Smell sample-based health state recognition method and related equipment

Abstract

The invention relates to a health state recognition method based on an odor sample and related equipment, and the method comprises the following steps: collecting the odor of a user to obtain a user odor sample, and carrying out the chemical analysis of the user odor sample through employing a preset chromatography-mass spectrometry technology to obtain odor component information; performing quantitative analysis and qualitative identification on the smell component information to obtain molecular composition characteristics corresponding to the smell components; performing principal component feature extraction on the molecular composition features to obtain principal component extraction features; performing automatic feature mining on the principal component extraction features to obtain potential health state features, and performing time sequence analysis on the potential health state features to obtain dynamic health state trend information; classifying the dynamic health state trend information to identify the health state of the user; the technical problem that odor characteristics are unstable possibly due to the change of odor components among different individuals and the change of odor components of the same individual in different time periods is solved.

Description

Technical Field

[0001] The present invention relates to the field of odor technology, and in particular to a health status identification method based on odor samples and related equipment. Background Art

[0002] With the development of science and technology and people's increasing attention to health management, finding new, non-invasive health monitoring methods has become a research hotspot. As a biomarker, odor has gradually attracted attention for its uniqueness and ability to reflect human health. The health status recognition method based on odor samples attempts to judge the health status of an individual by analyzing the odor emitted by the human body. This method is not only non-invasive, but also can achieve continuous monitoring, providing the possibility for personalized health management.

[0003] However, research in this field still faces many challenges. First, the composition of human body odor is complex and easily affected by environmental factors. How to accurately collect and analyze the effective components in odor samples is a technical challenge. In addition, changes in odor components between different individuals and within the same individual over different time periods may lead to unstable odor characteristics, which makes it difficult to establish a universal odor component database. Furthermore, the current understanding of which specific odor components can indicate a certain health status is not sufficient, and further research is needed to clarify these biomarkers and their relationship with health status.

[0004] In addition, from the perspective of technical implementation, although chromatography-mass spectrometry can provide high-precision analysis of odor components, this process is often time-consuming and costly, making it difficult to achieve real-time monitoring and widespread application. Therefore, how to simplify the analysis process, reduce costs, and improve efficiency has become an urgent problem to be solved. At the same time, since the identification of health status depends on the processing and analysis of a large amount of data, how to ensure the accuracy and reliability of data analysis and avoid misdiagnosis or missed diagnosis is also a major obstacle that this method must overcome in practical applications. Summary of the invention

[0005] The main purpose of the present invention is to provide a health status identification method based on odor samples and related equipment, which solves the technical problem that changes in odor components between different individuals and within different time periods of the same individual may lead to unstable odor characteristics.

[0006] To achieve the above object, the present invention provides a method for identifying health status based on odor samples, comprising the following steps: The user's odor is collected to obtain a user odor sample, and a preset chromatography-mass spectrometry technique is used to chemically analyze the user odor sample to obtain odor component information; Quantitatively analyzing and qualitatively identifying the odor component information to obtain molecular composition characteristics corresponding to the odor component; Performing principal component feature extraction on the molecular composition feature to obtain a principal component extraction feature; wherein the principal component extraction feature is a key feature associated with the health status corresponding to the user; Automatically mining the principal component extraction features to obtain potential health status features, and performing time series analysis on the potential health status features to obtain dynamic health status trend information; The dynamic health status trend information is classified to identify the user's health status; wherein the user's health status includes normal, warning and abnormal.

[0007] Furthermore, the user's odor is collected to obtain a user odor sample, and a preset chromatography-mass spectrometry technology is used to perform chemical analysis on the user's odor sample to obtain odor component information, including: The user's odor is collected non-invasively through a preset breath collection device to obtain a user odor sample; Filtering the user odor sample by using a one-way valve technology to obtain a filtered odor sample; Digitally processing the filtered odor sample to obtain filtered odor sample data; Preprocessing the filtered odor sample data to obtain odor preprocessed sample data, and using the odor preprocessed sample data as the user odor sample; Performing chemical analysis on the odor preprocessing sample data to obtain an odor component spectrum; Odor component information is obtained based on the odor component spectrum.

[0008] Furthermore, the odor component information is quantitatively analyzed and qualitatively identified to obtain the molecular composition characteristics corresponding to the odor components, including: Performing mass spectrometry analysis on the odor component information to obtain the molecular mass distribution of the odor component; Performing isotope pattern fitting on the molecular mass distribution to obtain the isotope abundance of the odor component; Inferring the chemical structure of the odor component based on the isotope abundance to obtain chemical structure information of the odor component; wherein the chemical structure information includes the skeleton structure and functional group type of the molecule; Performing topological analysis and graph theory modeling on the chemical structure information to obtain topological characteristics of the odor components; wherein the topological characteristics of the odor components include the number of rings of the odor molecules, the number of branches of the odor molecules, and the connection mode of the odor molecules, and the connection mode of the odor molecules includes the bonding type and spatial arrangement between atoms; Performing feature encoding on the topological features to obtain odor coding components; A molecular fingerprint is generated for the odor component based on the odor coding component to obtain a molecular composition feature corresponding to the odor component.

[0009] Further, the molecular composition features are subjected to principal component feature extraction to obtain principal component extraction features, including: Performing high-dimensional data preprocessing on the molecular composition features to obtain molecular feature standardized data; wherein the molecular feature standardized data is a feature vector with zero mean and unit variance; Performing local nonlinear dimensionality reduction on the molecular feature standardized data to obtain a local low-dimensional embedding; Performing sparse representation on the local low-dimensional embedding to obtain a sparse coefficient matrix; Performing graph Laplacian matrix construction on the sparse coefficient matrix to obtain a molecular feature graph; Performing principal component screening on the molecular feature graph by using elastic network regression technology to obtain screened principal components; The molecular composition features are characterized based on the screened principal components to obtain principal component extraction features.

[0010] Furthermore, the automatic feature mining of the principal component extraction features to obtain potential health status features, and the time series analysis of the potential health status features to obtain dynamic health status trend information includes: Performing nonlinear dynamic analysis on the principal component extraction features to obtain dynamic characteristic parameters of the odor; wherein the dynamic characteristic parameters of the odor are indicators of characteristics that change over time; Performing abstract feature extraction on the dynamic feature parameters of the odor to obtain extracted abstract features; Adaptively selecting and mining the extracted abstract features to obtain a key feature subset; wherein the key feature subset is a significant variable and a dynamic change trend feature in a principal component analysis; Performing time series analysis on the key feature subset using a preset long short memory algorithm to obtain time series features; Performing latent space mapping on the time series features to obtain latent health status features; A health status trend analysis is performed based on the potential health status characteristics to obtain dynamic health status trend information.

[0011] Further, the dynamic health status trend information is classified to identify the user's health status, including: Performing multi-scale spectral clustering on the dynamic health status trend information to obtain a multi-scale clustering result; wherein the multi-scale clustering result is a health status clustering label at different time scales; The multi-scale clustering results are embedded in a low dimension by using a preset Laplace eigenmapping technology to obtain a low-dimensional health status representation; Performing health state multi-view fusion on the low-dimensional health state representation to obtain a fused health state feature; wherein the fused health state feature is a comprehensive health state feature under multiple perspectives; Performing topological data analysis on the fused health status feature to obtain an analysis topological feature; wherein the analysis topological feature includes edges and nodes, the nodes represent health indicators, and the edges represent the speed of transfer between health indicators; Performing granular analysis on the analysis topological features to obtain health status granular features; wherein the health status granular features represent the details of health status changes at different time scales; The user's health status is identified and classified based on the health status granularity feature to obtain a user health status classification result.

[0012] Furthermore, the topological features are subjected to granular analysis to obtain health status granular features, including: Performing multi-scale granularity decomposition on the topological features to obtain topological features at different granularity levels; Performing network structure analysis on the topological features at different granularity levels to obtain network structure features; wherein the network structure features include node degree, average path length and clustering coefficient; the node degree is the number of connections of the node at each time point, the average path length is the average value of the shortest path length between the nodes, which is used to reflect the overall connectivity of the health status network, and the clustering coefficient is a measure of the density of neighbors around the node, reflecting the similarity within the health status; Performing granular feature fusion on the network structure features to obtain fused granular features; Performing topological invariance analysis on the fused granularity features to obtain invariance features; Performing integrated analysis on the invariance features to obtain integrated granularity features; Based on the integrated granularity features, a fine-grained analysis of the health status is performed to obtain the health status granularity features.

[0013] The present invention also provides a health status identification device based on odor samples, comprising: A collection module is used to collect the user's odor to obtain a user's odor sample, and use a preset chromatography-mass spectrometry technology to chemically analyze the user's odor sample to obtain odor component information; An identification module is used to perform quantitative analysis and qualitative identification on the odor component information to obtain molecular composition characteristics corresponding to the odor component; An extraction module, used to extract principal component features from the molecular composition features to obtain principal component extraction features; wherein the principal component extraction features are key features associated with the health status corresponding to the user; A mining module, used for automatically mining the principal component extraction features to obtain potential health status features, and performing time series analysis on the potential health status features to obtain dynamic health status trend information; The classification module is used to classify the dynamic health status trend information to identify the user's health status; wherein the user's health status includes normal, warning and abnormal.

[0014] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0015] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0016] The health status identification method based on odor samples provided by the present invention includes the following steps: collecting the user's odor to obtain the user's odor sample, and using a preset chromatography-mass spectrometry technology to chemically analyze the user's odor sample to obtain odor component information; quantitatively analyzing and qualitatively identifying the odor component information to obtain molecular composition characteristics corresponding to the odor components; extracting principal component characteristics from the molecular composition characteristics to obtain principal component extraction characteristics; automatically mining the principal component extraction characteristics to obtain potential health status characteristics, and performing time series analysis on the potential health status characteristics to obtain dynamic health status trend information; classifying the dynamic health status trend information to identify the user's health status; the above-mentioned technical solution solves the technical problem that changes in odor components between different individuals and the same individual in different time periods may lead to unstable odor characteristics, and realizes that the chemical component information in the odor sample can be quickly and accurately obtained by using chromatography-mass spectrometry technology, and the molecular composition characteristics related to the health status can be effectively identified through quantitative analysis and qualitative identification of these information. This method helps to increase the possibility of early detection of diseases, thereby buying more time for treatment, and through principal component feature extraction and automatic feature mining technology, it can capture the trend information of the user's health status changing over time. Such dynamic monitoring can help doctors better understand the development of patients' conditions and adjust treatment plans in a timely manner. It also facilitates personal health management and prevents the occurrence and development of diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1is a schematic diagram of the steps of a method for identifying health status based on odor samples in one embodiment of the present invention; Figure 2 is a structural block diagram of a health status identification device based on odor samples in one embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0018] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0020] like Figure 1 As shown, Figure 1 is a schematic diagram of steps of a method for identifying health status based on odor samples in one embodiment of the present invention; In one embodiment of the present invention, the present invention provides a method for identifying health status based on odor samples, comprising the following steps: Step S1, collecting the user's odor to obtain a user odor sample, and using a preset chromatography-mass spectrometry technology to perform chemical analysis on the user's odor sample to obtain odor component information.

[0021] Specifically, in order to collect the user's odor, obtain the user's odor sample, and use the preset chromatography-mass spectrometry technology to chemically analyze the user's odor sample to obtain the odor component information, it is first necessary to establish an effective odor collection system. This system must be able to efficiently and accurately capture the odor emitted by the user's body. Common collection methods include using specially designed odor collectors, which can be patches attached to the skin, handheld devices, or sensors placed in a specific environment. These devices need to be strictly calibrated and sterilized to ensure that the collected odor samples are not contaminated or interfered with by the outside world. Once the odor sample is successfully collected, the sample will be sealed and stored in a special container to prevent the sample from volatilizing or chemically changing. Next, the sample will be sent to the laboratory for analysis. In the laboratory, the sample first undergoes a pretreatment step, such as concentration or dilution, to ensure that it is suitable for subsequent analysis. Next, the sample will be introduced into the chromatography-mass spectrometry system for analysis. In the chromatography part, the different components in the odor sample will be separated one by one according to their chemical and physical properties. The selection of the stationary phase and mobile phase filled in the chromatographic column is crucial because they determine the separation effect of each component in the sample. For example, gas chromatography (GC) uses gas as the mobile phase and is suitable for analyzing volatile organic compounds; liquid chromatography (LC) uses liquid as the mobile phase and is suitable for analyzing non-volatile or thermally unstable compounds. After the separation is completed, it enters the mass spectrometry part. The mass spectrometer ionizes the sample molecules through the ionization source to generate charged ions. These ions are separated according to the mass-to-charge ratio (m / z) under the action of electric and magnetic fields, and finally generate signals on the detector. By analyzing these signals, the mass spectrometer can provide molecular weight and structural information of each component. Taking a practical application as an example, suppose you want to detect whether a person has a certain disease, you can collect volatile organic compounds (VOCs) in their exhaled breath. Through gas chromatography-mass spectrometry technology, you can separate and identify the individual VOCs components in the exhaled breath, and determine the concentration of each component through quantitative analysis. Then, these data can be compared with the data of healthy people to determine whether the person has abnormalities. In summary, through precise odor collection, strict sample storage, fine chromatographic separation and highly sensitive mass spectrometry analysis, the chemical composition information of user odor samples can be efficiently acquired, thereby providing reliable data support for subsequent health status identification.

[0022] Step S2, performing quantitative analysis and qualitative identification on the odor component information to obtain molecular composition characteristics corresponding to the odor components.

[0023] Specifically, in order to achieve "quantitative analysis and qualitative identification of the odor component information, and obtain the molecular composition characteristics corresponding to the odor component", it is first necessary to rely on the odor component information obtained by chromatography-mass spectrometry in the previous step. The odor component information is generated by the user's odor sample after chromatographic separation and mass spectrometry detection, including the quality, quantity and respective physicochemical characteristics of all chemical components in the sample. On this basis, the goal of quantitative analysis and qualitative identification is to determine the specific molecular composition and concentration of each odor component. Quantitative analysis mainly depends on the detection signal intensity of the mass spectrometer. By comparing the ion peak intensity detected by the mass spectrometer with the known component peak in the standard sample, the relative content of each component in the sample can be derived. For example, in the analysis of human breath samples, certain diseases can cause the concentration of specific volatile organic compounds (VOCs) to increase. By comparing with normal values, the abnormal levels of these specific components can be quantitatively identified. Qualitative identification is to identify the molecular structure of each odor component through the mass-to-charge ratio (m / z) value displayed in the mass spectrometer, combined with the known molecular mass spectrometry data in the database. The mass spectrometer will generate a characteristic fingerprint spectrum for each odor component, which is matched with known compounds in the standard database, and the specific molecular structure of each component in the odor sample can be inferred. For example, in an odor sample, if a specific molecular formula C6H12O (cyclohexanol) is identified by mass spectrometry, the molecular composition characteristics of the odor component can be qualitatively confirmed. For example, assuming that a patient's sweat sample is analyzed, quantitative analysis may show that the acetic acid content is significantly higher than the normal level, and qualitative identification determines the presence of acetic acid by matching molecular mass and molecular formula. This molecular composition feature can be used as an important indicator of the user's health status, and combined with other features, it can be used to infer the possibility of disease. Through the above quantitative analysis and qualitative identification, the specific molecular composition of the odor component can be clearly revealed, thereby providing a scientific basis for the identification of health status in the subsequent steps. This process not only ensures the accuracy of the analysis, but also reveals potential molecular characteristics, providing strong support for individualized health management.

[0024] Step S3, extracting principal component features from the molecular composition features to obtain principal component extraction features; wherein the principal component extraction features are key features associated with the health status corresponding to the user.

[0025] Specifically, in order to achieve "extracting the principal component features from the molecular composition features to obtain the principal component extraction features; wherein the principal component extraction features are key features associated with the health status corresponding to the user", it is first necessary to start from the molecular composition features obtained in the previous step and extract the key information highly associated with the user's health status through a data processing method. Principal component feature extraction is a multidimensional data dimensionality reduction technology, which aims to reduce redundant data while maintaining the main information of the data, so as to more efficiently analyze and classify the health status. In the specific implementation process, the principal component feature extraction mainly adopts the principal component analysis (PCA) algorithm. PCA transforms multiple related variables into a few unrelated variables through linear transformation, and these variables are called principal components. By performing PCA processing on the molecular composition features of the odor sample, those key features that have a significant impact on the change of health status can be screened out, while those noise or minor features will be ignored. For example, if the concentration fluctuation of a certain volatile organic compound (VOCs) is closely related to the change of a specific health status of the user, this molecular composition feature will be marked as a principal component. During the extraction process, the system first identifies which molecular composition features have an important impact on the classification of health status by training a large number of healthy and abnormal samples. Next, these features are converted into principal component features to simplify the data dimension, reduce the complexity of the model, and improve the efficiency of analysis. For example, if the content of certain fatty acids in the user's sweat sample is related to fatigue status, the molecular composition features of these fatty acids will be extracted as principal components in PCA and become the basis for determining the key health status. In this way, principal component feature extraction helps convert a large amount of complex molecular data into a small number of key features that are easy to analyze and classify. These principal component extraction features are highly correlated with the user's health status, can accurately reflect the user's health trends, and provide strong data support for subsequent health status classification and diagnosis. In this way, by performing principal component feature extraction on molecular composition features, the system can effectively filter out irrelevant information and focus on core features related to health status, providing a scientific basis for accurate health monitoring and early warning.

[0026] Step S4, performing automatic feature mining on the principal component extraction features to obtain potential health status features, and performing time series analysis on the potential health status features to obtain dynamic health status trend information.

[0027] Specifically, in order to achieve "automatic feature mining of the principal component extraction features to obtain potential health status features, and time series analysis of the potential health status features to obtain dynamic health status trend information", it is first necessary to start with the principal component extraction features obtained in the previous steps, and use automatic feature mining technology to identify and extract potential health status features. This process mainly relies on machine learning algorithms and data mining technology, which can automatically discover complex patterns and relationships related to health status from principal component features. In the specific implementation process, automatic feature mining technology includes two main parts: feature selection and feature construction. Feature selection uses algorithms to filter out features that have the greatest contribution to health status prediction, ignoring redundant and irrelevant features. These algorithms include decision trees, random forests, support vector machines (SVMs), etc., which can train principal component features and automatically determine which features are important. For example, when analyzing a user's sweat sample data, fluctuations in the concentration of specific volatile organic compounds may be associated with changes in health status, and these key features can be automatically mined through these algorithms. Once the potential health status features are determined, time series analysis is required to track the changing trends of these features over time. Time series analysis techniques include moving average, exponential smoothing, time series regression models, etc., which are used to analyze the dynamic changes of feature data over time. These techniques can reveal long-term trends and periodic changes in health status. For example, by performing time series analysis on the odor sample data collected by users in different time periods, the gradual deterioration or improvement trend of certain health conditions can be identified, thereby providing early warning information or verifying the effectiveness of health intervention measures. For example, if the principal component feature extraction shows that the concentration of certain specific volatile organic compounds is changing during long-term monitoring of a user, the relationship between the changes in these components and potential health problems can be identified through automatic feature mining. Subsequently, the application of time series analysis technology to track the changing trends of the concentrations of these components can reveal the dynamic evolution of the user's health status, such as the development or improvement of chronic diseases. Through the above method, automatic feature mining and time series analysis of the principal component extraction features can not only extract key potential health status features, but also monitor and predict changes in health status in real time, thereby achieving more accurate health management and intervention.

[0028] Step S5, classifying the dynamic health status trend information to identify the user's health status; wherein the user's health status includes normal, warning and abnormal.

[0029] Specifically, in order to achieve "classifying the dynamic health status trend information to identify the user's health status; wherein the user's health status includes normal, warning and abnormal", it is first necessary to effectively classify the dynamic health status trend information obtained from the above steps. This process mainly relies on the classification algorithm to determine the user's health status by analyzing the patterns and features in the trend information. The classification process first needs to build a classification model, which is trained using the labeled health status data. These data include dynamic trend information under different health states (such as normal, warning, and abnormal). Through training, the classification model learns the characteristic patterns and change laws of each health state. For example, the dynamic trend of the normal state may be stable or slightly fluctuating, while the warning state may show an early abnormal change trend, and the abnormal state may be accompanied by a significant deterioration of health indicators. In actual classification, the algorithms used may include logistic regression, support vector machine (SVM), decision tree, random forest, etc. These algorithms predict the user's health status based on the learned pattern by inputting the dynamic health status trend information into the model. The algorithm compares the trend information with the known health status pattern in the model to determine whether the user's health status is normal, warning or abnormal. For example, when monitoring a user's health status over a long period of time, dynamic health status trend information shows that some of the user's key health indicators (such as the concentration of a specific chemical component) have a trend of continuous increase. If the classification model is trained to recognize that this trend is usually associated with abnormal health status, the model will mark the user's health status as abnormal. On the contrary, if the trend is stable or within the normal range, the model will mark it as normal. In the case in between, such as when the trend fluctuates but does not reach the abnormal threshold, the model may mark it as a warning state. Through this classification process, the user's health status can be accurately divided into normal, warning or abnormal, thereby providing a basis for further health management and intervention. This classification not only helps identify potential health problems, but also provides early warnings, helping users take necessary measures in time to maintain health.

[0030] In a specific embodiment, the user's odor is collected to obtain a user odor sample, and a preset chromatography-mass spectrometry technology is used to perform chemical analysis on the user's odor sample to obtain odor component information, including: The user's odor is collected non-invasively through a preset breath collection device to obtain a user odor sample; Filtering the user odor sample by using a one-way valve technology to obtain a filtered odor sample; Digitally processing the filtered odor sample to obtain filtered odor sample data; Preprocessing the filtered odor sample data to obtain odor preprocessed sample data, and using the odor preprocessed sample data as the user odor sample; Performing chemical analysis on the odor preprocessing sample data to obtain an odor component spectrum; Odor component information is obtained based on the odor component spectrum.

[0031] Specifically, in the technical solution described above, the core is to collect, process and analyze the user's odor sample through a series of steps to obtain odor component information. The implementation of this process involves a variety of technologies, including the collection, filtering, digital processing and chemical analysis of odor samples. Here, I will explain the implementation of each step in detail and give examples in combination with practical applications to help understand this complex technical process. First of all, the above-mentioned "collecting the user's odor to obtain the user's odor sample" is the first step of the entire technical process. The collection of user odor samples is non-invasive, which is particularly important because non-invasive collection technology can effectively avoid physiological effects on users while ensuring the safety and convenience of sampling. To achieve this, a "preset breath collection device" is used. This device can usually ensure that the user's odor sample can be collected quickly and completely by designing a unique collection port. For example, the breath collection device can be used to collect the gas exhaled by the user to capture the key chemical components in the user's odor. This collection method is similar to the collection method in breath alcohol testing, except that the focus here is on odor rather than alcohol concentration. Next, "filtering the user odor sample through the one-way valve technology to obtain a filtered odor sample" is a key step in further processing the odor sample. The application of the one-way valve technology is to ensure the purity of the odor sample and avoid the mixing of external air or other impurities, thereby ensuring the authenticity of the odor sample. This filtering process can be achieved by a one-way valve installed in the collection device. For example, similar to the one-way valve design in some high-end air purification equipment, the gas exhaled by the user can only enter the collection device through the valve, while the external air cannot enter, so that a relatively pure user odor sample can be obtained for the next step of analysis. After the odor sample is filtered by the one-way valve, the above describes the step of "digital processing of the filtered odor sample to obtain filtered odor sample data". The "digital processing" here refers to converting the collected odor sample from a physical sample into digital data that can be processed by a computer. This usually requires the use of precision sensors or detection equipment to sense the chemical components in the odor sample, and converting these chemical components into electronic signals through a converter, and then further processing them into digital signals. For example, in modern medical or chemical laboratories, when analyzing odor samples using a gas chromatography-mass spectrometry instrument (GC-MS), the gas samples are usually first digitized so that they can be analyzed and processed by a computer. This digitization helps to convert the data of the original odor sample into an electronic format that is easier to handle. The next step mentioned above mentions "preprocessing the filtered odor sample data to obtain odor preprocessed sample data." Preprocessing generally refers to the process of cleaning, denoising or standardizing the raw data. In actual operation, preprocessing sample data involves removing noise signals or standardizing the data format, so that subsequent analysis is more accurate.For example, in odor analysis, the collected odor samples may contain multiple unrelated odor signals or data noise caused by equipment errors. Therefore, preprocessing is used to extract components related to the target user's odor to ensure the accuracy of the data. Next, it is further explained that "the odor preprocessed sample data is used as the user odor sample", which means that the odor sample data after preprocessing is more in line with the analysis requirements and can be formally used as the representative data of the user's odor for analysis. After that, the above-mentioned "chemical analysis of the odor preprocessed sample data to obtain an odor component spectrum" is the core part of the entire technical implementation. The acquisition of odor component spectra usually relies on advanced chemical analysis equipment, such as the gas chromatography-mass spectrometry (GC-MS) technology mentioned above. This technology can separate, identify and quantitatively analyze the chemical components in the odor sample, and finally draw an odor component spectrum. For example, when analyzing a user's breath sample, GC-MS can separate the volatile organic compounds therein, and analyze the molecular weight and structure of these compounds by mass spectrometry, thereby obtaining a clear odor component spectrum. Finally, it is mentioned that "odor component information is obtained based on the odor component spectrum". This process is to interpret and analyze the odor spectra, with the goal of extracting key information from the odor component spectra to determine the chemical components and their concentrations in the user's odor. For example, if a user's odor sample contains specific volatile organic compounds, such as acetaldehyde or ketones, the quantitative analysis of these chemical components can further understand the user's health status, eating habits, and even mood changes. This analysis has a wide range of applications in many fields, including medical testing, food analysis, and environmental monitoring. For example, some high-end health testing equipment has been able to identify early signs of certain metabolic diseases, such as diabetes or cancer, by analyzing the user's breath odor. In summary, the technical solution described above achieves the collection, processing, and analysis of user odors through a series of complex and sophisticated steps. This process not only relies on advanced collection and filtering technologies, but also relies on sophisticated chemical analysis instruments, such as gas chromatography-mass spectrometry, to ultimately obtain the composition information of the user's odor. This technology is expected to play an important role in health monitoring, personalized health care, and smart devices.

[0032] In a specific embodiment, the odor component information is quantitatively analyzed and qualitatively identified to obtain the molecular composition characteristics corresponding to the odor components, including: Performing mass spectrometry analysis on the odor component information to obtain the molecular mass distribution of the odor component; Performing isotope pattern fitting on the molecular mass distribution to obtain the isotope abundance of the odor component; Inferring the chemical structure of the odor component based on the isotope abundance to obtain chemical structure information of the odor component; wherein the chemical structure information includes the skeleton structure and functional group type of the molecule; Performing topological analysis and graph theory modeling on the chemical structure information to obtain topological characteristics of the odor components; wherein the topological characteristics of the odor components include the number of rings of the odor molecules, the number of branches of the odor molecules, and the connection mode of the odor molecules, and the connection mode of the odor molecules includes the bonding type and spatial arrangement between atoms; Performing feature encoding on the topological features to obtain odor coding components; A molecular fingerprint is generated for the odor component based on the odor coding component to obtain a molecular composition feature corresponding to the odor component.

[0033] Specifically, in order to achieve "quantitative analysis and qualitative identification of the odor component information, and obtain the molecular composition characteristics corresponding to the odor component", several important steps need to be completed in sequence, from mass spectrometry to molecular fingerprint generation. The following is a detailed explanation of how to achieve these steps. First, mass spectrometry analysis of the odor component information is the key first step. The molecular mass distribution of the odor component can be obtained through mass spectrometry. The mass spectrometer ionizes the odor sample and measures its mass-to-charge ratio (m / z) to generate a mass spectrum, which shows the molecular mass of each component and its relative abundance. Mass spectrometry can provide highly sensitive and high-resolution data to help identify complex compounds in the sample. Next, the molecular mass distribution is fitted with an isotope pattern to obtain the isotope abundance of the odor component. Isotope pattern fitting is to identify the presence and abundance of different isotopes by analyzing the subtle changes in each peak in the mass spectrum. Isotope abundance information can further improve the accuracy of identifying odor components because the isotope distribution of different compounds is unique. Isotope pattern fitting is usually performed using computing software. By comparing experimental data with theoretical models, the relative abundance of each isotope is accurately calculated. After obtaining the isotope abundance, the chemical structure of the odor component is inferred based on the isotope abundance, thereby obtaining the chemical structure information of the odor component. The purpose of this step is to use the isotope abundance and mass distribution information to infer the specific chemical structure of the molecule. This includes determining the skeleton structure and functional group type of the molecule. For example, by analyzing the mass and abundance of each peak in the mass spectrum, the arrangement and connection mode of atoms such as carbon, hydrogen, and oxygen in the molecule, as well as the presence of functional groups such as hydroxyl and carboxyl groups, can be determined. Then, the chemical structure information is subjected to topological analysis and graph theory modeling to obtain the topological characteristics of the odor component. Topological analysis converts the chemical structure of the molecule into a graph theory model to analyze the connection mode of each atom and bond in the molecule. The topological characteristics of the odor component include the number of rings, the number of branches, and the connection mode of the odor molecule. Specifically, the connection mode includes the type of bonding between atoms (such as single bonds, double bonds, triple bonds) and the spatial arrangement mode (such as cis, trans, etc.). Graph theory modeling can systematically describe and analyze these complex molecular structure information through mathematical algorithms, thereby extracting meaningful topological features. Next, the topological features are feature encoded to obtain odor coding components. This step converts the topological features into standardized coding forms so that these features can be further processed and analyzed. Feature coding methods include using hash functions, binary coding or other mathematical conversion techniques to compress complex topological information into easy-to-process coding forms. These coding forms are not only convenient for storage and transmission, but also can improve the efficiency of subsequent analysis. Finally, based on the odor coding component, a molecular fingerprint is generated for the odor component to obtain the molecular composition characteristics corresponding to the odor component.Molecular fingerprints are highly concentrated molecular descriptors. By further processing the odor coding components, a unique and concise molecular fingerprint can be generated. Molecular fingerprints are usually represented in binary arrays or other forms, and contain all the important structural features and connection patterns in the molecule. This molecular fingerprint can not only quickly identify and compare different odor components, but can also be used in machine learning models to help predict and analyze health status. Through the above steps, from mass spectrometry analysis to molecular fingerprint generation, the odor component information can be systematically quantitatively analyzed and qualitatively identified to obtain the molecular composition characteristics corresponding to the odor components. This series of steps ensures that the chemical information extracted from the odor sample can be accurately and comprehensively parsed, providing a solid data foundation for further health monitoring and analysis.

[0034] In a specific embodiment, the molecular composition feature is subjected to principal component feature extraction to obtain the principal component extraction feature, including: Performing high-dimensional data preprocessing on the molecular composition features to obtain molecular feature standardized data; wherein the molecular feature standardized data is a feature vector with zero mean and unit variance; Performing local nonlinear dimensionality reduction on the molecular feature standardized data to obtain a local low-dimensional embedding; Performing sparse representation on the local low-dimensional embedding to obtain a sparse coefficient matrix; Performing graph Laplacian matrix construction on the sparse coefficient matrix to obtain a molecular feature graph; Performing principal component screening on the molecular feature graph by using elastic network regression technology to obtain screened principal components; The molecular composition features are characterized based on the screened principal components to obtain principal component extraction features.

[0035] Specifically, in order to achieve "extracting the principal component features of the molecular composition features to obtain the principal component extraction features", it is necessary to process and analyze the molecular composition features through multiple steps. The following is a detailed explanation of how to implement these steps. First, the molecular composition features are preprocessed for high-dimensional data to obtain molecular feature standardized data. The goal of this step is to convert the original molecular composition features into feature vectors with zero mean and unit variance for standardization. High-dimensional data preprocessing includes adjusting the mean of each feature to zero and the variance to unit variance, so that all features are compared on the same scale. Standardization helps to eliminate bias and scale effects in the data, making subsequent dimensionality reduction and analysis more accurate and effective. Next, the molecular feature standardized data is subjected to local nonlinear dimensionality reduction to obtain a local low-dimensional embedding. Local nonlinear dimensionality reduction technology aims to embed high-dimensional data into a lower-dimensional space while retaining the local structural information of the data as much as possible. This method can identify potential nonlinear relationships in the data and map the data to a low-dimensional space, simplifying the complexity of subsequent analysis. For example, techniques such as local linear embedding (LLE) or isomap can effectively capture the local features of the data, so that the structure and similarity of the data can be maintained in the low-dimensional space. After obtaining the local low-dimensional embedding, it is sparsely represented to obtain a sparse coefficient matrix. The sparse representation method aims to find a sparse coefficient matrix by representing the data as a linear combination of several bases, so that the data can be effectively represented with fewer non-zero coefficients. The sparse coefficient matrix can not only compress the dimension of the data, but also highlight the important features in the data, suppress noise and redundant information, and thus improve the quality of subsequent analysis. Then, the sparse coefficient matrix is ​​constructed by graph Laplacian matrix to obtain the molecular feature map. The graph Laplacian matrix is ​​a matrix used for graph data analysis, which can capture the relationship and similarity between data points. By constructing the graph Laplacian matrix, the sparse coefficient matrix can be converted into a graph data representation, so that the similarity and difference of the data points can be reflected in the structure of the graph. The graph Laplacian matrix reflects the structural characteristics of the graph, such as the connection pattern and weight between nodes, and provides a basis for further principal component screening. The molecular feature map is subjected to principal component screening by elastic network regression technology to obtain the screened principal components. Elastic network regression is a regression method that combines L1 regularization and L2 regularization, which can perform feature selection and data dimension reduction while maintaining the complexity of the model. This technology selects features that have a significant impact on the principal components by optimizing the loss function, thereby screening out the most representative principal components. These principal components can retain the main information and structural characteristics of the data while reducing the data dimension. Finally, the molecular composition features are characterized based on the screened principal components to obtain the principal component extraction features.This step applies the screened principal components to the original molecular composition features to generate a new feature representation that effectively summarizes the main information of the data in a low-dimensional space. The principal component extraction features can be used for subsequent data analysis, model training, and prediction, providing a concise and meaningful representation of the molecular composition features. Through the above steps, from high-dimensional data preprocessing to principal component extraction feature generation, the molecular composition features can be systematically analyzed and optimized to obtain the principal component extraction features. This process ensures the accuracy and effectiveness of data analysis and provides a solid foundation for further scientific research and application.

[0036] In a specific embodiment, the automatic feature mining of the principal component extraction feature to obtain the potential health status feature, and the time series analysis of the potential health status feature to obtain the dynamic health status trend information includes: Performing nonlinear dynamic analysis on the principal component extraction features to obtain dynamic characteristic parameters of the odor; wherein the dynamic characteristic parameters of the odor are indicators of characteristics that change over time; Performing abstract feature extraction on the dynamic feature parameters of the odor to obtain extracted abstract features; Adaptively selecting and mining the extracted abstract features to obtain a key feature subset; wherein the key feature subset is a significant variable and a dynamic change trend feature in a principal component analysis; Performing time series analysis on the key feature subset using a preset long short memory algorithm to obtain time series features; Performing latent space mapping on the time series features to obtain latent health status features; A health status trend analysis is performed based on the potential health status characteristics to obtain dynamic health status trend information.

[0037] Specifically, in order to achieve "automatic feature mining of the principal component extraction features to obtain potential health status features, and time series analysis of the potential health status features to obtain dynamic health status trend information", several detailed steps are required, which will help to mine health status-related information from the principal component extraction features and analyze its dynamic change trend. First, the principal component extraction features are automatically feature mined in order to identify potential health status features. Automatic feature mining technology uses advanced algorithms to automatically extract meaningful features from large amounts of data. These features can reflect the potential health status patterns in the data and provide a basis for subsequent analysis. This step automatically identifies important features in the data through algorithms without manual intervention, thereby improving the efficiency and accuracy of the analysis. Next, the potential health status features are analyzed in time series to obtain dynamic health status trend information. This includes performing nonlinear dynamic analysis on the principal component extraction features to obtain dynamic characteristic parameters of odor. The dynamic characteristic parameters of odor are indicators that reflect the time-varying characteristics of odor, and can capture the time series changes of health status. It can be understood that "indicators of time-varying characteristics" (i.e., dynamic characteristic parameters of odor) are numerical values ​​or parameters used to measure the time-varying changes of the main component characteristics extracted from the odor sample. These indicators can reflect the change pattern of the odor sample at different time points, thereby helping to identify the trend of changes in the health status of an individual. For example, these indicators can be the rate of change of concentration of certain chemical substances, the frequency of occurrence, etc. By analyzing these indicators, potential features related to health status can be further mined. This analysis helps understand how health status evolves over time, thereby revealing potential health problems. Further, the dynamic characteristic parameters of odor are subjected to abstract feature extraction to obtain extracted abstract features. Abstract feature extraction converts dynamic feature parameters into higher-level representations, removes noise and redundancy in the data, and retains key health status information. The goal of this step is to extract simplified features that can effectively reflect health status from complex dynamic features. Then, adaptive feature selection and mining are performed on the extracted abstract features to obtain a subset of key features. The adaptive feature selection method can automatically select and extract the most representative features according to the actual situation of the data. These features include significant variables and dynamic change trend features in principal component analysis. The key feature subset is a feature that has a significant impact on the change of health status, which helps to accurately describe the change of health status. The key feature subset is subjected to time series analysis by a preset long short-term memory algorithm to obtain time series features. The long short-term memory algorithm (LSTM) is a neural network model that can process long time series data and is suitable for capturing long-term dependencies in data. Using LSTM to perform time series analysis on the key feature subset can extract important patterns and change trends in the time series and obtain time series features. Finally, the time series features are mapped to a latent space to obtain potential health status features.Latent space mapping technology maps time series features into latent space, thereby revealing the health status patterns hidden in the data. These latent health status features can reflect the health trends of individuals. Based on these features, health status trend analysis is performed to obtain dynamic health status trend information. Through this analysis, changes in health status can be effectively monitored and predicted, providing valuable health management information.

[0038] In a specific embodiment, the dynamic health status trend information is classified to identify the user's health status, including: Performing multi-scale spectral clustering on the dynamic health status trend information to obtain a multi-scale clustering result; wherein the multi-scale clustering result is a health status clustering label at different time scales; The multi-scale clustering results are embedded in a low dimension by using a preset Laplace eigenmapping technology to obtain a low-dimensional health status representation; Performing health state multi-view fusion on the low-dimensional health state representation to obtain a fused health state feature; wherein the fused health state feature is a comprehensive health state feature under multiple perspectives; Performing topological data analysis on the fused health status feature to obtain an analysis topological feature; wherein the analysis topological feature includes edges and nodes, the nodes represent health indicators, and the edges represent the speed of transfer between health indicators; Performing granular analysis on the analysis topology feature to obtain health status granular features; wherein the health status granular features represent the details of health status changes at different time scales; The user's health status is identified and classified based on the health status granularity feature to obtain a user health status classification result.

[0039] Specifically, in order to achieve "classifying the dynamic health status trend information to identify the user's health status", it is necessary to perform multiple steps in sequence, which are used to process the dynamic health status trend information and finally achieve the classification of the user's health status. First, multi-scale spectral clustering is performed on the dynamic health status trend information to obtain a multi-scale clustering result. Multi-scale spectral clustering is a clustering method that can cluster data at different time scales. Through this method, the clustering labels of data at different time scales can be identified, thereby revealing the patterns and trends of health status at different time scales. The multi-scale clustering results are health status clustering labels at different time scales, which can help us understand the changes and associations of health status in the short and long term. Next, the multi-scale clustering results are low-dimensionally embedded through the preset Laplace eigenmapping technology to obtain a low-dimensional health status representation. Laplace eigenmapping technology is a dimensionality reduction method that can embed high-dimensional data into a low-dimensional space while maintaining the local structure and global characteristics of the data. By embedding the multi-scale clustering results in low dimensions, the complexity of the data can be simplified, making the representation of the health status more intuitive and easy to analyze. The low-dimensional health status representation can reflect the main characteristics of the health status in fewer dimensions, which is convenient for subsequent analysis and processing. Then, the low-dimensional health status representation is subjected to health status multi-view fusion to obtain a fused health status feature. The health status multi-view fusion method integrates information from different perspectives to generate a comprehensive health status feature representation. Multi-view fusion can combine the results of different data sources and different analysis methods to generate more comprehensive and accurate health status features. The fused health status feature is a comprehensive health status feature from multiple perspectives, which can provide richer health status information and improve the accuracy and reliability of classification. Then, the fused health status feature is subjected to topological data analysis to obtain an analytical topological feature. Topological data analysis is a method for analyzing data structure and relationships, which can reveal topological characteristics in data, such as the relationship between nodes and edges. In health status analysis, nodes represent health indicators. In addition, for example, health indicators can be interactions between physiological parameters: for example, edges can represent the relationship between heart rate variability and blood pressure and their rate of change, while edges represent the speed of transfer between health indicators. This refers to the speed at which different health indicators are converted or influenced by each other. Through topological data analysis, the relationship and change pattern between health indicators can be identified, thereby obtaining analytical topological features. These features include the connection pattern and change rate of health indicators, which provide a basis for subsequent granular analysis. Further, the analytical topological features are subjected to granular analysis to obtain health status granular features. Granular analysis methods can analyze data at different time scales and levels of detail, revealing detailed features of the data.In health status analysis, health status granularity features represent the details of health status changes at different time scales. Through granularity analysis, the change patterns and details of health status at different time scales can be identified, providing rich information for the identification and classification of user health status. Finally, the user health status is identified and classified based on the health status granularity features to obtain the user health status classification results. The health status identification and classification method uses the health status granularity features, combined with machine learning and statistical analysis techniques, to classify the user's health status. Through this method, the user's health status can be divided into different categories, such as health, sub-health or disease status, so as to provide personalized health management and intervention measures. The user health status classification results can help medical professionals and users themselves understand the health status and make corresponding health decisions and adjustments. Through the above steps, from multi-scale spectral clustering to user health status classification, each step provides key information and technical support for the final health status identification and classification. Such systematic processing and analysis ensures the accuracy and effectiveness of health status classification and provides a scientific basis for user health management and medical decision-making.

[0040] In a specific embodiment, the topological features are subjected to granular analysis to obtain health status granular features, including: Performing multi-scale granularity decomposition on the topological features to obtain topological features at different granularity levels; Performing network structure analysis on the topological features at different granularity levels to obtain network structure features; wherein the network structure features include node degree, average path length and clustering coefficient; the node degree is the number of connections of the node at each time point, the average path length is the average value of the shortest path length between the nodes, which is used to reflect the overall connectivity of the health status network, and the clustering coefficient is a measure of the density of neighbors around the node, reflecting the similarity within the health status; Performing granular feature fusion on the network structure features to obtain fused granular features; Performing topological invariance analysis on the fused granularity features to obtain invariance features; Performing integrated analysis on the invariance features to obtain integrated granularity features; Based on the integrated granularity features, a fine-grained analysis of the health status is performed to obtain the health status granularity features.

[0041] Specifically, in order to achieve "performing granular analysis on the topological features to obtain granular features of health status", a series of steps need to be performed, which are helpful to extract detailed health status information from the topological features and perform granular level analysis. First, the topological features are multi-scale granular decomposition to obtain topological features at different granularity levels. Multi-scale granular decomposition technology is to decompose and analyze data at different scales to identify fine-grained and coarse-grained information. The purpose of this step is to decompose the topological features at multiple granularity levels in order to capture the characteristics of health status at different levels. Through this decomposition, the changes and patterns of health status at various granularity levels can be understood. Next, the topological features at different granularity levels are subjected to network structure analysis to obtain network structure characteristics. The network structure analysis method is used to analyze the network structure in the data and reveal the relationship between nodes and edges. In health status analysis, network structure characteristics include node degree, average path length and clustering coefficient. The node degree is the number of connections of the node at each time point, reflecting the influence and importance of each node in the health state network; the average path length is the average of the shortest path lengths between the nodes, which is used to reflect the overall connectivity of the health state network, and it can reveal the efficiency and speed of information propagation in the health state network; the clustering coefficient is a measure of the density of neighbors around the node, reflecting the similarity within the health state, and this indicator can reveal the tightness of the local structure in the health state. Then, the network structure features are fused with granular features to obtain fused granular features. The granular feature fusion method integrates network structure features at different granularity levels to generate a comprehensive feature representation. The purpose of this step is to combine multi-scale network structure features to form a comprehensive feature representation that can reflect the characteristics of the health state at all scales. The fused granular features can provide rich information and provide a basis for subsequent analysis. Next, the fused granular features are subjected to topological invariance analysis to obtain invariant features. The topological invariance analysis method is used to identify stable and invariant topological structure features in the data. Through this analysis, key features in the health state that are not affected by time and scale changes can be identified. Invariant features reflect stable patterns and structures in health status, and these features are of great significance in health status analysis because they can provide reliable health status information. Further, the invariant features are integrated and analyzed to obtain integrated granular features. Integrated analysis technology integrates invariant features to generate a comprehensive feature representation. Integrated granular features can combine the information of different invariant features to form a complete feature representation, which can more comprehensively reflect the characteristics of health status. The goal of this step is to integrate the invariant features to generate a comprehensive feature representation, which provides a basis for fine-grained analysis of health status. Finally, fine-grained analysis of health status is performed based on the integrated granular features to obtain health status granular features.The health status fine-grained analysis method uses integrated granular features combined with advanced analysis techniques to conduct a detailed analysis of the health status. Through this analysis, the changing patterns and characteristics of the health status at various levels of detail can be identified. The health status granular feature is a detailed description of the health status, which can provide rich information to help identify and classify the user's health status. Through the above steps, from multi-scale granularity decomposition to health status fine-grained analysis, each step provides key information and technical support for the final health status granular feature. Such systematic processing and analysis ensures the accuracy and effectiveness of health status analysis and provides a scientific basis for health management and medical decision-making.

[0042] The above describes the health status recognition method based on odor samples in the embodiment of the present invention. The following describes the health status recognition device based on odor samples in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, a device for identifying a health status based on an odor sample includes: The collection module 21 is used to collect the user's odor to obtain the user's odor sample, and use a preset chromatography-mass spectrometry technology to perform chemical analysis on the user's odor sample to obtain odor component information; The identification module 22 is used to perform quantitative analysis and qualitative identification on the odor component information to obtain the molecular composition characteristics corresponding to the odor component; An extraction module 23 is used to extract principal component features from the molecular composition features to obtain principal component extraction features; wherein the principal component extraction features are key features associated with the health status corresponding to the user; A mining module 24 is used to perform automatic feature mining on the principal component extraction features to obtain potential health status features, and perform time series analysis on the potential health status features to obtain dynamic health status trend information; The classification module 25 is used to classify the dynamic health status trend information to identify the user's health status; wherein the user's health status includes normal, warning and abnormal.

[0043] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the description in the above method embodiment, which will not be repeated here.

[0044] Reference Figure 3 The present invention also provides a computer device in an embodiment, wherein the internal structure of the computer device can be as follows: Figure 3As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0045] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0046] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0047] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0048] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0049] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for identifying health status based on odor samples, characterized in that: The following steps are involved: The user's odor is collected to obtain a user odor sample, and a preset chromatography-mass spectrometry technique is used to chemically analyze the user odor sample to obtain odor component information; Quantitatively analyzing and qualitatively identifying the odor component information to obtain molecular composition characteristics corresponding to the odor component; Performing principal component feature extraction on the molecular composition feature to obtain a principal component extraction feature; wherein the principal component extraction feature is a key feature associated with the health status corresponding to the user; Automatically mining the principal component extraction features to obtain potential health status features, and performing time series analysis on the potential health status features to obtain dynamic health status trend information; The dynamic health status trend information is classified to identify the user's health status; wherein the user's health status includes normal, warning and abnormal.

2. The method for identifying health status based on odor samples according to claim 1, characterized in that: The user's odor is collected to obtain a user odor sample, and a preset chromatography-mass spectrometry technology is used to perform chemical analysis on the user's odor sample to obtain odor component information, including: The user's odor is collected non-invasively through a preset breath collection device to obtain a user odor sample; Filtering the user odor sample by using a one-way valve technology to obtain a filtered odor sample; Digitally processing the filtered odor sample to obtain filtered odor sample data; Preprocessing the filtered odor sample data to obtain odor preprocessed sample data, and using the odor preprocessed sample data as the user odor sample; Performing chemical analysis on the odor preprocessing sample data to obtain an odor component spectrum; Odor component information is obtained based on the odor component spectrum.

3. The method for identifying health status based on odor samples according to claim 1, characterized in that: The odor component information is quantitatively analyzed and qualitatively identified to obtain the molecular composition characteristics corresponding to the odor component, including: Performing mass spectrometry analysis on the odor component information to obtain the molecular mass distribution of the odor component; Performing isotope pattern fitting on the molecular mass distribution to obtain the isotope abundance of the odor component; Inferring the chemical structure of the odor component based on the isotope abundance to obtain chemical structure information of the odor component; wherein the chemical structure information includes the skeleton structure and functional group type of the molecule; Performing topological analysis and graph theory modeling on the chemical structure information to obtain topological characteristics of the odor components; wherein the topological characteristics of the odor components include the number of rings of the odor molecules, the number of branches of the odor molecules, and the connection mode of the odor molecules, and the connection mode of the odor molecules includes the bonding type and spatial arrangement between atoms; Performing feature encoding on the topological features to obtain odor coding components; A molecular fingerprint is generated for the odor component based on the odor coding component to obtain a molecular composition feature corresponding to the odor component.

4. The method for identifying health status based on odor samples according to claim 1, characterized in that: The principal component feature is extracted from the molecular composition feature to obtain the principal component extraction feature, including: Performing high-dimensional data preprocessing on the molecular composition features to obtain molecular feature standardized data; wherein the molecular feature standardized data is a feature vector with zero mean and unit variance; Performing local nonlinear dimensionality reduction on the molecular feature standardized data to obtain a local low-dimensional embedding; Performing sparse representation on the local low-dimensional embedding to obtain a sparse coefficient matrix; Performing graph Laplacian matrix construction on the sparse coefficient matrix to obtain a molecular feature graph; Performing principal component screening on the molecular feature graph by using elastic network regression technology to obtain screened principal components; The molecular composition features are characterized based on the screened principal components to obtain principal component extraction features.

5. The method for identifying health status based on odor samples according to claim 1, characterized in that: The automatic feature mining of the principal component extraction features to obtain potential health status features, and the time series analysis of the potential health status features to obtain dynamic health status trend information include: Performing nonlinear dynamic analysis on the principal component extraction features to obtain dynamic characteristic parameters of the odor; wherein the dynamic characteristic parameters of the odor are indicators of characteristics that change over time; Performing abstract feature extraction on the dynamic feature parameters of the odor to obtain extracted abstract features; Adaptively selecting and mining the extracted abstract features to obtain a key feature subset; wherein the key feature subset is a significant variable and a dynamic change trend feature in a principal component analysis; Performing time series analysis on the key feature subset using a preset long short memory algorithm to obtain time series features; Performing latent space mapping on the time series features to obtain latent health status features; A health status trend analysis is performed based on the potential health status characteristics to obtain dynamic health status trend information.

6. The method for identifying health status based on odor samples according to claim 1, characterized in that: Classifying the dynamic health status trend information to identify the user's health status includes: Performing multi-scale spectral clustering on the dynamic health status trend information to obtain a multi-scale clustering result; wherein the multi-scale clustering result is a health status clustering label at different time scales; The multi-scale clustering results are embedded in a low dimension by using a preset Laplace eigenmapping technology to obtain a low-dimensional health status representation; Performing health state multi-view fusion on the low-dimensional health state representation to obtain a fused health state feature; wherein the fused health state feature is a comprehensive health state feature under multiple perspectives; Performing topological data analysis on the fused health status feature to obtain an analysis topological feature; wherein the analysis topological feature includes edges and nodes, the nodes represent health indicators, and the edges represent the speed of transfer between health indicators; Performing granular analysis on the analysis topological features to obtain health status granular features; wherein the health status granular features represent the details of health status changes at different time scales; The user's health status is identified and classified based on the health status granularity feature to obtain a user health status classification result.

7. The method for identifying health status based on odor samples according to claim 6, characterized in that: The topological features are subjected to granular analysis to obtain health status granular features, including: Performing multi-scale granularity decomposition on the topological features to obtain topological features at different granularity levels; Performing network structure analysis on the topological features at different granularity levels to obtain network structure features; wherein the network structure features include node degree, average path length and clustering coefficient; the node degree is the number of connections of the node at each time point, the average path length is the average value of the shortest path length between the nodes, which is used to reflect the overall connectivity of the health status network, and the clustering coefficient is a measure of the density of neighbors around the node, reflecting the similarity within the health status; Performing granular feature fusion on the network structure features to obtain fused granular features; Performing topological invariance analysis on the fused granularity features to obtain invariance features; Performing integrated analysis on the invariance features to obtain integrated granularity features; Based on the integrated granularity features, a fine-grained analysis of the health status is performed to obtain the health status granularity features.

8. A health status identification device based on odor samples, characterized in that: include: A collection module is used to collect the user's odor to obtain a user's odor sample, and use a preset chromatography-mass spectrometry technology to chemically analyze the user's odor sample to obtain odor component information; An identification module is used to perform quantitative analysis and qualitative identification on the odor component information to obtain molecular composition characteristics corresponding to the odor component; An extraction module, used to extract principal component features from the molecular composition features to obtain principal component extraction features; wherein the principal component extraction features are key features associated with the health status corresponding to the user; A mining module, used for automatically mining the principal component extraction features to obtain potential health status features, and performing time series analysis on the potential health status features to obtain dynamic health status trend information; The classification module is used to classify the dynamic health status trend information to identify the user's health status; wherein the user's health status includes normal, warning and abnormal.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images