Furniture harmful substance detection and analysis system

Abstract

The invention provides a furniture harmful substance detection and analysis system, and relates to the field of intelligent detection.The furniture harmful substance detection and analysis system comprises the steps that firstly, furniture VOC preliminary measurement values and corresponding temperature and humidity time sequence data are synchronously collected, and then the synergistic effect value of temperature and humidity at each time point is calculated; meanwhile, physicochemical property text description of the VOC components is extracted, and semantic coding is carried out to convert the physicochemical property text description into quantifiable features. And then, carrying out independence soft constraint deep correlation analysis, capturing a specific response rule of the VOC component under dynamic temperature and humidity, and finally dynamically generating a compensation coefficient so as to carry out nonlinear correction on the initial measurement value. Thus, through environment coupling dynamic modeling and material characteristic knowledge fusion, the defect that a traditional fixed compensation mode has insufficient characterization on the temperature and humidity synergistic effect and material difference is effectively overcome, and the detection accuracy of different VOC components in a complex environment is remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more particularly, in an embodiment of the present application, to a furniture harmful substance detection and analysis system. Background Art

[0002] As consumers pay more and more attention to the health of indoor environment, furniture harmful substance detection technology has become an important part of ensuring the safety of human living environment. The release of volatile organic compounds (VOC) in furniture products is significantly affected by the dynamics of environmental parameters. Among them, the release characteristics of harmful components such as formaldehyde and benzene series show nonlinear characteristics under changes in temperature and humidity, which leads to severe challenges for traditional detection systems.

[0003] Existing technologies usually use fixed environmental compensation coefficients or single parameter calibration methods, which makes it difficult to dynamically capture the nonlinear effects of the synergistic effects of temperature and humidity on sensor characteristics. That is, when the ambient temperature and humidity change synergistically, especially in actual application scenarios with large temperature differences between day and night and frequent humidity changes, the physical properties and electrochemical responses of the sensor's sensitive materials will produce time-varying drifts, and traditional compensation models are often unable to accurately decouple the complex relationship between environmental interference and actual VOC emissions, resulting in systematic deviations between measured values ​​and actual emissions. In addition, due to differences in physical and chemical properties, different VOC components also have specific response mechanisms to the effects of temperature and humidity, and traditional methods lack the ability to jointly model material properties and environmental parameters, resulting in systematic errors in compensated measurements.

[0004] Therefore, an optimized furniture harmful substance detection and analysis solution is desired. Summary of the invention

[0005] In order to solve the above-mentioned technical problems, the present application is proposed. The embodiment of the present application provides a furniture harmful substance detection and analysis system, which first synchronously collects the preliminary VOC measurement values ​​of the furniture and the corresponding temperature and humidity time series data, and then calculates the synergistic effect value of temperature and humidity at each time point, and extracts the text description of the physical and chemical properties of the VOC components and converts them into quantifiable features through semantic encoding. Subsequently, an independent soft-constrained deep correlation analysis is performed to capture the specific response laws of the VOC components under dynamic temperature and humidity, and finally a compensation coefficient is dynamically generated to perform nonlinear correction on the preliminary measurement values. In this way, through the fusion of environmental coupling dynamic modeling and material property knowledge, the insufficient characterization of the synergistic effect of temperature and humidity and material differences in the traditional fixed compensation mode is effectively overcome, and the detection accuracy of different VOC components in complex environments is significantly improved.

[0006] According to one aspect of the present application, a furniture harmful substance detection and analysis system is provided, which includes: The preliminary measurement module is used to obtain the preliminary measurement value of furniture VOC; An environmental parameter acquisition module, used to receive a temperature time series data set and a humidity time series data set collected by a temperature sensor and a humidity sensor during a measurement process; A data extraction module, used to extract a preliminary measurement value of a first VOC component from the furniture VOC preliminary measurement value, and extract a property text description of the first VOC component from a backend database by querying; A compensation and correction module is used to perform response analysis on the temperature time series data set, the humidity time series data set and the text description of the properties of the first VOC component to compensate and correct the preliminary measurement value of the first VOC component, wherein the compensation and correction module includes: a collaborative analysis unit, used to perform a collaborative effect time series analysis on the temperature time series data set and the humidity time series data set to obtain a temperature and humidity collaborative time series feature; a correction unit, used to obtain a corrected measurement value of the first VOC component based on a soft constraint property-environmental response association representation between the temperature and humidity collaborative time series feature and the text description of the properties of the first VOC component.

[0007] Compared with the prior art, the present application provides a furniture harmful substance detection and analysis system, which first synchronously collects the preliminary VOC measurement values ​​of the furniture and the corresponding temperature and humidity time series data, and then calculates the synergistic effect value of temperature and humidity at each time point, and extracts the text description of the physical and chemical properties of the VOC components and converts them into quantifiable features through semantic encoding. Subsequently, an independent soft-constrained deep correlation analysis is performed to capture the specific response laws of the VOC components under dynamic temperature and humidity, and finally a compensation coefficient is dynamically generated to perform nonlinear correction on the preliminary measurement values. In this way, through the fusion of environmental coupling dynamic modeling and material property knowledge, the insufficient representation of the synergistic effect of temperature and humidity and material differences in the traditional fixed compensation mode can be effectively overcome, and the detection accuracy of different VOC components in complex environments can be significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 4 is a system block diagram of a furniture harmful substance detection and analysis system according to an embodiment of the present application.

[0010] Figure 2 Schematic diagram of data flow of the furniture harmful substance detection and analysis system according to an embodiment of the present application.

[0011] Figure 34 is a block diagram of a compensation and correction module in a furniture harmful substance detection and analysis system according to an embodiment of the present application.

[0012] Figure 4 4 is a block diagram of a collaborative analysis unit in a furniture harmful substance detection and analysis system according to an embodiment of the present application.

[0013] Figure 5 4 is a block diagram of a correction unit in a furniture harmful substance detection and analysis system according to an embodiment of the present application.

[0014] Figure 6 4 is a block diagram of a property-environment response association subunit in a furniture harmful substance detection and analysis system according to an embodiment of the present application.

[0015] Figure 7 4 is a block diagram of a secondary subunit of inter-modal independence soft constraint in a furniture harmful substance detection and analysis system according to an embodiment of the present application. DETAILED DESCRIPTION

[0016] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0017] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0018] In addition, in order to better illustrate the present application, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0019] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0020] As consumers pay more attention to the health of indoor environments, the detection of harmful substances in furniture has become the key to ensuring the safety of the living environment. Volatile organic compounds (VOCs) in furniture are significantly affected by changes in temperature and humidity. In particular, the release of harmful substances such as formaldehyde and benzene under different environmental conditions shows nonlinear characteristics, which poses a challenge to traditional detection systems. Existing technologies are difficult to accurately capture the nonlinear effects of temperature and humidity changes on sensors, resulting in systematic deviations between measured values ​​and actual release amounts. In addition, the response mechanisms of different VOCs vary due to differences in physical and chemical properties. Traditional methods fail to effectively combine material properties with environmental parameters and cannot eliminate errors after compensation.

[0021] In response to the above technical problems, the technical concept of this application is to use AI-based data analysis and natural language processing technology to build a dynamic environmental coupling analysis and material property response model to achieve accurate compensation under the synergistic effect of temperature and humidity. Specifically, first, the preliminary measurement value of furniture VOC and the corresponding temperature and humidity time series data are collected synchronously, and the synergistic effect value of temperature and humidity at each time point (such as the accelerating effect of high temperature and high humidity environment on formaldehyde release) is calculated to generate time series features that characterize the dynamic coupling effect of the environment; at the same time, the text description of the physical and chemical properties of the VOC components (such as "benzene is easily soluble in lipid substances") is extracted and converted into quantifiable features through semantic coding. Subsequently, the environmental time series features and the material property vectors are subjected to independent soft constraint deep correlation analysis to capture the specific response laws of VOC components under dynamic temperature and humidity (such as the adsorption hysteresis effect of toluene when humidity changes suddenly). Finally, the compensation coefficient is dynamically generated to perform nonlinear correction on the preliminary measurement value. This processing effectively overcomes the insufficient characterization of the synergistic effect of temperature and humidity and material differences in the traditional fixed compensation mode through the fusion of environmental coupling dynamic modeling and material property knowledge, and significantly improves the detection accuracy of different VOC components in complex environments.

[0022] In particular, the present application proposes a furniture harmful substance detection and analysis system. Figure 1 4 is a system block diagram of a furniture harmful substance detection and analysis system according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of a furniture harmful substance detection and analysis system according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the furniture harmful substance detection and analysis system 100 of the embodiment of the present application, it includes: a preliminary measurement module 110, which is used to obtain a preliminary measurement value of furniture VOC; an environmental parameter acquisition module 120, which is used to receive a temperature time series data set and a humidity time series data set in a measurement process acquired by a temperature sensor and a humidity sensor; a data extraction module 130, which is used to extract a preliminary measurement value of a first VOC component from the preliminary measurement value of the furniture VOC, and extract a text description of the properties of the first VOC component from a background database by query; a compensation correction module 140, which is used to perform response analysis on the temperature time series data set, the humidity time series data set and the text description of the properties of the first VOC component to compensate and correct the preliminary measurement value of the first VOC component.

[0023] In the above-mentioned furniture harmful substance detection and analysis system 100, the preliminary measurement module 110 is used to obtain the preliminary measurement value of furniture VOC. It should be understood that in the technical solution of the present application, the concentration level of VOC in the air around the furniture is monitored and recorded in real time based on sensing technology. Among them, the sensing technology consists of a series of highly sensitive chemical sensors, which are designed to respond to a variety of common VOC components, including but not limited to formaldehyde, benzene and other substances that pose a potential threat to human health. Considering the cross-sensitivity problem that may exist among different VOC components, multiple types of sensors are often integrated inside the module to achieve effective distinction and quantification of complex mixtures. Specifically, when performing preliminary measurements, the furniture to be tested must first be placed in a closed or semi-enclosed space, which needs to have good sealing performance so as to minimize the impact of external air flow on the test results. Then, within a set time period, air samples are continuously collected by sensors installed in the space and converted into electrical signals. These signals are then sent to the data acquisition system for digital conversion after a series of pre-processing steps such as amplification and filtering, and finally form raw measurement data for subsequent analysis. In this way, the concentration information of various VOC components released by furniture is captured through precise sensor technology. These data reflect the emission of harmful substances from furniture products under specific conditions.

[0024] In the above-mentioned furniture harmful substance detection and analysis system 100, the environmental parameter acquisition module 120 is used to receive the temperature time series data set and the humidity time series data set collected by the temperature sensor and the humidity sensor during the measurement process. It should be understood that in the furniture harmful substance detection and analysis environment, temperature and humidity are two key environmental variables, which have a significant impact on the release rate and pattern of VOC in furniture materials. For example, under high temperature and high humidity conditions, some VOC components such as formaldehyde may accelerate the release, while under low temperature and low humidity conditions, their release rate may be slowed down. In addition, different types of VOC components also have different response mechanisms to changes in temperature and humidity. This difference makes it difficult to accurately describe and predict the actual release of VOCs by relying solely on a single environmental factor or a fixed compensation model. By real-time monitoring and recording the temperature and humidity changes during the entire measurement process, the specific impact of these environmental factors on the VOC release behavior can be more comprehensively captured, thereby improving the accuracy and reliability of the detection results. Specifically, the temperature sensor and humidity sensor of the technical solution of the present application have the characteristics of high precision, fast response time and good long-term stability to ensure that small environmental changes can be accurately captured. In addition, considering the needs of different application scenarios, it is necessary to select temperature sensors and humidity sensor products with wide range and strong anti-interference ability. It is worth mentioning that in order to ensure the integrity and accuracy of the data, the distance between sensors needs to be reasonably set to avoid data redundancy caused by too close distance or signal attenuation caused by too far distance. At the same time, the sensors need to be calibrated regularly to ensure that they are always in the best working state. Once the sensors start working, they will continuously generate a large amount of raw data streams. These data need to go through a series of preprocessing steps before they can be effectively used. The first is signal amplification and filtering. By amplifying and filtering the original electrical signal, noise interference can be effectively removed, the signal-to-noise ratio can be improved, and subsequent data analysis can be more reliable. Subsequently, the preprocessed data will be sent to the data acquisition system, which is responsible for converting analog signals into digital signals, and sampling and storing the data at predetermined time intervals to obtain temperature time series data sets and humidity time series data sets. In order to ensure the real-time and synchronization of the data, high-speed data acquisition cards or dedicated data acquisition equipment are usually used to complete this task. In terms of data transmission, this technology relies on two methods: wired transmission and wireless transmission. Wired transmission usually relies on the traditional RS-485 bus or other industrial standard interfaces, which has high stability and security and is suitable for application scenarios in fixed places. Wireless transmission, such as ZigBee, Wi-Fi or Bluetooth protocols, is easy to deploy and maintain, and can achieve remote monitoring and data sharing, greatly improving the flexibility and scalability of the system.

[0025] In the above-mentioned furniture harmful substance detection and analysis system 100, the data extraction module 130 is used to extract the preliminary measurement value of the first VOC component from the furniture VOC preliminary measurement value, and extract the property text description of the first VOC component from the background database by query. It should be understood that different VOC components have different physical and chemical properties, and their response mechanisms affected by environmental factors such as temperature and humidity are specific. The furniture VOC preliminary measurement value is a comprehensive measurement result of multiple VOC components. To accurately analyze the release of each VOC component in a dynamic environment, the measurement values ​​of different components need to be processed separately. At the same time, the physical and chemical properties of each VOC component have an important influence on its release characteristics under temperature and humidity changes. These property information is usually stored in the background database in text form, so it is necessary to extract the property text description of the corresponding component from the database for further analysis. Specifically, in a specific embodiment of the present application, first, the collected furniture VOC preliminary measurement value is input into a preprocessing mechanism, which is responsible for removing noise and outliers to ensure that subsequent analysis is based on high-quality data. Then, the pre-processed furniture VOC preliminary measurement values ​​are jointly analyzed in the time and frequency domains, and the characteristic response patterns of each VOC component in a specific frequency band are extracted using wavelet packet transform, and then the fingerprints of different components are identified by combining gas chromatography-mass spectrometry (GC-MS) technology to extract the preliminary measurement value of the first VOC component. Among them, for complex scenarios with overlapping spectral peaks, this application uses a non-negative matrix decomposition algorithm for blind source separation to extract the preliminary measurement value of the first VOC component. Subsequently, based on the unique identifier of the first VOC component, this application queries the background database to obtain a text description of the properties of the first VOC component including physical and chemical properties such as boiling point, polarity, and adsorption coefficient.

[0026] In the above-mentioned furniture harmful substance detection and analysis system 100, the compensation and correction module 140 is used to perform response analysis on the temperature time series data set, the humidity time series data set and the text description of the properties of the first VOC component to compensate and correct the preliminary measurement value of the first VOC component. It should be understood that since traditional methods usually adopt fixed environmental compensation coefficients or single parameter calibration methods, it is difficult to dynamically capture the nonlinear effect of the synergistic effect of temperature and humidity on sensor characteristics. The impact of changes in temperature and humidity in the actual environment on sensor characteristics has significant time-varying synergy, and the temperature or humidity time series data alone cannot fully characterize the nonlinear drift phenomenon caused by this composite effect. Therefore, in the technical solution of the present application, by performing response analysis on the temperature time series data set, the humidity time series data set and the text description of the properties of the first VOC component, the insufficient characterization of the synergistic effect of temperature and humidity and material differences in the traditional fixed compensation mode is overcome, and the detection accuracy of different VOC components in complex environments is significantly improved.

[0027] Figure 3 FIG. 1 is a block diagram of a compensation and correction module in a furniture harmful substance detection and analysis system according to an embodiment of the present application. Figure 3 As shown, the compensation correction module 140 includes: a collaborative analysis unit 141, which is used to perform a collaborative effect time series analysis on the temperature time series data set and the humidity time series data set to obtain a temperature and humidity collaborative time series feature; a correction unit 142, which is used to obtain a corrected measurement value of the first VOC component based on a soft constraint property-environmental response association representation between the temperature and humidity collaborative time series feature and the property text description of the first VOC component.

[0028] Figure 4 FIG. 1 is a block diagram of a collaborative analysis unit in a furniture harmful substance detection and analysis system according to an embodiment of the present application. Figure 4 As shown, in an embodiment of the present application, the collaborative analysis unit 141 includes: a temperature value collaborative effect value synchronization calculation subunit 1411, which is used to perform time synchronization on the temperature time series data set and the humidity time series data set and calculate the collaborative effect value between the temperature value and the humidity value at the same time point to obtain a time queue of the temperature value collaborative effect value; a collaborative time series feature extraction subunit 1412, which is used to extract the temperature and humidity collaborative time series feature from the time queue of the temperature value collaborative effect value to obtain a temperature and humidity collaborative time series feature encoding vector as the temperature and humidity collaborative time series feature.

[0029] Specifically, the temperature value synergy effect value synchronization calculation subunit 1411 is used to synchronize the temperature time series data set and the humidity time series data set and calculate the synergy effect value between the temperature value and the humidity value at the same time point to obtain the time queue of the temperature value synergy effect value. The process is expressed by the formula:

[0030] in, It is the first The temperature value at a time point, It is the first The humidity value at a time point, is a sine function, is to calculate the absolute value, yes and The synergistic effect value of the temperature values ​​between them. It should be understood that, considering that the effects of temperature and humidity changes on sensor characteristics are time-varying and synergistic, a single temperature or humidity time series data cannot characterize the nonlinear drift phenomenon caused by the superposition of the two. For example, when a high temperature environment is accompanied by a sudden drop in humidity, the sensor sensitive membrane may cause a sudden change in gas permeability due to rapid dehydration. This composite environmental effect is far beyond the linear superposition range of a single parameter change. If the temperature and humidity data are processed independently, the modulation effect of humidity changes on the temperature sensitivity curve is ignored, making it impossible to accurately model the response distortion of the sensor under high temperature and high humidity conditions during the rainy season. Based on this, the present application synchronizes the temperature time series data set and the humidity time series data set and calculates the synergistic effect value between the temperature value and the humidity value at the same time point to obtain a time queue of the synergistic effect value of the temperature value. In the present application, this step eliminates the sensor sampling clock deviation through strict time synchronization and constructs a quantitative model of the joint action of temperature and humidity parameters. Its necessity stems from the spatiotemporal correlation characteristics of temperature and humidity fluctuations in the real environment. For example, the afternoon temperature peak is often accompanied by an anti-phase fluctuation pattern of the relative humidity valley. This dynamic coupling relationship directly determines the nonlinear trajectory of the VOC sensor baseline drift.

[0031] In an embodiment of the present application, the synergistic time series feature extraction subunit 1412 is used to: use a time series feature extractor based on an LSTM model to extract the temperature and humidity synergistic time series features from the time queue of the temperature value synergistic effect value to obtain the temperature and humidity synergistic time series feature encoding vector. It should be understood that the dynamic synergistic effect of environmental parameters on the sensor has significant time series correlation characteristics. Since the evolution pattern of the temperature and humidity synergistic effect in the time dimension directly affects the VOC release dynamics, such as the cumulative effect of the duration of high temperature and high humidity environment on the sensitivity attenuation of formaldehyde sensors, or the secondary release of benzene series caused by the temperature rise after the humidity suddenly changes, it is impossible to capture the time-varying propagation law of environmental interference by simply relying on discrete synergistic effect values. The time queue of the temperature value synergistic effect value records the relevant information of the synergistic effect of temperature and humidity at different time points, but this information is relatively primitive and scattered. In order to more effectively use this information to analyze the release of volatile organic compounds (VOCs) in a dynamic temperature and humidity environment, it is necessary to extract key features that can represent the synergistic changes of temperature and humidity from the time queue. Therefore, the present application extracts the temperature and humidity coordinated time series features from the time queue of the temperature value synergy effect value to extract the key features that can represent the coordinated change of temperature and humidity, and obtains the temperature and humidity coordinated time series feature encoding vector. In particular, in a specific example of the present application, a time series feature extractor based on the LSTM model is used to extract the temperature and humidity coordinated time series features from the time queue of the temperature value synergy effect value to obtain the temperature and humidity coordinated time series feature encoding vector. Specifically, the long short-term memory network (LSTM), as a special recurrent neural network (RNN), can effectively process time series data with long-term dependencies. For example, when a composite environmental pattern in which the temperature continues to rise and the humidity steps down is detected, the forget gate of the LSTM can automatically identify the residual influence weight of the previous high temperature and high humidity stage on the current sensor characteristics, while the input gate captures the update intensity of the current temperature and humidity coordinated mutation on the feature vector. This time series modeling capability enables the encoding vector to characterize the enhanced trajectory of the hysteresis effect of the benzene series sensor in non-steady-state environments such as "humidity slow drop accompanied by temperature oscillation rise", or the modulation law of the formaldehyde diffusion rate in "periodic fluctuation of humidity during temperature stability". Through the time-delivery property of LSTM hidden states, the system can trace the propagation path of environmental coupling on the time axis, thereby providing a dynamic environmental feature expression with temporal causal correlation for the subsequent compensation model.

[0032] Figure 5 FIG. 1 is a block diagram of a correction unit in a furniture harmful substance detection and analysis system according to an embodiment of the present application. Figure 5As shown, in an embodiment of the present application, the correction unit 142 includes: a first VOC component property semantic encoding subunit 1421, which is used to semantically encode the property text description of the first VOC component to obtain a first VOC component property semantic encoding vector; a property-environment response association subunit 1422, which is used to perform property-environment hidden state independence soft constraint response analysis on the first VOC component property semantic encoding vector and the temperature and humidity collaborative temporal feature encoding vector to obtain a property-environment deep state response association encoding vector as the soft constraint property-environment response association representation; a first VOC component correction subunit 1423, which is used to obtain a corrected measurement value of the first VOC component based on the property-environment deep state response association encoding vector.

[0033] Specifically, the first VOC component property semantic encoding subunit 1421 is used to semantically encode the property text description of the first VOC component to obtain the first VOC component property semantic encoding vector. It should be understood that, considering that the property text description of the first VOC component contains rich semantic information, these semantic information are crucial to understanding the behavior of the component under different environmental conditions. For example, a text description such as "benzene is easily soluble in lipid substances", in which words such as "easy to dissolve in" convey certain physical and chemical properties of benzene. Therefore, in order to capture and mine these semantic information and convert them into numerical vectors, the present application obtains the first VOC component property semantic encoding vector by semantically encoding the property text description of the first VOC component. In particular, in a specific example of the present application, first, the property text description of the first VOC component is preprocessed, including removing stop words (such as "of", "is" and other common but meaningless words), word segmentation (splitting sentences into words or phrases), stemming or word form restoration (normalizing different forms of the same word to the basic form) operations, thereby obtaining a clean and structured text dataset. Next, these keywords are semantically embedded using a deep learning model. Semantic embedding is a technique that maps high-dimensional sparse text data to a low-dimensional dense vector space, so that words with similar semantics are close in the vector space. Common embedding methods include Word2Vec, GloVe, and BERT. Taking BERT as an example, it captures contextual information through a bidirectional Transformer architecture and can generate more accurate semantic representations. The input to the BERT model is the processed keyword sequence, and the output is the high-dimensional vector representation corresponding to each keyword. Next, the BERT model is used to more accurately capture the key information in the semantic description and convert it into a high-quality semantic encoding vector. Finally, the vector representations of all keywords are aggregated to form a comprehensive semantic encoding vector of the first VOC component properties. Common aggregation methods include simple averaging, weighted averaging, or using recurrent neural networks (RNNs) for sequence modeling. Among them, when the natural language processing model performs deep semantic analysis on the text "saturated vapor pressure changes exponentially with temperature", it can not only extract the quantitative features of temperature dependence, but also identify the potential impact pattern of this feature on the sensor baseline drift. This processing enables the subsequent hidden state response analysis to associate the text semantics of "aniline substances are easily oxidized" with the polarization phenomenon of the sensor electrode in a high humidity environment, thereby dynamically adjusting the generation logic of the compensation coefficient.

[0034] Specifically, the property-environment response association subunit 1422 is used to perform property-environment hidden state independence soft constraint response analysis on the first VOC component property semantic coding vector and the temperature and humidity collaborative time series feature coding vector to obtain the property-environment deep state response association coding vector as the soft constraint property-environment response association representation. It should be understood that, considering that in the actual furniture VOC release scenario, the release of VOC components is not only affected by temperature and humidity environmental factors, but also closely related to its own physical and chemical properties. When processing material properties and temperature and humidity time series characteristics, traditional methods often simply splice different modal data directly, resulting in the intrinsic association between the hydrophilicity description of the material and the humidity fluctuation characteristics being submerged in redundant information. For example, when the system simultaneously processes the semantic feature of "formaldehyde is easily soluble in water" and the time series pattern of humidity drop, it is difficult for traditional fusion methods to identify the specific modulation relationship of humidity change on the response of the hydrophilic VOC sensor. Based on this, the present application performs property-environment hidden state independence soft constraint response analysis on the first VOC component property semantic coding vector and the temperature and humidity coordinated time series feature coding vector to obtain the property-environment deep state response association coding vector. In this way, the property-environment hidden state independence soft constraint response analysis is aimed at solving the two core problems in cross-modal feature interaction: first, there is scale heterogeneity between the semantic space of material properties and the environmental time series feature space, such as the text description of the boiling point of the material and the numerical fluctuation of the temperature time series show nonlinear differences in feature distribution; second, there are both complementary associations between modes (such as the synergistic effect of humidity changes and material solubility) and independent information (such as the inherent influence of material molecular weight on sensor response), and information optimization needs to be achieved through learnable independence constraints.

[0035] Figure 6 FIG. 1 is a block diagram of a property-environment response association subunit in a furniture harmful substance detection and analysis system according to an embodiment of the present application. Figure 6As shown, in an embodiment of the present application, the property-environment response association subunit 1422 includes: an inter-modal independence soft constraint secondary subunit 1422-1, which is used to perform principal component analysis and linear transformation on the temperature and humidity collaborative time series feature coding vector to obtain a set of temperature and humidity collaborative time series principal component feature linear transformation coding vectors, and then perform inter-modal independence soft constraint calculation on the set of temperature and humidity collaborative time series principal component feature linear transformation coding vectors and the first VOC component property semantic coding vector to obtain a set of property-environment deep state inter-modal independence soft constraint factors; a property-environment deep state response secondary subunit 1422-2, which is used to perform adaptive response analysis on the first VOC component property semantic coding vector and the set of temperature and humidity collaborative time series principal component feature linear transformation coding vectors based on the set of property-environment deep state inter-modal independence soft constraint factors to obtain the property-environment deep state response association coding vector.

[0036] Figure 7 FIG. 4 is a block diagram of a secondary subunit of inter-modal independence soft constraint in a furniture harmful substance detection and analysis system according to an embodiment of the present application. Figure 7 As shown, in an embodiment of the present application, the inter-modal independence soft constraint secondary subunit 1422-1 includes: a temperature and humidity collaborative time series principal component analysis tertiary subunit 1422-11, which is used to perform principal component analysis on the temperature and humidity collaborative time series feature coding vector to obtain a set of temperature and humidity collaborative time series principal component feature coding vectors; a temperature and humidity collaborative time series principal component feature linear transformation tertiary subunit 1422-12, which is used to perform linear transformation on each temperature and humidity collaborative time series principal component feature coding vector in the set of the temperature and humidity collaborative time series principal component feature coding vector to obtain a set of the temperature and humidity collaborative time series principal component feature linear transformation coding vectors, wherein each temperature and humidity collaborative time series principal component feature linear transformation coding vector in the set of the temperature and humidity collaborative time series principal component feature linear transformation coding vectors is a linear transformation of the temperature and humidity collaborative time series principal component feature coding vectors. The temporal principal component feature linear transformation coding vector and the first VOC component property semantic coding vector have the same characteristic scale; the property-environment depth state inter-modal independence coding tertiary sub-unit 1422-13 is used to perform inter-modal independence coding on each temperature and humidity collaborative temporal principal component feature linear transformation coding vector in the set of the first VOC component property semantic coding vector and the temperature and humidity collaborative temporal principal component feature linear transformation coding vector to obtain a set of property-environment depth state inter-modal independence coding matrices; the independence soft constraint factor calculation tertiary sub-unit 1422-14 is used to calculate a set of property-environment depth state inter-modal independence soft constraint factors based on the set of property-environment depth state inter-modal independence coding matrices.

[0037] Specifically, the temperature and humidity coordinated time series principal component analysis tertiary subunit 1422-11 is used to perform principal component analysis on the temperature and humidity coordinated time series feature coding vector to obtain a set of temperature and humidity coordinated time series principal component feature coding vectors, which is expressed as follows using the temperature and humidity coordinated time series principal component analysis formula:

[0038] in, is the temperature and humidity coordinated time series feature encoding vector, For principal component analysis, is the The number of eigenvalues ​​in , is the covariance matrix of the temperature and humidity coordinated time series samples, is the orthogonal matrix of the principal component of the temperature and humidity coordinated time series, that is, the set of principal component feature coding vectors of the temperature and humidity coordinated time series. , , and are the first, second, and third principal component feature encoding vectors of the temperature and humidity collaborative time series. and The principal component feature encoding vector of temperature and humidity coordinated time series, is the temperature and humidity coordinated time series diagonal matrix, The diagonal elements of the matrix are , , and The temperature and humidity coordinated time series diagonal matrix, , , and They are , , and Corresponding eigenvalues. It should be understood that due to the nonlinear superposition effect of temperature and humidity in the time series dimension, the characteristic dimensions generated by their synergy often show high redundancy. For example, temperature gradient changes and humidity oscillations may produce similar characteristic response patterns on different time scales. This redundancy will not only increase the computational load of subsequent modal interactions, but also cause the characteristic space to be distorted due to multicollinearity problems, so that the correlation between sensor response and VOC emission is masked by noise. Therefore, in the technical solution of the present application, by performing principal component analysis on the temperature and humidity collaborative time series feature coding vector, the high-dimensional temperature and humidity collaborative time series feature coding vector is reduced in dimension and decoupled in features, which is essentially information purification of modal data. The application of principal component analysis here is essentially to construct a feature space transformation mechanism that enhances information entropy. By projecting the original temperature and humidity collaborative time series feature coding vector onto a set of ordered low-dimensional principal component axes through orthogonal transformation, a small number of head principal components can effectively capture the main variation information of the data and screen out the feature subset with the highest explanatory power for VOC emission variation. For example, when there is a dominant variation direction in the temperature and humidity synergistic effect, PCA can automatically extract the characteristic components of this direction, so that the subsequent inter-modal independence constraint analysis can focus on the key modes that truly affect the material response. This dimensionality reduction process essentially completes the "elimination of false and retention of true" of environmental interference signals, so that the time-varying law of sensor baseline drift can be presented in a purer form. In this way, the redundant associations and multicollinearity problems between temperature and humidity parameters can be eliminated through feature decoupling, and the high-variance principal components can be retained to maintain a high concentration of original information while reducing the feature dimension. In addition, the more compact feature subspace reconstructed by PCA alleviates the fusion conflict between environmental parameters and semantic information of material properties in traditional methods, so that text descriptions such as "benzene is easily soluble in lipid substances" can establish accurate associations with humidity mutation features at a deep semantic level, so as to focus more on complementary information rather than repeated fragments, thereby improving information utilization efficiency and the discriminability of interaction results.

[0039] Specifically, the temperature and humidity coordinated time series principal component feature linear transformation tertiary subunit 1422-12 is used to linearly transform each temperature and humidity coordinated time series principal component feature coding vector in the set of the temperature and humidity coordinated time series principal component feature coding vector to obtain the set of the temperature and humidity coordinated time series principal component feature linear transformation coding vector, wherein each temperature and humidity coordinated time series principal component feature linear transformation coding vector in the set of the temperature and humidity coordinated time series principal component feature linear transformation coding vector has the same characteristic scale as the first VOC component property semantic coding vector, and is expressed by the temperature and humidity coordinated time series principal component feature linear transformation formula as follows:

[0040] in, is the transpose operation, To perform a linear transformation operation on the set of principal component feature encoding vectors of the temperature and humidity collaborative time series, , , and They are the first, second, and third linear transformation encoding vectors of the temperature and humidity coordinated time series principal component features. and The linear transformation encoding vector of the principal component characteristics of the temperature and humidity coordinated time series, It is a set of linear transformation encoding vectors of the main component features of the temperature and humidity coordinated time series. It should be understood that, considering that the main component features of the temperature and humidity coordinated time series usually present Gaussian distribution characteristics, and the semantic encoding vector of the first VOC component property after semantic encoding has a sparse or asymmetric distribution, this mismatch between dimensions and distribution, if the data of different modalities with different scales are directly input into the modal interaction layer for interaction without processing, will lead to the direction of parameter optimization in the gradient update process. For example, the high amplitude fluctuation of the high temperature feature may mask the weak change signal of the semantic feature of "hydrophilic substance", causing the model to reduce the detection sensitivity of humidity-sensitive VOC, thereby leading to an imbalance in contribution and generating information bias. In the technical solution of the present application, the linear transformation can construct a mathematical isomorphic space for cross-modal feature interaction, so as to map the numerical distribution of the main component features of the temperature and humidity coordinated time series to a scale range that matches the semantic encoding vector of the first VOC component property, thereby finely balancing the importance of different modal features and ensuring the fairness and effectiveness of subsequent interactions. This spatial reconstruction essentially establishes a computable association framework between environmental parameters and material properties, providing a mathematical basis for deep feature interaction. This is not only an adjustment of the numerical range, but also promotes the initial alignment of cross-modal feature spaces, laying the foundation for deeper feature semantic alignment.

[0041] Specifically, the property-environment depth state inter-modal independence encoding third-level subunit 1422-13 is used to perform inter-modal independence encoding on each temperature and humidity coordinated temporal principal component feature linear transformation encoding vector in the set of the first VOC component property semantic encoding vector and the temperature and humidity coordinated temporal principal component feature linear transformation encoding vector to obtain a set of property-environment depth state inter-modal independence encoding matrices, which is expressed as a property-environment depth state inter-modal independence encoding formula:

[0042] in, and is a feature mapping function, such as a linear mapping or a nonlinear kernel function, is the semantic encoding vector of the first VOC component property, yes Length, yes and The inter-modal independence encoding matrix of the properties between the environmental deep states. It should be understood that when the system needs to fuse the linear transformation encoding vector of the temperature and humidity collaborative time series principal component features and the semantic encoding vector of the properties of the first VOC component, traditional methods often find it difficult to distinguish between the complementary information and redundant noise in the two modalities. If feature fusion is performed directly, the noise component will interfere with the model's capture of key association rules, resulting in systematic deviations when the compensation coefficients are generated. This step constructs an information optimization mechanism through inter-modal independence encoding, and makes modal independence explicit as a measurable and learnable constraint, which transcends the limitations of implicit learning of modal relationships in traditional methods. Inter-modal independence encoding aims to construct a learnable independence representation space, rather than simply calculating independence indicators, so as to pay more attention to the complementary information between modalities, because the essence of modal fusion is to utilize complementary rather than redundant information.

[0043] In an embodiment of the present application, the independence soft constraint factor calculation tertiary subunit 1422-14 is used to calculate a set of property-environment depth state inter-modal independence soft constraint factors based on the set of property-environment depth state inter-modal independence coding matrices, including: performing intrinsic optimization based on the canonical field on each property-environment depth state inter-modal independence coding matrix in the set of property-environment depth state inter-modal independence coding matrices to obtain a set of optimized property-environment depth state inter-modal independence coding matrices; using the square of the Frobenius norm of each optimized property-environment depth state inter-modal independence coding matrix in the set of optimized property-environment depth state inter-modal independence coding matrices as the set of property-environment depth state inter-modal independence soft constraint factors.

[0044] Specifically, each property-environment depth state modal independence encoding matrix in the set of property-environment depth state modal independence encoding matrices is subjected to eigenoptimization based on the canonical field to obtain a set of optimized property-environment depth state modal independence encoding matrices, which is expressed as the property-environment eigenoptimization formula:

[0045]

[0046]

[0047] in, yes Each eigenvalue of yes The corresponding eigenvectors, It is added by location point. is matrix multiplication, yes The corresponding topological gauge field vector, yes The optimized property-environment depth state modality independence encoding matrix after optimization. It should be understood that based on the characteristics of the property-environment depth state modality independence encoding matrix as a topological order matrix, it can characterize the eigenvector and The neighbor-non-neighbor coupling characteristics are obtained under the framework of short-range-long-range eigenvalue scale correlation, and the global correlation topological configuration is realized through asymmetric integration. In order to maintain the structural stability of the intrinsic gauge field representation of the asymmetric system, thereby providing a basis for the subsequent refined modal fusion, the eigenvectors composed of the eigenvalues ​​of the property-environment deep state modal independence encoding matrix are defined as the eigenvalue gauge field representation under the topological order, thereby constructing a topological invariant stabilization mechanism under asymmetric conditions. In obtaining the eigenvectors corresponding to the property-environment deep state modal independence encoding matrix After that, the high-order trace terms are ignored in the framework of the eigenstatement of non-limiting distribution, and the linear gauge field approximation property - the topological gauge field vector corresponding to the independence encoding matrix between the environmental depth state modes is derived through the gauge field-connection 1-form. Then, for the phase accumulation process of topological correlation, the topological gauge field vector corresponding to the property-environment depth state modal independence encoding matrix is ​​used. The autocorrelation matrix of the property-environment depth state modal independence encoding matrix is ​​optimized in the gauge field to ensure that the asymmetric system is robust to local configuration disturbances, and finally an optimized property-environment depth state modal independence encoding matrix representation with structural stability is obtained.

[0048] Specifically, the square of the Frobenius norm of each optimized property-environment depth state inter-modal independence encoding matrix in the set of the optimized property-environment depth state inter-modal independence encoding matrix is ​​used as the set of the property-environment depth state inter-modal independence soft constraint factors, and is expressed in the property-environment depth state inter-modal independence soft constraint formula as follows:

[0049] in, To calculate the square of the Frobenius norm, yes The corresponding soft constraint factor of inter-modal independence of property-environment deep state. It should be understood that the introduction of the soft constraint factor of inter-modal independence of property-environment deep state reflects the refined control of the modal fusion strategy. It guides the model to perform information fusion in a flexible way rather than rigidly excluding dependencies. The soft constraint factor of inter-modal independence of property-environment deep state has become a key bridge connecting modal independence modeling and feature interaction, enabling the model to adaptively weigh the independence and correlation between modalities. Compared with hard constraints, this flexible constraint is more in line with the complexity of real-world data and can make full use of various modal information. In the specific implementation, the design of the soft constraint factor of inter-modal independence of property-environment deep state needs to carefully consider the information aggregation strategy and extract key independence information from the optimization of the encoding matrix of inter-modal independence of property-environment deep state. The dynamic adaptability of the soft constraint factor of inter-modal independence of property-environment deep state is its core advantage. It can dynamically adjust the constraint strength according to the input features, strengthen the focus on independence when independence is high, and relax it relatively when independence is low. This dynamic characteristic gives the architecture stronger flexibility and generalization ability.

[0050] In an embodiment of the present application, the property-environment depth state response secondary sub-unit 1422-2 is used to: perform feature interaction response on each temperature and humidity collaborative time-series principal component feature linear transformation coding vector in the set of the first VOC component property semantic coding vector and the temperature and humidity collaborative time-series principal component feature linear transformation coding vector to obtain a set of property-environment depth state principal component fine-grained modal response interaction coding vectors; based on the set of property-environment depth state modal independence soft constraint factors, perform adaptive response analysis on the set of property-environment depth state principal component fine-grained modal response interaction coding vectors to obtain the property-environment depth state response association coding vector.

[0051] Specifically, each temperature and humidity coordinated time series principal component feature linear transformation coding vector in the set of the first VOC component property semantic coding vector and the temperature and humidity coordinated time series principal component feature linear transformation coding vector is subjected to feature interactive response to obtain a set of property-environment deep state principal component fine-grained modal response interactive coding vectors, which is expressed as a property-environment feature interactive response formula:

[0052] in, It is the point multiplication by position. It is divided by position. For cascade operation, and They are The corresponding interaction response weight matrix and interaction response bias vector, yes and The property-environment deep state principal component fine-grained modal response interaction encoding vector between them. It should be understood that the fine-grained feature interactive response realizes the deep alignment and complementary information mining of cross-modal features through mathematical means. The first VOC component property semantic encoding vector carries the physicochemical properties of the VOC component, while the temperature and humidity coordinated time series principal component feature linear transformation encoding vector represents the dynamic evolution pattern of environmental parameters. Through operations such as dot multiplication, dot division, and cascade, the model can go deep into the feature dimension level and analyze the potential connection between the two in terms of numerical scale, change trend, and semantic association. This fine-grained interaction goes deep into the feature dimension level, excavates nonlinear and hierarchical interactions, thereby capturing the complementarity, dependency, and semantic association between different modal features, and generating a more expressive property-environment deep state principal component fine-grained modal response interaction feature representation.

[0053] Specifically, based on the set of the property-environment depth state modal independence soft constraint factors, the set of the property-environment depth state principal component fine-grained modal response interaction coding vectors is adaptively analyzed to obtain the property-environment depth state response association coding vector, which is expressed as the property-environment adaptive response analysis formula:

[0054] in, is the normalization function, is the number of sets of property-environment deep state principal component fine-grained modal response interaction encoding vectors, It is the property-environment deep state response associated coding vector. It should be understood that the property-environment deep state modal independence soft constraint factor is essentially a quantitative indicator of modal independence. It not only reflects the complementarity of different modal features (such as VOC semantic attributes and temperature and humidity time series patterns) in information distribution, but also reveals the potential correlation between the two in physical mechanism. In the technical solution of the present application, under the guidance of the property-environment deep state modal independence soft constraint, the property-environment deep state principal component fine-grained modal response is selectively fused to generate a property-environment deep state response associated coding vector with global awareness. The goal of adaptive response analysis is global information integration and selective fusion. It is not a simple average, but dynamically adjusts the contribution weights of the fine-grained modal responses of the principal components of each property-environment deep state according to the property-environment deep state modal independence soft constraint factor, preferably retaining information with strong independence and suppressing redundant information. The final obtained property-environment deep state response association encoding vector is a refined representation of multimodal features fused under the soft constraint optimization of independence between property-environment deep state modalities. It not only contains rich association information, but also minimizes redundancy and noise, providing efficient and generalized feature representation for subsequent tasks.

[0055] In an embodiment of the present application, the first VOC component correction subunit 1423 is used to: input the property-environment depth state response associated coding vector into a decoder-based compensation estimation module to obtain a first compensation coefficient; and correct the preliminary measurement value of the first VOC component based on the first compensation coefficient to obtain a corrected measurement value of the first VOC component.

[0056] Specifically, the property-environmental deep state response association coding vector is input into the compensation estimation module based on the decoder to obtain the first compensation coefficient. That is, the property-environmental deep state response association coding vector obtained by response association using the first VOC component property semantic coding vector and the temperature and humidity collaborative temporal feature coding vector is decoded to obtain the first compensation coefficient. It should be understood that the compensation estimation module based on the decoder can parse and process such high-dimensional and complex coding vectors, and mine the information related to the measurement value compensation. That is, by inputting the property-environmental deep state response association coding vector into the compensation estimation module based on the decoder, the first compensation coefficient for the first VOC component can be obtained. This compensation coefficient is calculated comprehensively based on the properties of the component and the temperature and humidity environment in which it is located, and can more accurately reflect the influence of environmental factors and material characteristics on the measurement value, thereby providing an accurate compensation basis for the subsequent correction of the measurement value of the first VOC component in the preliminary measurement value of furniture VOC.

[0057] Specifically, the preliminary measurement value of the first VOC component is corrected based on the first compensation coefficient to obtain the corrected measurement value of the first VOC component. In this way, by correcting the preliminary measurement value of the first VOC component based on the first compensation coefficient, the measurement deviation caused by environmental factors and material characteristics can be eliminated or reduced, so that the corrected measurement value obtained is closer to the actual release amount of the component. In this way, the accuracy of the detection of the first VOC component can be improved, and more reliable data support can be provided for accurately evaluating the harmful substances in furniture in the indoor environment. Specifically, during the correction process, the first compensation coefficient and the preliminary measurement value of the first VOC component are multiplied to achieve nonlinear correction of the original data. This step eliminates the systematic deviation of the sensor caused by the synergistic effect of temperature and humidity by dynamically adjusting the dimension and amplitude. For example, when it is detected that the sensitivity of the formaldehyde sensor decreases due to a high temperature and high humidity environment, the compensation coefficient will automatically increase the output value to compensate for the signal attenuation; conversely, if the sensor responds excessively under low temperature and low humidity conditions, the compensation coefficient will reduce the measurement value accordingly. It is worth noting that the correction process is not a simple global scaling, but a refined correction is achieved through the spatial transformation of local response characteristics. The system's built-in adaptive calibration engine continuously monitors the statistical characteristics of the corrected data. If the residual distribution deviates from the preset threshold, the back-propagation mechanism is triggered to update the decoder parameters, thereby further improving the compensation accuracy in subsequent detections. This closed-loop correction mechanism enables the system to continuously optimize the compensation strategy in actual operation, adapting to the time-varying laws of furniture material release characteristics and the complex fluctuations of environmental parameters.

[0058] In summary, the furniture harmful substance detection and analysis system 100 based on the embodiment of the present application is explained, which first synchronously collects the preliminary measurement values ​​of furniture VOC and the corresponding temperature and humidity time series data, and then calculates the synergistic effect value of temperature and humidity at each time point, and extracts the text description of the physical and chemical properties of the VOC components and converts them into quantifiable features through semantic encoding. Subsequently, an independent soft-constrained deep correlation analysis is performed to capture the specific response laws of VOC components under dynamic temperature and humidity, and finally a compensation coefficient is dynamically generated to perform nonlinear correction on the preliminary measurement values. In this way, through the fusion of environmental coupling dynamic modeling and material property knowledge, the insufficient representation of the synergistic effect of temperature and humidity and material differences in the traditional fixed compensation mode can be effectively overcome, and the detection accuracy of different VOC components in complex environments can be significantly improved.

[0059] As described above, the furniture harmful substance detection and analysis system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the furniture harmful substance detection and analysis system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the furniture harmful substance detection and analysis system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the furniture harmful substance detection and analysis system 100 can also be one of the many hardware modules of the terminal device.

[0060] Alternatively, in another example, the furniture harmful substance detection and analysis system 100 and the terminal device may also be separate devices, and the furniture harmful substance detection and analysis system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0061] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0062] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0063] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0064] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0065] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0066] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The second and other words are used to indicate names, but not to indicate any particular order.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit of the technical solution of the present invention.

Claims

1. A furniture harmful substance detection and analysis system, characterized in that: include: The preliminary measurement module is used to obtain the preliminary measurement value of furniture VOC; An environmental parameter acquisition module, used to receive a temperature time series data set and a humidity time series data set collected by a temperature sensor and a humidity sensor during a measurement process; A data extraction module, used to extract a preliminary measurement value of a first VOC component from the furniture VOC preliminary measurement value, and extract a property text description of the first VOC component from a backend database by querying; A compensation and correction module is used to perform response analysis on the temperature time series data set, the humidity time series data set and the text description of the properties of the first VOC component to compensate and correct the preliminary measurement value of the first VOC component, wherein the compensation and correction module includes: a collaborative analysis unit, used to perform a synergistic effect time series analysis on the temperature time series data set and the humidity time series data set to obtain a temperature and humidity collaborative time series feature; a correction unit, used to obtain a corrected measurement value of the first VOC component based on a soft constraint property-environmental response association representation between the temperature and humidity collaborative time series feature and the text description of the properties of the first VOC component.

2. The furniture harmful substance detection and analysis system according to claim 1, characterized in that: The collaborative analysis unit comprises: A temperature value synergy effect value synchronization calculation subunit is used to synchronize the temperature time series data set and the humidity time series data set and calculate the synergy effect value between the temperature value and the humidity value at the same time point to obtain a time queue of the temperature value synergy effect value; The collaborative time series feature extraction subunit is used to extract the temperature and humidity collaborative time series feature from the time queue of the temperature value collaborative effect value to obtain the temperature and humidity collaborative time series feature encoding vector as the temperature and humidity collaborative time series feature.

3. The furniture harmful substance detection and analysis system according to claim 2, characterized in that: The collaborative time series feature extraction subunit is used to: use a time series feature extractor based on an LSTM model to extract the temperature and humidity collaborative time series features from the time queue of the temperature value collaborative effect value to obtain the temperature and humidity collaborative time series feature encoding vector.

4. The furniture harmful substance detection and analysis system according to claim 1, characterized in that: The correction unit comprises: A first VOC component property semantic encoding subunit, configured to semantically encode the property text description of the first VOC component to obtain a first VOC component property semantic encoding vector; A property-environment response association subunit is used to perform property-environment hidden state independence soft constraint response analysis on the first VOC component property semantic coding vector and the temperature and humidity coordinated temporal feature coding vector to obtain a property-environment deep state response association coding vector as the soft constraint property-environment response association representation; The first VOC component correction subunit is used to obtain a corrected measurement value of the first VOC component based on the property-environment depth state response associated coding vector.

5. The furniture harmful substance detection and analysis system according to claim 4, characterized in that: The property-environment response association subunit includes: The inter-modal independence soft constraint secondary subunit is used to perform principal component analysis and linear transformation on the temperature and humidity collaborative time series feature coding vector to obtain a set of temperature and humidity collaborative time series principal component feature linear transformation coding vectors, and then perform inter-modal independence soft constraint calculation on the set of temperature and humidity collaborative time series principal component feature linear transformation coding vectors and the first VOC component property semantic coding vector to obtain a set of property-environment depth state inter-modal independence soft constraint factors; The property-environment deep state response secondary subunit is used to perform adaptive response analysis on the set of the first VOC component property semantic coding vector and the temperature and humidity collaborative time series principal component feature linear transformation coding vector based on the set of soft constraint factors of independence between the property-environment deep state modes to obtain the property-environment deep state response associated coding vector.

6. The furniture harmful substance detection and analysis system according to claim 5, characterized in that: The inter-modal independence soft constraint secondary subunit includes: The temperature and humidity coordinated time series principal component analysis tertiary subunit is used to perform principal component analysis on the temperature and humidity coordinated time series feature coding vector to obtain a set of temperature and humidity coordinated time series principal component feature coding vectors; The third-level sub-unit of the linear transformation of the temperature and humidity coordinated time series principal component feature is used to linearly transform each temperature and humidity coordinated time series principal component feature coding vector in the set of the temperature and humidity coordinated time series principal component feature coding vectors to obtain the set of the temperature and humidity coordinated time series principal component feature linear transformation coding vectors, wherein each temperature and humidity coordinated time series principal component feature linear transformation coding vector in the set of the temperature and humidity coordinated time series principal component feature linear transformation coding vectors has the same characteristic scale as the first VOC component property semantic coding vector; The third-level sub-unit of property-environment deep state inter-modal independence encoding is used to perform inter-modal independence encoding on each temperature and humidity coordinated temporal principal component feature linear transformation encoding vector in the set of the first VOC component property semantic encoding vector and the temperature and humidity coordinated temporal principal component feature linear transformation encoding vector to obtain a set of property-environment deep state inter-modal independence encoding matrices; The third-level subunit for calculating the independence soft constraint factor is used to calculate a set of property-environment depth state inter-modal independence soft constraint factors based on the set of property-environment depth state inter-modal independence coding matrices.

7. The furniture harmful substance detection and analysis system according to claim 6, characterized in that: The independence soft constraint factor calculation three-level sub-unit is used for: Performing a standard field-based eigenoptimization on each property-environment depth state inter-modal independence encoding matrix in the set of property-environment depth state inter-modal independence encoding matrices to obtain a set of optimized property-environment depth state inter-modal independence encoding matrices; The square of the Frobenius norm of each optimized property-environment depth state inter-modal independence coding matrix in the set of the optimized property-environment depth state inter-modal independence coding matrices is used as the set of soft constraint factors for the property-environment depth state inter-modal independence.

8. The furniture harmful substance detection and analysis system according to claim 7, characterized in that: The property-environment depth state response secondary subunit is used to: Performing feature interaction response on each temperature and humidity coordinated time series principal component feature linear transformation coding vector in the set of the first VOC component property semantic coding vector and the temperature and humidity coordinated time series principal component feature linear transformation coding vector to obtain a set of property-environment deep state principal component fine-grained modal response interaction coding vectors; Based on the set of soft constraint factors of independence between the property-environment depth state modes, an adaptive response analysis is performed on the set of property-environment depth state principal component fine-grained modal response interaction coding vectors to obtain the property-environment depth state response association coding vector.

9. The furniture harmful substance detection and analysis system according to claim 8, characterized in that: The first VOC component correction subunit is used to: Inputting the property-environment depth state response association encoding vector into a decoder-based compensation estimation module to obtain a first compensation coefficient; The preliminary measurement value of the first VOC component is corrected based on the first compensation coefficient to obtain a corrected measurement value of the first VOC component.

Images