a pale tongue and weak pulse, the syndromes are aversion to cold and spleen yang deficiency, and astragalus is recommended as a reference intervention. The hypergraph structure represents the relationship chain formed by nodes and hyperedges. Each hypergraph structure corresponds to one relationship chain. According to professional domain knowledge (such as doctors' opinions and medical books), various corresponding nodes are connected using hyperedges to form multiple hypergraph structures.

[0050] The hypergraph neural network model includes a hypergraph convolutional network, which contains multiple hypergraph convolutional layers. These layers learn the relationships between nodes to form corresponding hyperedges, thereby constructing the corresponding hypergraph structure. The core operation of the hypergraph convolutional layer is divided into two stages: (1) The features of a hyperedge are aggregated from the features of all the nodes it connects to, specifically: ;in, Let e ​​be the output feature of the hyperedge e in the i-th hypergraph convolutional layer, i.e., the hypergraph structure; 'e' is a node; 'e' is a hyperedge; For nodes The entity set in the i-th hypergraph convolutional layer, after computation in each hypergraph convolutional layer, consists of nodes. Based on the initial entity, the entities of all nodes within the corresponding hyperedge are aggregated to form the entity set. For example, after the calculation of the hypergraph convolution layer, the "spleen yang deficiency" node receives information from directly related symptom nodes such as "pale tongue" and "weak pulse", forming the entity set "spleen yang deficiency-pale tongue-weak pulse".

[0051] (2) A new feature of a node is formed by aggregating all the hyperedge features that contain that node, specifically: ;in, For nodes The set of entities in the (i+1)th hypergraph convolutional layer; It is a non-linear activation function; Let E be the learnable parameter matrix in the i-th hypergraph convolutional layer; E is the set of hyperedges corresponding to hyperedge e, i.e., the corresponding medical knowledge.

[0052] In summary, in a hypergraph convolutional network, each hypergraph convolutional layer receives the output features from the previous layer and calculates the output features of the current layer using the formula described above. Thus, by stacking multiple hypergraph convolutional layers, the hypergraph neural network model can achieve multi-hop propagation of information and higher-order inference on the hypergraph, ultimately using the final output features of the last layer to determine the health status.

[0053] In the hypergraph neural network model, a classifier (such as a fully connected layer + softmax function) is also set up to calculate its probability distribution. The final output features of the hypergraph convolutional network are input into the classifier to determine the health status, obtaining the corresponding evaluation results, the probability distribution of the evaluation results, and intervention measures, thus forming the final health status evaluation result. The evaluation result with the highest probability distribution is the current primary health status. Specifically, the classifier converts the output features into corresponding scores through a fully connected layer. ;in, This is the score for the k-th syndrome (evaluation result) in the final output feature; This is the feature vector corresponding to the kth syndrome in the final output features; is the first learnable parameter of the classifier; This is the second learnable parameter of the classifier.

[0054] The classifier uses the Softmax function to convert the result into a corresponding probability distribution, specifically: ;in, This represents the probability distribution of the kth syndrome in the final output features (the probability distribution of the evaluation result). It is a natural exponential function; Let be the score of the j-th syndrome in the final output feature; K is the total number of syndromes in the output feature.

[0055] Therefore, the final health status assessment result is obtained by calculating through a classifier.

[0056] In some alternative embodiments, the health status assessment model can also be a pre-trained machine learning model that can output the user's health status assessment results based on the input physiological characteristic data. It can be constructed based on various algorithms such as deep learning, support vector machine, and decision tree, and trained using historical physiological characteristic data of multiple users in the database and corresponding historical health status assessment results (the historical health status assessment results are determined by senior doctors or existing health status assessment criteria).

[0057] Based on the probability distribution of each assessment result in the preliminary health status assessment, the corresponding information entropy is calculated to quantify the degree of uncertainty in the assessment results. Information entropy is an indicator that measures the uncertainty or randomness of this probability distribution; the higher the information entropy value, the higher the uncertainty of the assessment result, and the more ambiguous the model's judgment on the final health status.

[0058] The specific formula for calculating information entropy is as follows: ;in, Information entropy; Let be the probability distribution of the k-th evaluation result; m is the total number of evaluation results; k is the k-th evaluation result, k≤m.

[0059] The system determines the reliability of the preliminary health status assessment by checking if the information entropy exceeds a preset uncertainty threshold. If the information entropy is less than or equal to the preset uncertainty threshold, the preliminary health status assessment is considered reliable and is accepted as the final health status assessment. If the information entropy exceeds the preset uncertainty threshold, the preliminary health status assessment is considered unreliable and is not directly adopted. Instead, the differences between the assessment results are calculated, and those with differences less than the preset assessment threshold are extracted. Related physiological characteristic questions are then retrieved from a preset question bank. By having users answer these questions, more specific and personalized physiological characteristic data is obtained, updating the original physiological characteristic data. The updated physiological characteristic data is then re-input into the health status assessment model, and the above assessment steps are repeated until the information entropy corresponding to the new preliminary health status assessment result is less than or equal to the preset uncertainty threshold, ensuring the final health status assessment result has high reliability. The preset uncertainty threshold and preset assessment threshold can be set according to actual needs.

[0060] By introducing information entropy as a reliability criterion and combining it with a user-interactive question-and-answer mechanism, dynamic updates of physiological characteristic data and iterative optimization of the assessment process were achieved. This not only significantly improved the accuracy and reliability of health status assessment results, avoiding misjudgments caused by fuzzy model outputs, but also enhanced the intelligence and personalization of the assessment process, enabling users to obtain more accurate health status feedback and providing a solid foundation for subsequent health management and intervention.

[0061] For example, suppose a user's physiological characteristic data within a preset period is completed and input into a preset health status assessment model. The model's preliminary health status assessment results show that the user has a probability of "spleen yang deficiency" of 0.45, a probability of "kidney yang deficiency" of 0.42, and a probability of "liver fire excess" of 0.13. Because the probability distributions of the assessment results "Spleen Yang Deficiency" and "Kidney Yang Deficiency" are relatively similar, the calculated information entropy is high, and this information entropy is greater than the preset uncertainty threshold. Therefore, "Spleen Yang Deficiency" and "Kidney Yang Deficiency" will not be directly used as the final assessment results. Instead, the differences between these assessment results will be calculated. For example, the probability difference between "Spleen Yang Deficiency" and "Kidney Yang Deficiency" is 0.03, the probability difference between "Spleen Yang Deficiency" and "Liver Fire Excess" is 0.32, and the probability difference between "Kidney Yang Deficiency" and "Liver Fire Excess" is 0.29. Since the probability difference between "Spleen Yang Deficiency" and "Kidney Yang Deficiency" is less than the preset assessment threshold (e.g., 5%), related physiological characteristic questions associated with the states of "Spleen Yang Deficiency" and "Kidney Yang Deficiency" will be extracted from the preset question bank, such as "Have you had frequent urination at night in the past week?" and "Have your stools been formed in the past week?".

[0062] After users answer these questions based on their own circumstances, their physiological characteristic data will be updated according to their answers. For example, if a user answers that they have experienced frequent urination at night and loose stools in the past week, indicators related to "frequent urination at night" and "loose stools" will be added or adjusted in their physiological characteristic data. Subsequently, the updated physiological characteristic data will be re-input into the health status assessment model for evaluation.

[0063] After this iteration, the model may output new preliminary health status assessment results, such as a probability of "spleen yang deficiency" of 0.75, a probability of "kidney yang deficiency" of 0.22, and a probability of "liver fire excess" of 0.03. At this point, the information entropy is recalculated, and its value is found to be less than or equal to the preset uncertainty threshold. Therefore, the preliminary health status assessment result is determined to be reliable, and "spleen yang deficiency" is identified as the primary assessment result for the user's health status within the preset period. Through this iteration and interaction, the reliability of the final assessment result is ensured.

[0064] As shown above, this health status assessment method acquires the user's physiological characteristic data within a preset period, constructs this data into tensor data containing individual, characteristic, and time dimensions, and uses frequency domain transformation and singular value decomposition methods to complete the missing data in the tensor data, obtaining the completed physiological characteristic data. This completed physiological characteristic data is then input into a preset health status assessment model to calculate the user's health status assessment result within the preset period. Therefore, by inputting the completed physiological characteristic data obtained using frequency domain transformation and singular value decomposition into the preset health status assessment model, the method calculates the user's health status assessment result within the preset period. This addresses the problem that existing methods for completing missing physiological characteristic data fail to fully consider the integrity and periodicity of human physiological characteristic data, leading to significant deviations between the completed data and the actual physiological state, thus affecting the accuracy of health status assessment. This method effectively completes missing physiological characteristic data, improving the accuracy of health status assessment.

[0065] Referring to Figure 2, this application provides a health status assessment device for assessing health status, comprising: an acquisition module 1 for acquiring physiological characteristic data of a user within a preset period; a construction module 2 for constructing tensor data containing individual, feature, and time dimensions from the physiological characteristic data; a completion module 3 for completing missing data in the tensor data using frequency domain transformation and singular value decomposition methods to obtain completed physiological characteristic data; and a calculation module 4 for inputting the completed physiological characteristic data into a preset health status assessment model to calculate the user's health status assessment result within the preset period.

[0066] This health status assessment device calculates the user's health status assessment results within a preset period by inputting the completed physiological characteristic data obtained using frequency domain transformation and singular value decomposition methods into a preset health status assessment model. This solves the problem that existing physiological characteristic data completion methods fail to fully consider the integrity and periodicity of human physiological characteristic data, resulting in significant deviations between the completed data and the actual physiological state, thus affecting the accuracy of health status assessment. The device can effectively complete missing physiological characteristic data, thereby improving the accuracy of health status assessment.

[0067] Specifically, when module 1 is executed, it acquires the user's physiological characteristic data within a preset period. This physiological characteristic data refers to various biological indicators that reflect a person's health status. Specifically, it can include data from the four diagnostic methods of Traditional Chinese Medicine, encompassing multi-source heterogeneous data such as images, waveforms, sounds, and text. Examples include tongue images, pulse data, questionnaires, and voice data. This data can be collected through various methods such as image capture, medical-grade pulse sensors, questionnaires, and voice input. Furthermore, physiological characteristic data can also include heart rate, blood pressure, body temperature, blood oxygen saturation, sleep duration, and step count. This data can be collected through wearable devices, smart home devices, or medical-grade sensors. Further, physiological characteristic data can also include non-invasive surface image data and periodic physiological fluctuation signal data, such as CT scans, magnetic resonance imaging (MRI), surface electromyography (EMG), electrocardiograms (ECG), and electroencephalograms (EEG). This data can be obtained through external professional equipment (such as specialized medical examinations at hospitals). When acquiring physiological characteristic data, at least one of the above data types should be acquired. The more data acquired, the more accurate the final health status assessment result. The preset period refers to the time range used to assess a user's health status, such as one day, one week, one month, or one quarter, which can be set according to actual needs.

[0068] Specifically, during execution, module 2 constructs physiological characteristic data into tensor data containing individual, feature, and time dimensions, which can better preserve the multidimensional correlation between data. Tensor data is a multidimensional array used to represent physiological characteristic data; the individual dimension represents different users or individuals; the feature dimension represents different physiological characteristic indicators, such as heart rate and blood pressure; and the time dimension represents different time points within a preset period.

[0069] For example, suppose there are N users, each with M physiological characteristics (such as heart rate, blood pressure, and body temperature), and T time points within a preset period. This data can be organized into an N×M×T three-dimensional tensor. The first dimension represents the individual, the second represents the physiological characteristics, and the third represents time. This tensor structure can effectively represent the physiological state of different individuals and different characteristics at different time points, laying the foundation for subsequent missing data completion and health status assessment.

[0070] Specifically, when the completion module 3 uses frequency domain transformation and singular value decomposition to complete the missing data in the tensor data and obtain the completed physiological feature data, it performs the following steps: Based on the time dimension, it performs frequency domain transformation on the tensor data to represent the original time domain signal of the tensor data in the frequency domain, obtaining a frequency domain tensor; through singular value decomposition, it performs data repair on the frequency domain tensor to complete the missing data, obtaining the completed frequency domain tensor; based on the time dimension, it performs inverse frequency domain transformation on the completed frequency domain tensor to obtain the completed physiological feature data.

[0071] When completing module 3 is executed, physiological characteristic data often exhibits periodicity or quasi-periodicity. Frequency domain transformation can effectively reveal these periodic components and their corresponding frequency information. Therefore, by performing frequency domain transformation in the time dimension, such as using the Fourier transform method, the time series data of each individual in each feature dimension can be converted to the frequency domain, thus obtaining a frequency domain tensor. The purpose of this step is to transform the local missing data in the time domain signal into global or local anomalies in the frequency domain, facilitating subsequent data repair.

[0072] In the frequency domain, the properties of singular value decomposition (SVD) can be used to fill in missing data. SVD is a powerful matrix factorization technique capable of revealing the main structure and underlying patterns in data. In the frequency domain, physiological characteristic data typically exhibits stronger low-rank properties, meaning it can be approximated with fewer singular values. When data is missing, iterative optimization or low-rank approximation can be used to infer and complete the missing frequency domain components using existing data points. This allows for efficient and accurate recovery of missing physiological characteristic data by leveraging the inherent structure of the frequency domain data.

[0073] By employing inverse frequency domain transformation, such as the inverse Fourier transform, the repaired and completed frequency domain tensor can be converted back into a time domain signal, yielding completed physiological characteristic data. This step provides complete and continuous physiological characteristic data in the time dimension, thus providing reliable input for subsequent health status assessment.

[0074] Specifically, when the completion module 3 repairs the frequency domain tensor using the singular value decomposition method to complete the missing data and obtain the completed frequency domain tensor, it performs the following steps: extracting the complex value matrix for each frequency point from the frequency domain tensor; performing singular value decomposition on the complex value matrix to reconstruct and update the singular value elements in the complex value matrix to obtain the repaired frequency domain tensor; traversing the complex value matrix of all frequency points and summarizing all repaired frequency domain tensors to obtain the completed frequency domain tensor.

[0075] When the completion module 3 is executed, it decomposes the frequency domain tensor obtained after frequency domain transformation according to its frequency dimension, thereby obtaining a two-dimensional complex value matrix corresponding to each specific frequency point (i.e., the complex value matrix of each frequency point). This decomposes the high-dimensional frequency domain tensor into a lower-dimensional matrix that is easier to process, so that the singular value decomposition process can be performed independently for the characteristics of each frequency point.

[0076] Specifically, when the completion module 3 performs singular value decomposition on the complex numerical matrix to reconstruct and update the singular value elements in the complex numerical matrix and obtain the repaired frequency domain tensor, it performs the following steps: performing singular value decomposition on the complex numerical matrix to obtain the complex numerical matrix after singular value decomposition; updating the corresponding singular value elements based on the size relationship between each singular value element in the complex numerical matrix after singular value decomposition and the preset singular threshold to obtain the updated singular value elements; and updating the complex numerical matrix after singular value decomposition using the updated singular value elements to obtain the repaired frequency domain tensor.

[0077] When the completion module 3 is executed, it decomposes the complex numerical matrix into the product of three matrices, i.e., S=UMV. T In this matrix, S is the complex numerical matrix after singular value decomposition, M is a diagonal matrix whose diagonal elements are the singular values ​​(i.e., singular value elements), U is a multi-order orthogonal matrix, V is a 3rd-order orthogonal matrix, and the superscript T is the transpose symbol. The singular value elements in the diagonal matrix M reflect the importance or energy distribution of the original data in different directions.

[0078] Specifically, the completion module 3 updates the corresponding singular value elements based on the relationship between each singular value element in the complex numerical matrix after singular value decomposition and a preset singular threshold. When obtaining the updated singular value elements, it performs the following steps: sequentially determines whether each singular value element in the complex numerical matrix after singular value decomposition is greater than the preset singular threshold; if so, it calculates the difference between the corresponding singular value element and the preset singular threshold, and updates the corresponding singular value element to the difference to obtain the updated singular value element; if not, it replaces the corresponding singular value element with zero to obtain the updated singular value element.

[0079] When the completion module 3 is executed, it introduces a preset singularity threshold to distinguish different components in the data when updating singular value elements. First, it iterates through each singular value element in the complex numerical matrix after singular value decomposition and checks whether it is greater than the preset singularity threshold. If the result is yes, it indicates that the singular value element represents important information in the data. In this case, the difference between the singular value element and the preset singularity threshold is calculated, and the singular value element is updated to the difference, thereby preserving important information and performing a certain degree of noise reduction. If the result is no, that is, the singular value element is less than or equal to the preset singularity threshold, it is considered that the singular value element may represent noise or unimportant information in the data. In this case, the corresponding singular value element is directly replaced with zero, thereby directly removing noise or unimportant information.

[0080] The updated singular value elements are added to the complex value matrix after singular value decomposition to denoise and complete the original data in the frequency domain, reconstructing the repaired complex value matrix and obtaining the repaired frequency domain tensor. Specifically, when the completion module 3 traverses the complex value matrix of all frequency points and summarizes all repaired frequency domain tensors to obtain the repaired frequency domain tensor, it performs the following: traversing the complex value matrix of all frequency points to obtain the repaired frequency domain tensor corresponding to all frequency points; comparing the complex value matrix of all frequency points with the corresponding repaired frequency domain tensor to determine whether all singular value elements in the repaired frequency domain tensor have been updated; if so, summarizing all repaired frequency domain tensors to obtain the repaired frequency domain tensor; if not, comparing the size relationship between the unupdated singular value elements and the preset singular threshold, updating the unupdated singular value elements, and summarizing all repaired frequency domain tensors to obtain the repaired frequency domain tensor.

[0081] When the completion module 3 is executed, it processes the complex value matrix corresponding to each frequency point one by one, and generates the repaired frequency domain tensor for each frequency point according to the above singular value decomposition and update rules.

[0082] The complex value matrices of all frequency points are compared with the corresponding repaired frequency domain tensors to determine whether all singular value elements in the repaired frequency domain tensors have been updated. Once it is confirmed that all singular value elements have been updated, the repaired frequency domain tensors corresponding to all frequency points are integrated to form the final completed frequency domain tensor. If any unupdated singular value elements are found, the update logic is executed again, that is, its relationship with a preset singularity threshold is judged and adjusted (e.g., replaced with zero or the difference is calculated) until all singular value elements are processed. Then, the results are summarized to form the final completed frequency domain tensor.

[0083] Before summarizing the repaired frequency domain tensor, a comparison and judgment mechanism for the update status of all singular value elements is introduced to ensure that each singular value element is optimized strictly according to the preset singular threshold. This ensures that all missing data in the frequency domain tensor is fully and accurately filled in before the inverse frequency domain transformation, significantly improving the integrity and reliability of data repair.

[0084] Specifically, when the calculation module 4 inputs the completed physiological characteristic data into the preset health status assessment model to calculate the user's health status assessment result within a preset period, it performs the following steps: inputting the completed physiological characteristic data into the preset health status assessment model to obtain the user's preliminary health status assessment result within the preset period; calculating the corresponding information entropy based on the probability distribution of each assessment result in the preliminary health status assessment result; determining whether the information entropy is greater than a preset uncertainty threshold; if not, determining that the preliminary health status assessment result is reliable and determining the preliminary health status assessment result as the health status assessment result; if so, calculating the difference between each assessment result to extract the associated physiological characteristic questions corresponding to assessment results with differences less than the preset assessment threshold from the preset question bank, having the user answer the associated physiological characteristic questions, updating the corresponding physiological characteristic data based on the user's answer, and inputting the updated physiological characteristic data into the preset health status assessment model to repeat the health status assessment steps until the information entropy corresponding to the updated preliminary health status assessment result is less than or equal to the preset uncertainty threshold, and then determining the updated preliminary health status assessment result as the health status assessment result.

[0085] When the calculation module 4 is executed, it inputs the completed physiological characteristic data into the preset health status assessment model to obtain the user's preliminary health status assessment results within a preset period. The preliminary health status assessment results refer to the initial assessment output (or assessment output) given by the health status assessment model based on the completed physiological characteristic data, along with corresponding intervention measures. Specifically, this includes the prediction confidence or probability distribution (i.e., the probability distribution of assessment results) for different syndromes (i.e., each assessment result, such as liver fire excess, spleen yang deficiency, kidney yang deficiency, etc.). For example, the preliminary health status assessment results might show a probability of 0.45 for "spleen yang deficiency" and 0.42 for "kidney yang deficiency," with intervention measures including traditional Chinese medicine such as astragalus and angelica.

[0086] The health status assessment model is a pre-trained hypergraph neural network model (or graph neural network model). The hypergraph neural network model constructs a hypergraph structure through nodes and hyperedges. Nodes represent discrete entities, including physiological characteristic data, syndromes, and interventions; the data in a node is represented in vector form. Hyperedges represent higher-order associations between entities. A hyperedge can connect any number of nodes, i.e., it can associate multiple entities, thus forming a complete "syndrome (physiological characteristic data) - symptom (syndrome) - medicine (intervention)" relationship chain. For example, a hyperedge associating "pale tongue - weak pulse - aversion to cold - spleen yang deficiency - astragalus" indicates that the physiological characteristic data are a pale tongue and weak pulse, the syndromes are aversion to cold and spleen yang deficiency, and astragalus is recommended as a reference intervention. The hypergraph structure represents the relationship chain formed by nodes and hyperedges. Each hypergraph structure corresponds to one relationship chain. According to professional domain knowledge (such as doctors' opinions and medical books), various corresponding nodes are connected using hyperedges to form multiple hypergraph structures.

[0087] The hypergraph neural network model includes a hypergraph convolutional network, which contains multiple hypergraph convolutional layers. These layers learn the relationships between nodes to form corresponding hyperedges, thereby constructing the corresponding hypergraph structure. The core operation of the hypergraph convolutional layer is divided into two stages: (1) The features of a hyperedge are aggregated from the features of all the nodes it connects to, specifically: ;in, Let e ​​be the output feature of the hyperedge e in the i-th hypergraph convolutional layer, i.e., the hypergraph structure; 'e' is a node; 'e' is a hyperedge; For nodes The entity set in the i-th hypergraph convolutional layer, after computation in each hypergraph convolutional layer, consists of nodes. Based on the initial entity, the entities of all nodes within the corresponding hyperedge are aggregated to form the entity set. For example, after the calculation of the hypergraph convolution layer, the "spleen yang deficiency" node receives information from directly related symptom nodes such as "pale tongue" and "weak pulse", forming the entity set "spleen yang deficiency-pale tongue-weak pulse".

[0088] (2) A new feature of a node is formed by aggregating all the hyperedge features that contain that node, specifically: ;in, For nodes The set of entities in the (i+1)th hypergraph convolutional layer; It is a non-linear activation function; Let E be the learnable parameter matrix in the i-th hypergraph convolutional layer; E is the set of hyperedges corresponding to hyperedge e, i.e., the corresponding medical knowledge.

[0089] In summary, in a hypergraph convolutional network, each hypergraph convolutional layer receives the output features from the previous layer and calculates the output features of the current layer using the formula described above. Thus, by stacking multiple hypergraph convolutional layers, the hypergraph neural network model can achieve multi-hop propagation of information and higher-order inference on the hypergraph, ultimately using the final output features of the last layer to determine the health status.

[0090] In the hypergraph neural network model, a classifier (such as a fully connected layer + softmax function) is also set up to calculate its probability distribution. The final output features of the hypergraph convolutional network are input into the classifier to determine the health status, obtaining the corresponding evaluation results, the probability distribution of the evaluation results, and intervention measures, thus forming the final health status evaluation result. The evaluation result with the highest probability distribution is the current primary health status. Specifically, the classifier converts the output features into corresponding scores through a fully connected layer. ;in, This is the score for the k-th syndrome (evaluation result) in the final output feature; This is the feature vector corresponding to the kth syndrome in the final output features; is the first learnable parameter of the classifier; This is the second learnable parameter of the classifier.

[0091] The classifier uses the Softmax function to convert the result into a corresponding probability distribution, specifically: ;in, This represents the probability distribution of the kth syndrome in the final output features (the probability distribution of the evaluation result). It is a natural exponential function; Let be the score of the j-th syndrome in the final output feature; K is the total number of syndromes in the output feature.

[0092] Therefore, the final health status assessment result is obtained by calculating through a classifier.

[0093] In some alternative embodiments, the health status assessment model can also be a pre-trained machine learning model that can output the user's health status assessment results based on the input physiological characteristic data. It can be constructed based on various algorithms such as deep learning, support vector machine, and decision tree, and trained using historical physiological characteristic data of multiple users in the database and corresponding historical health status assessment results (the historical health status assessment results are determined by senior doctors or existing health status assessment criteria).

[0094] Based on the probability distribution of each assessment result in the preliminary health status assessment, the corresponding information entropy is calculated to quantify the degree of uncertainty in the assessment results. Information entropy is an indicator that measures the uncertainty or randomness of this probability distribution; the higher the information entropy value, the higher the uncertainty of the assessment result, and the more ambiguous the model's judgment on the final health status.

[0095] The specific formula for calculating information entropy is as follows: ;in, Information entropy; Let be the probability distribution of the k-th evaluation result; m is the total number of evaluation results; k is the k-th evaluation result, k≤m.

[0096] The system determines the reliability of the preliminary health status assessment by checking if the information entropy exceeds a preset uncertainty threshold. If the information entropy is less than or equal to the preset uncertainty threshold, the preliminary health status assessment is considered reliable and is accepted as the final health status assessment. If the information entropy exceeds the preset uncertainty threshold, the preliminary health status assessment is considered unreliable and is not directly adopted. Instead, the differences between the assessment results are calculated, and those with differences less than the preset assessment threshold are extracted. Related physiological characteristic questions are then retrieved from a preset question bank. By having users answer these questions, more specific and personalized physiological characteristic data is obtained, updating the original physiological characteristic data. The updated physiological characteristic data is then re-input into the health status assessment model, and the above assessment steps are repeated until the information entropy corresponding to the new preliminary health status assessment result is less than or equal to the preset uncertainty threshold, ensuring the final health status assessment result has high reliability. The preset uncertainty threshold and preset assessment threshold can be set according to actual needs.

[0097] By introducing information entropy as a reliability criterion and combining it with a user-interactive question-and-answer mechanism, dynamic updates of physiological characteristic data and iterative optimization of the assessment process were achieved. This not only significantly improved the accuracy and reliability of health status assessment results, avoiding misjudgments caused by fuzzy model outputs, but also enhanced the intelligence and personalization of the assessment process, enabling users to obtain more accurate health status feedback and providing a solid foundation for subsequent health management and intervention.

[0098] For example, suppose a user's physiological characteristic data within a preset period is completed and input into a preset health status assessment model. The model's preliminary health status assessment results show that the user has a probability of "spleen yang deficiency" of 0.45, a probability of "kidney yang deficiency" of 0.42, and a probability of "liver fire excess" of 0.13. Because the probability distributions of the assessment results "Spleen Yang Deficiency" and "Kidney Yang Deficiency" are relatively similar, the calculated information entropy is high, and this information entropy is greater than the preset uncertainty threshold. Therefore, "Spleen Yang Deficiency" and "Kidney Yang Deficiency" will not be directly used as the final assessment results. Instead, the differences between these assessment results will be calculated. For example, the probability difference between "Spleen Yang Deficiency" and "Kidney Yang Deficiency" is 0.03, the probability difference between "Spleen Yang Deficiency" and "Liver Fire Excess" is 0.32, and the probability difference between "Kidney Yang Deficiency" and "Liver Fire Excess" is 0.29. Since the probability difference between "Spleen Yang Deficiency" and "Kidney Yang Deficiency" is less than the preset assessment threshold (e.g., 5%), related physiological characteristic questions associated with the states of "Spleen Yang Deficiency" and "Kidney Yang Deficiency" will be extracted from the preset question bank, such as "Have you had frequent urination at night in the past week?" and "Have your stools been formed in the past week?".

[0099] After users answer these questions based on their own circumstances, their physiological characteristic data will be updated according to their answers. For example, if a user answers that they have experienced frequent urination at night and loose stools in the past week, indicators related to "frequent urination at night" and "loose stools" will be added or adjusted in their physiological characteristic data. Subsequently, the updated physiological characteristic data will be re-input into the health status assessment model for evaluation.

[0100] After this iteration, the model may output new preliminary health status assessment results, such as a probability of "spleen yang deficiency" of 0.75, a probability of "kidney yang deficiency" of 0.22, and a probability of "liver fire excess" of 0.03. At this point, the information entropy is recalculated, and its value is found to be less than or equal to the preset uncertainty threshold. Therefore, the preliminary health status assessment result is determined to be reliable, and "spleen yang deficiency" is identified as the primary assessment result for the user's health status within the preset period. Through this iteration and interaction, the reliability of the final assessment result is ensured.

[0101] As shown above, this health status assessment device acquires the user's physiological characteristic data within a preset period, constructs the physiological characteristic data into tensor data containing individual, characteristic, and time dimensions, and uses frequency domain transformation and singular value decomposition methods to complete the missing data in the tensor data, obtaining the completed physiological characteristic data. This completed physiological characteristic data is then input into a preset health status assessment model to calculate the user's health status assessment result within the preset period. Therefore, by inputting the completed physiological characteristic data obtained using frequency domain transformation and singular value decomposition methods into the preset health status assessment model, the device calculates the user's health status assessment result within the preset period. This addresses the problem that existing methods for completing missing physiological characteristic data fail to fully consider the integrity and periodicity of human physiological characteristic data, leading to significant deviations between the completed data and the actual physiological state, thus affecting the accuracy of health status assessment. The device effectively completes missing physiological characteristic data, improving the accuracy of health status assessment.

[0102] Please refer to Figure 3, which is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. This application provides an electronic device including: a processor 301 and a memory 302. The processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanism (not shown). The memory 302 stores a computer program executable by the processor 301. When the electronic device is running, the processor 301 executes the computer program to execute the health status assessment method in any optional implementation of the above embodiments, so as to achieve the following functions: acquiring the user's physiological characteristic data within a preset period, constructing the physiological characteristic data into tensor data containing individual dimension, feature dimension and time dimension, using frequency domain transformation method and singular value decomposition method to complete the missing data in the tensor data to obtain the completed physiological characteristic data, inputting the completed physiological characteristic data into a preset health status assessment model, and calculating the user's health status assessment result within the preset period.

[0103] This application provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it executes the health status assessment method in any optional implementation of the above embodiments to achieve the following functions: acquiring the user's physiological characteristic data within a preset period, constructing the physiological characteristic data into tensor data containing individual dimensions, feature dimensions, and time dimensions, using frequency domain transformation and singular value decomposition methods to complete the missing data in the tensor data to obtain the completed physiological characteristic data, inputting the completed physiological characteristic data into a preset health status assessment model, and calculating the user's health status assessment result within the preset period. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0104] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0105] Furthermore, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0106] Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0107] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0108] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A health status assessment method, used to assess health status, characterized in that, The steps include: acquiring the user's physiological characteristic data within a preset period; The physiological characteristic data is constructed into tensor data containing individual, feature, and time dimensions; the missing data in the tensor data is filled in using frequency domain transformation and singular value decomposition methods to obtain the filled physiological characteristic data. The completed physiological characteristic data is input into a preset health status assessment model to calculate the user's health status assessment result within a preset period.

2. The health status assessment method according to claim 1, characterized in that, Missing data in the tensor data is filled in using frequency domain transformation and singular value decomposition methods to obtain completed physiological feature data. This includes: performing a frequency domain transformation on the tensor data based on the time dimension to represent the original time-domain signal of the tensor data in the frequency domain, obtaining a frequency domain tensor; performing data repair on the frequency domain tensor using singular value decomposition to fill in the missing data, obtaining a completed frequency domain tensor; and performing an inverse frequency domain transformation on the completed frequency domain tensor based on the time dimension to obtain the completed physiological feature data.

3. The health status assessment method according to claim 2, characterized in that, The frequency domain tensor is repaired using singular value decomposition (SVD) to fill in missing data and obtain a repaired frequency domain tensor. This process includes: extracting a complex value matrix for each frequency point from the frequency domain tensor; performing SVD on the complex value matrix to reconstruct and update the singular value elements in the complex value matrix to obtain the repaired frequency domain tensor; and traversing the complex value matrix for all frequency points and summing all the repaired frequency domain tensors to obtain the repaired frequency domain tensor.

4. The health status assessment method according to claim 3, characterized in that, Performing singular value decomposition (SVD) on the complex numerical matrix to reconstruct and update the singular value elements in the complex numerical matrix to obtain a repaired frequency domain tensor includes: performing SVD on the complex numerical matrix to obtain a SVD-decomposed complex numerical matrix; updating the corresponding singular value elements based on the relationship between each singular value element in the SVD-decomposed complex numerical matrix and a preset singular threshold to obtain updated singular value elements; and updating the SVD-decomposed complex numerical matrix using the updated singular value elements to obtain a repaired frequency domain tensor.

5. The health status assessment method according to claim 4, characterized in that, Based on the relationship between the singular value elements in the complex numerical matrix after singular value decomposition and a preset singular threshold, the corresponding singular value elements are updated to obtain updated singular value elements. This includes: sequentially determining whether each singular value element in the complex numerical matrix after singular value decomposition is greater than the preset singular threshold; if so, calculating the difference between the corresponding singular value element and the preset singular threshold, and updating the corresponding singular value element to the difference to obtain updated singular value elements; if not, replacing the corresponding singular value element with zero to obtain updated singular value elements.

6. The health status assessment method according to claim 5, characterized in that, The process involves traversing the complex value matrix at all frequency points and summarizing all the repaired frequency domain tensors to obtain the completed frequency domain tensor. This includes: traversing the complex value matrix at all frequency points to obtain the repaired frequency domain tensor corresponding to each frequency point; comparing the complex value matrix at all frequency points with the corresponding repaired frequency domain tensor to determine whether all singular value elements in the repaired frequency domain tensor have been updated; if so, summarizing all the repaired frequency domain tensors to obtain the completed frequency domain tensor; if not, comparing the size relationship between the unupdated singular value elements and a preset singular threshold to update the unupdated singular value elements, and then summarizing all the repaired frequency domain tensors to obtain the completed frequency domain tensor.

7. The health status assessment method according to claim 1, characterized in that, The process involves inputting the completed physiological characteristic data into a preset health status assessment model to calculate the user's health status assessment result within a preset period. This includes: inputting the completed physiological characteristic data into the preset health status assessment model to obtain the user's preliminary health status assessment result within the preset period; calculating the corresponding information entropy based on the probability distribution of each assessment result in the preliminary health status assessment result; determining whether the information entropy is greater than a preset uncertainty threshold; if not, determining that the preliminary health status assessment result is reliable and identifying it as the health status assessment result; if yes, calculating the difference between each assessment result to extract associated physiological characteristic questions from a preset question bank corresponding to assessment results with differences less than the preset assessment threshold; having the user answer the associated physiological characteristic questions; updating the corresponding physiological characteristic data based on the user's answer; and inputting the updated physiological characteristic data into the preset health status assessment model to repeat the health status assessment steps until the information entropy corresponding to the updated preliminary health status assessment result is less than or equal to the preset uncertainty threshold, at which point the updated preliminary health status assessment result is identified as the health status assessment result.

8. A health status assessment device for assessing health status, characterized in that, include: The acquisition module is used to acquire the user's physiological characteristic data within a preset period; The construction module is used to construct the physiological feature data into tensor data containing individual dimension, feature dimension and time dimension; the completion module is used to complete the missing data in the tensor data using frequency domain transformation method and singular value decomposition method to obtain the completed physiological feature data. The calculation module is used to input the completed physiological characteristic data into a preset health status assessment model and calculate the user's health status assessment result within a preset period.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a computer program executable by the processor, which, when executed by the processor, performs the steps of the health status assessment method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it performs the steps in the health status assessment method as described in any one of claims 1-7.

Images