Quantitative detection method and system based on Raman spectrum

Abstract

The invention discloses a raman spectrum-based quantitative detection method and system, and the method comprises the steps: obtaining a raman spectrogram of a to-be-detected sample, employing a data analysis and processing technology based on artificial intelligence to extract each characteristic peak and an intensity value from the raman spectrogram, carrying out the One-hot coding of each characteristic peak, and carrying out the one-hot coding of each characteristic peak, thereby obtaining the raman spectrum of the to-be-detected sample, and obtaining the raman spectrum of the to-be-detected sample. The method comprises the following steps: firstly, carrying out one-hot coding on each characteristic peak, adding an intensity value at the tail of one-hot coding of each characteristic peak to form a characteristic peak low-dimensional embedded coding feature, then, carrying out significance enhancement processing on each characteristic peak low-dimensional embedded coding feature, and carrying out linear aggregation according to each enhanced characteristic peak low-dimensional embedded coding feature to form a characteristic peak low-dimensional embedded coding feature; and the obtained characteristic peak information is jointly expressed to intelligently obtain the concentration value of the target substance. Compared with a traditional method, the method has the advantages that the efficiency and the precision of Raman spectrum quantitative analysis are remarkably improved, and particularly, the method has higher adaptive capacity when being used for processing complex or overlapped spectrum peaks, so that more reliable and more intelligent Raman spectrum quantitative analysis is realized.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to a quantitative detection method and system based on Raman spectroscopy. Background Art

[0002] Raman spectroscopy is a powerful analytical tool widely used in chemistry, materials science, biomedicine and other fields. It provides information about molecular structure and chemical environment by measuring the scattering of monochromatic light by molecules. This non-destructive detection method is not only applicable to liquid, solid and gas samples, but also can directly analyze complex systems without sample preparation, making it an important means to study material properties and diagnose diseases.

[0003] Traditional quantitative detection of Raman spectroscopy relies on operators to manually test each sample. However, manual identification of characteristic peaks and establishment of calibration curves are highly dependent on the experience and skills of the operator. Different operators may have different judgment criteria, resulting in highly subjective results and lack of uniformity and objectivity. In addition, manual testing of each sample takes a lot of time and effort, especially when dealing with large batches of samples. The work efficiency is low and it is difficult to meet the needs of rapid analysis.

[0004] Therefore, an optimized Raman spectroscopy-based quantitative detection scheme is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a quantitative detection method and system based on Raman spectroscopy, which obtains the Raman spectrum of the sample to be tested, and uses data analysis and processing technology based on artificial intelligence to extract each characteristic peak and intensity value from the Raman spectrum, and each characteristic peak is uniquely encoded, and the intensity value is added at the end of each characteristic peak uniquely encoded to form a characteristic peak low-dimensional embedded coding feature, and then, each characteristic peak low-dimensional embedded coding feature is significantly enhanced, and the characteristic peak information obtained by linearly aggregating the low-dimensional embedded coding features of each enhanced characteristic peak is represented jointly to intelligently obtain the concentration value of the target substance. In this way, by extracting characteristic peaks and intensity values ​​by automated means, the analysis time of the operator can be greatly shortened, and the detection efficiency is improved. At the same time, the characteristic peaks and the measured intensity values ​​can be more accurately identified, especially when processing complex samples, avoiding errors caused by human subjectivity, thereby achieving more reliable and intelligent Raman spectroscopy quantitative analysis.

[0006] According to one aspect of the present application, a quantitative detection method based on Raman spectroscopy is provided, comprising:

[0007] Obtaining a Raman spectrum of the sample to be tested;

[0008] Extracting the intensity value of each characteristic peak from the Raman spectrum;

[0009] Performing one-hot encoding on each of the characteristic peaks to obtain a set of characteristic peak one-hot encoding vectors;

[0010] Adding the intensity value of each characteristic peak at the end of each characteristic peak one-hot encoding vector in the set of characteristic peak one-hot encoding vectors to obtain a set of characteristic peak low-dimensional embedded encoding vectors;

[0011] The set of the feature peak low-dimensional embedded coding vectors is subjected to feature peak low-dimensional sequence feature enhancement to obtain a set of feature peak low-dimensional embedded coding vectors, including: mapping each feature peak low-dimensional embedded coding vector in the set of the feature peak low-dimensional embedded coding vectors to a hyperbolic space to obtain a set of feature peak low-dimensional embedded coding mapping feature vectors; calculating a feature significance description factor of each feature peak low-dimensional embedded coding mapping feature vector in the set of the feature peak low-dimensional embedded coding mapping feature vectors to obtain a set of feature peak low-dimensional embedded coding significance description factors; based on the set of feature peak low-dimensional embedded coding significance description factors, the set of the feature peak low-dimensional embedded coding vectors is subjected to feature significance enhancement to obtain a set of feature peak low-dimensional embedded coding enhancement coding vectors;

[0012] Aggregating the set of the feature peak low-dimensional embedded enhanced coding vectors to obtain a joint representation of feature peak information;

[0013] Based on the joint representation of the characteristic peak information, a detection result is obtained, and the detection result is a decoded value of the concentration of the target substance.

[0014] According to another aspect of the present application, a quantitative detection system based on Raman spectroscopy is provided, comprising:

[0015] A Raman spectrum acquisition module for the sample to be tested, used to obtain the Raman spectrum of the sample to be tested;

[0016] A characteristic peak intensity value extraction module, used to extract the intensity value of each characteristic peak from the Raman spectrum;

[0017] A characteristic peak one-hot encoding module, used for performing one-hot encoding on each characteristic peak to obtain a set of characteristic peak one-hot encoding vectors;

[0018] A characteristic peak low-dimensional embedding coding module, used for adding the intensity value of each characteristic peak at the end of each characteristic peak one-hot encoding vector in the set of characteristic peak one-hot encoding vectors to obtain a set of characteristic peak low-dimensional embedding coding vectors;

[0019] A feature peak sequence enhancement module is used to perform feature peak low-dimensional sequence feature enhancement on the set of feature peak low-dimensional embedded coding vectors to obtain a set of feature peak low-dimensional embedded enhanced coding vectors, wherein the feature peak sequence enhancement module includes: a feature peak low-dimensional mapping unit, used to map each feature peak low-dimensional embedded coding vector in the set of feature peak low-dimensional embedded coding vectors to a hyperbolic space to obtain a set of feature peak low-dimensional embedded coding mapping feature vectors; a feature peak significance description factor calculation unit, used to calculate the feature significance description factor of each feature peak low-dimensional embedded coding mapping feature vector in the set of feature peak low-dimensional embedded coding mapping feature vectors to obtain a set of feature peak low-dimensional embedded coding significance description factors; a feature peak low-dimensional embedding enhancement unit, used to perform feature significance enhancement on the set of feature peak low-dimensional embedded coding vectors based on the set of feature peak low-dimensional embedded coding significance description factors to obtain the set of feature peak low-dimensional embedded enhanced coding vectors;

[0020] A feature peak information joint module, used for aggregating the set of feature peak low-dimensional embedded enhanced coding vectors to obtain a feature peak information joint representation;

[0021] The radar signal transmission analysis module is used to obtain a detection result based on the joint representation of the characteristic peak information, and the detection result is a decoded value of the concentration of the target substance.

[0022] Compared with the prior art, the present application provides a quantitative detection method and system based on Raman spectroscopy, which obtains the Raman spectrum of the sample to be tested, and uses data analysis and processing technology based on artificial intelligence to extract each characteristic peak and intensity value from the Raman spectrum, and each characteristic peak is uniquely encoded, and the intensity value is added at the end of each characteristic peak uniquely encoded to form a characteristic peak low-dimensional embedded coding feature, and then, each characteristic peak low-dimensional embedded coding feature is significantly enhanced, and the characteristic peak information joint representation obtained by linear aggregation between the low-dimensional embedded coding features of each enhanced characteristic peak is intelligently obtained. The concentration value of the target substance. In this way, the characteristic peaks and intensity values ​​are extracted by automated means, which can greatly shorten the analysis time of the operator and improve the detection efficiency. At the same time, the characteristic peaks and measured intensity values ​​can be more accurately identified, especially when processing complex samples, avoiding errors caused by human subjectivity, thereby achieving more reliable and intelligent Raman spectroscopy quantitative analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] 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.

[0024] Figure 1 The flowchart is a quantitative detection method based on Raman spectroscopy according to an embodiment of the present application.

[0025] Figure 2 Schematic diagram of data flow of a quantitative detection method based on Raman spectroscopy according to an embodiment of the present application.

[0026] Figure 3 The present invention is a flowchart of performing characteristic peak low-dimensional sequence feature enhancement on the set of characteristic peak low-dimensional embedded coding vectors in the quantitative detection method based on Raman spectroscopy according to an embodiment of the present application to obtain a set of characteristic peak low-dimensional embedded enhanced coding vectors.

[0027] Figure 4 This is a flowchart of performing feature significant enhancement on the set of characteristic peak low-dimensional embedded coding vectors based on the set of characteristic peak low-dimensional embedded coding significance descriptor factors in the quantitative detection method based on Raman spectroscopy according to an embodiment of the present application to obtain the set of characteristic peak low-dimensional embedded enhanced coding vectors.

[0028] Figure 5 4 is a block diagram of a quantitative detection system based on Raman spectroscopy according to an embodiment of the present application. DETAILED DESCRIPTION

[0029] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, but rather these embodiments are provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0030] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0031] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0032] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0033] Raman spectroscopy is a powerful analytical tool widely used in chemistry, materials science, biomedicine and other fields. It provides information about molecular structure and chemical environment by measuring the scattering of monochromatic light by molecules. This non-destructive detection method is not only applicable to liquid, solid and gas samples, but also can directly analyze complex systems without sample preparation, making it an important means to study material properties and diagnose diseases.

[0034] It should be understood that Raman spectroscopy is a non-destructive analysis technique used to study molecular structure. It analyzes by observing the energy changes in the process of material scattering monochromatic light (usually laser). Specifically, during the Raman scattering process, the incident light (called "excitation light") interacts with the molecules in the substance, causing the energy of some photons to change. This phenomenon is called inelastic scattering. The intensity of Raman scattering is usually proportional to the concentration of the target substance in the sample, which means that under certain conditions, the intensity of a specific peak in the Raman spectrum can reflect the concentration level of the substance. That is, by analyzing the intensity of the characteristic peaks in the Raman spectrum, the concentration of the target substance can be quantitatively determined.

[0035] Therefore, in view of the above technical problems, the technical concept of the present application is to obtain the Raman spectrum of the sample to be tested, and use data analysis and processing technology based on artificial intelligence to extract each characteristic peak and intensity value from the Raman spectrum, and by performing unique heat encoding on each characteristic peak, and adding an intensity value at the end of each characteristic peak unique heat encoding to form a characteristic peak low-dimensional embedded coding feature, and then, the low-dimensional embedded coding feature of each characteristic peak is significantly enhanced, and the characteristic peak information obtained by linearly aggregating the low-dimensional embedded coding features of each enhanced characteristic peak is jointly represented to intelligently obtain the concentration value of the target substance. In this way, by extracting characteristic peaks and intensity values ​​by automated means, the analysis time of the operator can be greatly shortened and the detection efficiency is improved. At the same time, characteristic peaks and measured intensity values ​​can be more accurately identified, especially when processing complex samples, errors caused by human subjectivity are avoided, thereby achieving more reliable and intelligent Raman spectroscopy quantitative analysis.

[0036] Figure 1 The flowchart is a quantitative detection method based on Raman spectroscopy according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of data flow of a quantitative detection method based on Raman spectroscopy according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the quantitative detection method based on Raman spectroscopy of the embodiment of the present application, the method includes: S110, obtaining a Raman spectrum of the sample to be tested; S120, extracting the intensity value of each characteristic peak from the Raman spectrum; S130, performing one-hot encoding on the each characteristic peak to obtain a set of characteristic peak one-hot encoding vectors; S140, adding the intensity value of each characteristic peak at the end of each characteristic peak one-hot encoding vector in the set of characteristic peak one-hot encoding vectors to obtain a set of characteristic peak low-dimensional embedded encoding vectors; S150, performing characteristic peak low-dimensional sequence feature enhancement on the set of characteristic peak low-dimensional embedded encoding vectors to obtain a set of characteristic peak low-dimensional embedded enhanced encoding vectors; S160, aggregating the set of characteristic peak low-dimensional embedded enhanced encoding vectors to obtain a joint representation of characteristic peak information; S170, obtaining a detection result based on the joint representation of characteristic peak information, and the detection result is a decoded value of the concentration of the target substance.

[0037] In step S110, a Raman spectrum of the sample to be tested is obtained. It should be understood that, considering that the Raman spectrum provides important information about the molecular structure in the sample, specifically, the position of the characteristic peak on the Raman spectrum corresponds to the vibration frequency of the chemical bond in the molecule, which depends on the mass of the atom and the force constant of the chemical bond, and the intensity of the peak can provide information about the concentration of the molecule in the sample, and substances with different concentrations will cause the peak intensity to change. Based on this, in order to more accurately analyze the concentration of the target substance, in the technical solution of the present application, a Raman spectrum of the sample to be tested is obtained.

[0038] In step S120, the intensity value of each characteristic peak is extracted from the Raman spectrum. Accordingly, considering that the position of the characteristic peak can be used to qualitatively identify the specific chemical bonds or functional groups present in the sample, and there is usually a certain relationship between the intensity of the characteristic peak and the concentration of the corresponding substance in the sample, in order to understand and analyze each characteristic peak and intensity information more carefully and accurately, so as to improve the accuracy of the target substance concentration detection, in the technical solution of the present application, the intensity value of each characteristic peak is extracted from the Raman spectrum.

[0039] In particular, in a specific embodiment of the present application that can be realized, the intensity values ​​of each characteristic peak extracted from the Raman spectrum can be extracted in the following manner. Specifically, 1. Threshold detection: First, a reasonable threshold is set according to the characteristics and experience of the Raman spectrum. The selection of the threshold needs to take into account the noise level and the actual intensity of the characteristic peak. Then, traverse each point in the spectrum to find all points above the threshold. After that, from the points above the threshold, further confirm which are the real peaks, usually need to consider the local maximum principle, that is, a point must be the highest point in its neighborhood to be considered as a peak. 2. Gaussian fitting: First, Gaussian fitting is performed on each peak. In particular, the goal of fitting is to find the best fitting curve of a Gaussian function, which should be able to describe the shape of the suspected peak well. Then, the center position (wave number or wavelength) and intensity (height) of each peak are determined by Gaussian fitting. 3. Peak positioning: Read the center position of the peak from the result of Gaussian fitting (usually the mean of the Gaussian function). 4. Peak intensity measurement: Determine by measuring the maximum intensity value at the peak. Alternatively, the maximum value of the Gaussian function may be used as the intensity value, or the amplitude of the Gaussian function may be used to obtain the intensity value of each characteristic peak.

[0040] In step S130, each characteristic peak is one-hot encoded to obtain a set of characteristic peak one-hot encoded vectors. It should be understood that in order to convert the position information of each characteristic peak into a format that can be understood and processed by the machine learning algorithm, so as to better understand the information of each characteristic peak, in the technical solution of the present application, each characteristic peak is one-hot encoded to obtain a set of characteristic peak one-hot encoded vectors. That is, the one-hot encoding converts the position of each characteristic peak into a binary vector, so that different characteristic peaks can be clearly distinguished, making the position information clearer.

[0041] In step S140, the intensity values ​​of each characteristic peak are added to the end of each characteristic peak unique heat encoding vector in the set of characteristic peak unique heat encoding vectors to obtain a set of characteristic peak low-dimensional embedded coding vectors. Accordingly, considering that relying solely on the position information of the characteristic peak is sometimes not enough to distinguish different compounds, especially when the sample has a similar characteristic peak position, therefore, in the technical solution of the present application, the intensity values ​​of each characteristic peak are added to the end of each characteristic peak unique heat encoding vector in the set of characteristic peak unique heat encoding vectors to obtain a set of characteristic peak low-dimensional embedded coding vectors. It should be understood that since the characteristic peaks at the same position may represent different substances or concentrations due to different intensities, the ability to distinguish different compounds in a substance can be improved by adding intensity values, so that the obtained characteristic peak low-dimensional embedded coding vector contains both position information and information about intensity, so that the model can more comprehensively understand the characteristics of each characteristic peak, and provide rich input for subsequent detection of target substance concentration.

[0042] In step S150, feature peak low-dimensional sequence feature enhancement is performed on the set of feature peak low-dimensional embedded coding vectors to obtain a set of feature peak low-dimensional embedded enhanced coding vectors. Figure 3 The present invention is a flowchart of performing feature peak low-dimensional sequence feature enhancement on the set of feature peak low-dimensional embedded coding vectors in the quantitative detection method based on Raman spectroscopy according to an embodiment of the present application to obtain a set of feature peak low-dimensional embedded enhanced coding vectors. Specifically, in the embodiment of the present application, Figure 3 As shown, the set of feature peak low-dimensional embedded coding vectors is subjected to feature peak low-dimensional sequence feature enhancement to obtain a set of feature peak low-dimensional embedded enhanced coding vectors, including: S210, mapping each feature peak low-dimensional embedded coding vector in the set of feature peak low-dimensional embedded coding vectors to a hyperbolic space to obtain a set of feature peak low-dimensional embedded coding mapping feature vectors; S220, calculating the feature significance description factor of each feature peak low-dimensional embedded coding mapping feature vector in the set of feature peak low-dimensional embedded coding mapping feature vectors to obtain a set of feature peak low-dimensional embedded coding significance description factors; S230, based on the set of feature peak low-dimensional embedded coding significance description factors, performing feature significance enhancement on the set of feature peak low-dimensional embedded coding vectors to obtain the set of feature peak low-dimensional embedded enhanced coding vectors.

[0043] It should be understood that in order to improve the model's ability to distinguish different compounds and enhance the information content of the characteristic peak low-dimensional embedded coding vector, which makes the feature representation richer and more comprehensive and can better capture the subtle differences in the data, in the technical solution of the present application, the set of the characteristic peak low-dimensional embedded coding vector is subjected to characteristic peak low-dimensional sequence feature enhancement to obtain a set of characteristic peak low-dimensional embedded enhanced coding vectors. In this way, even when the characteristic peak positions are similar, the enhanced feature vector can better reflect the intensity and other detailed information.

[0044] Specifically, first, each characteristic peak low-dimensional embedding coding vector in the set of characteristic peak low-dimensional embedding coding vectors is mapped to a hyperbolic space to obtain a set of characteristic peak low-dimensional embedding coding mapping feature vectors. It should be understood that the hyperbolic space has negative curvature, and this geometric structure is very suitable for embedding data with a hierarchical structure or a tree structure. The characteristic peaks in the Raman spectrum often have a hierarchical distribution characteristic, and mapping to the hyperbolic space can better capture this hierarchical structure.

[0045] Next, the feature significance description factor of each feature peak low-dimensional embedded coding mapping feature vector in the set of the feature peak low-dimensional embedded coding mapping feature vector is calculated to obtain a set of feature peak low-dimensional embedded coding significance description factors. In other words, the significance description factor can be used to adjust the weight of the feature so that the important feature peak features are more prominent, thereby improving the distinguishing ability of the model. In particular, in the technical solution of the present application, the squared Poincare norm of each feature peak low-dimensional embedded coding mapping feature vector is used to perform significance description modeling on the feature significance of each feature peak low-dimensional embedded coding mapping feature vector, that is, the Poincare norm is an effective metric in hyperbolic geometry, which can measure the "length" of the feature vector in the hyperbolic space, thereby reflecting its significance, to better understand the contribution of each feature peak.

[0046] Then, based on the set of the feature peak low-dimensional embedded coding significance descriptor factors, the set of the feature peak low-dimensional embedded coding vectors is significantly enhanced to obtain the set of the feature peak low-dimensional embedded enhanced coding vectors. That is, by significantly enhancing the original features, those features that are more important in the classification or regression task can be highlighted, so as to improve the accuracy of subsequent target concentration detection. Specifically, first, each significance descriptor is hierarchically modulated to selectively enhance certain features and weaken other features, which helps to more clearly mark the key feature peak information in the feature space while ignoring those noise or insignificant features. In particular, in the present technical solution, the hierarchical mask modulation strategy adopted sets two thresholds, which are used to divide the feature significance descriptor sequence into three different levels. And different degrees of feature scaling are implemented for the feature significance descriptor in each level, so that different degrees of attention can be applied to them on the basis of retaining the diversified feature components, thereby more effectively emphasizing the key patterns and reducing the influence of non-relevant features. Finally, the set of low-dimensional embedding coding vectors of feature peaks is weighted by using the modulated weights to highlight and enhance those feature peaks with higher significance, making it easier for the model to identify and distinguish different feature peaks, and obtaining a set of low-dimensional embedding enhanced coding vectors of feature peaks.

[0047] Specifically, in step S210, each feature peak low-dimensional embedded coding vector in the set of feature peak low-dimensional embedded coding vectors is mapped to a hyperbolic space to obtain a set of feature peak low-dimensional embedded coding mapping feature vectors, including: after matrix-vector multiplication of a first weight matrix and the feature peak low-dimensional embedded coding vector, vector-matrix multiplication of the obtained first feature peak low-dimensional embedded coding weight vector and a second weight matrix to obtain the feature peak low-dimensional embedded coding mapping feature vector.

[0048] Specifically, in step S220, the feature significance description factor of each feature peak low-dimensional embedded coding mapping feature vector in the set of the feature peak low-dimensional embedded coding mapping feature vector is calculated to obtain a set of feature peak low-dimensional embedded coding significance description factors, including: calculating the square of the one-norm of the feature peak low-dimensional embedded coding mapping feature vector to obtain the feature peak low-dimensional embedded coding mapping feature norm value; adding and subtracting the constant one from the feature peak low-dimensional embedded coding mapping feature norm value to obtain a first feature peak low-dimensional embedded coding mapping feature norm modulation value and a second feature peak low-dimensional embedded coding mapping feature norm modulation value, and dividing the first feature peak low-dimensional embedded coding mapping feature norm modulation value by the second feature peak low-dimensional embedded coding mapping feature norm modulation value to obtain the feature peak low-dimensional embedded coding mapping feature norm modulation value; calculating the logarithmic function value of the logarithm of the feature peak low-dimensional embedded coding mapping feature norm modulation value with base two to obtain the feature peak low-dimensional embedded coding significance description factor.

[0049] Figure 4 The present invention is a flowchart of performing feature enhancement on the set of feature peak low-dimensional embedded coding vectors based on the set of feature peak low-dimensional embedded coding significance descriptors in the quantitative detection method based on Raman spectroscopy according to the embodiment of the present application to obtain the set of feature peak low-dimensional embedded enhanced coding vectors. Specifically, in the embodiment of the present application, Figure 4 As shown, based on the set of feature peak low-dimensional embedded coding significance descriptor factors, the set of feature peak low-dimensional embedded coding vectors is subjected to feature significance enhancement to obtain the set of feature peak low-dimensional embedded enhanced coding vectors, including: S310, performing hierarchical mask modulation on the set of feature peak low-dimensional embedded coding significance descriptor factors to obtain a set of feature peak low-dimensional embedded coding significance modulation weights; S320, using each feature peak low-dimensional embedded coding significance modulation weight in the set of feature peak low-dimensional embedded coding significance modulation weights as a weight, and performing weighted enhancement on the set of feature peak low-dimensional embedded coding vectors to obtain the set of feature peak low-dimensional embedded enhanced coding vectors.

[0050] Specifically, in step S310, the set of feature peak low-dimensional embedded coding significance descriptor factors is subjected to hierarchical mask modulation to obtain a set of feature peak low-dimensional embedded coding significance modulation weights, including: normalizing the set of feature peak low-dimensional embedded coding significance descriptor factors to obtain a set of normalized feature peak low-dimensional embedded coding significance descriptor factors; and hierarchically modulating the set of normalized feature peak low-dimensional embedded coding significance descriptor factors to obtain a set of feature peak low-dimensional embedded coding significance modulation weights.

[0051] More specifically, in an embodiment of the present application, the set of feature peak low-dimensional embedded coding significance descriptor factors is normalized to obtain a set of normalized feature peak low-dimensional embedded coding significance descriptor factors, including: calculating an exponential function value with a natural constant e as the base and each feature peak low-dimensional embedded coding significance descriptor factor in the set of feature peak low-dimensional embedded coding significance descriptor factors as the exponent to obtain a set of feature peak low-dimensional embedded coding index significance descriptor factors; calculating the positional sum of the set of feature peak low-dimensional embedded coding index significance descriptor factors to obtain a feature peak low-dimensional embedded coding index total sum significance descriptor factor; calculating the positional division between the set of feature peak low-dimensional embedded coding index significance descriptor factors and the feature peak low-dimensional embedded coding index total sum significance descriptor factor to obtain the set of feature peak low-dimensional embedded coding index significance descriptor factors.

[0052] More specifically, in an embodiment of the present application, the set of normalized feature peak low-dimensional embedded coding significance descriptor factors is hierarchically modulated to obtain a set of feature peak low-dimensional embedded coding significance modulation weights, including: comparing each normalized feature peak low-dimensional embedded coding significance descriptor factor in the set of normalized feature peak low-dimensional embedded coding significance descriptor factors with a preset threshold to obtain a set of feature peak low-dimensional embedded coding significance modulation weights; wherein, in response to the normalized feature peak low-dimensional embedded coding significance descriptor factor being greater than a second preset threshold, the normalized feature peak low-dimensional embedded coding significance descriptor factor is hierarchically modulated to obtain a set of feature peak low-dimensional embedded coding significance modulation weights; The product of the factor and the first feature scaling factor is used as the significant modulation weight of the low-dimensional embedded coding of the feature peak; in response to the normalized feature peak low-dimensional embedded coding significance description factor being greater than the first preset threshold and less than or equal to the second preset threshold, the normalized feature peak low-dimensional embedded coding significance description factor is used as the significant modulation weight of the low-dimensional embedded coding of the feature peak; in response to the normalized feature peak low-dimensional embedded coding significance description factor being less than or equal to the first preset threshold, the product of the normalized feature peak low-dimensional embedded coding significance description factor and the second feature scaling factor is used as the significant modulation weight of the low-dimensional embedded coding of the feature peak.

[0053] In the embodiment of the present application, specifically, the set of the characteristic peak low-dimensional embedded coding vectors is subjected to characteristic peak low-dimensional sequence feature enhancement, which can be expressed by the formula:

[0054] X={x1,x2,...,x i ,...,x n}

[0055] h i =W1x i W2

[0056]

[0057] w i =mask(o i ′)

[0058]

[0059] Y={y1,y2,...,y i ,...,y n}

[0060] y i =w i ·x i

[0061] Where X is the set of low-dimensional embedding coding vectors of the feature peaks, x1, x2, ..., x i ,...,x n are the 1st, 2nd, ..., ith, ..., nth feature peak low-dimensional embedding coding vectors in the set of the feature peak low-dimensional embedding coding vectors, W1 and W2 are the first weight matrix and the second weight matrix, respectively, h i is the i-th feature peak low-dimensional embedding coding mapping feature vector in the set of feature peak low-dimensional embedding coding mapping feature vectors, ‖·‖ 2 is the square of the one-norm of the vector, log represents the logarithmic function value with base 2, o i is the i-th feature peak low-dimensional embedding coding significant description factor in the set of feature peak low-dimensional embedding coding significant description factors, exp(·) represents an exponential function with the natural constant e as the base, o i ′ is the ith normalized feature peak low-dimensional embedded coding significant description factor in the set of normalized feature peak low-dimensional embedded coding significant description factors, mask(·) is the masking process, θ is the preset threshold, λ and η are feature scaling factors, and λ>1, η<1, w i is the i-th feature peak low-dimensional embedding coding significant modulation weight in the set of feature peak low-dimensional embedding coding significant modulation weights, Y is the set of feature peak low-dimensional embedding enhanced coding vectors, y1, y2, ..., y i ,...,y n It is the 1st, 2nd, ..., ith, ..., nth feature peak low-dimensional embedding enhanced coding vector in the set of the feature peak low-dimensional embedding enhanced coding vector.

[0062] In step S160, the set of the characteristic peak low-dimensional embedded enhanced coding vectors is aggregated to obtain a joint representation of characteristic peak information. Specifically, in an embodiment of the present application, the set of the characteristic peak low-dimensional embedded enhanced coding vectors is aggregated to obtain a joint representation of characteristic peak information, including: inputting the set of the characteristic peak low-dimensional embedded enhanced coding vectors into a linear aggregation network to obtain a joint representation vector of characteristic peak information as the joint representation of characteristic peak information. Accordingly, in order to more comprehensively and comprehensively display the representation of the characteristic peaks in the entire Raman spectrum, so that the model can more comprehensively understand the characteristic peak pattern in the entire image, in the technical solution of the present application, the set of the characteristic peak low-dimensional embedded enhanced coding vectors is input into a linear aggregation network to obtain a joint representation vector of characteristic peak information. In particular, in a specific embodiment of the present application, the vector obtained by cascading the set of the characteristic peak low-dimensional embedded enhanced coding vectors can be used as the joint representation vector of the characteristic peak information.

[0063] In step S160, based on the joint representation of the characteristic peak information, a detection result is obtained, and the detection result is a decoded value of the concentration of the target substance. Specifically, in an embodiment of the present application, based on the joint representation of the characteristic peak information, a detection result is obtained, including: inputting the joint representation vector of the characteristic peak information into a target substance concentration detection module based on a decoder to obtain the detection result. That is, the joint representation of the characteristic peak information obtained by first gathering the set of the characteristic peak low-dimensional embedded enhanced coding vectors is classified and processed to intelligently obtain the concentration value of the target substance. In this way, by extracting characteristic peaks and intensity values ​​through automated means, the analysis time of the operator can be greatly shortened and the detection efficiency can be improved. At the same time, characteristic peaks and measured intensity values ​​can be more accurately identified, especially when processing complex samples, avoiding errors caused by human subjectivity, thereby achieving more reliable and intelligent Raman spectroscopy quantitative analysis.

[0064] It should be understood that in the technical solution of the present application, it is considered that each characteristic peak low-dimensional embedded coding vector in the set of characteristic peak low-dimensional embedded coding vectors respectively represents the attribute description of each characteristic peak in the Raman spectrum of the sample to be tested and the mixed embedded coding representation of the intensity value. After the characteristic peak low-dimensional sequence feature enhancement is performed, the feature expression significance of each characteristic peak low-dimensional embedded coding vector in the set of characteristic peak low-dimensional embedded coding vectors will be enhanced, but at the same time, the nonlinear correlation expression between each characteristic peak low-dimensional embedded coding vector in the set of characteristic peak low-dimensional embedded coding vectors will be suppressed. This makes the fused feature representation of the characteristic peak information joint representation vector obtained by inputting the set of characteristic peak low-dimensional embedded enhanced coding vectors into the linear aggregation network have a complex spatial structure, thereby expecting to improve its decoding regression convergence and generalization effect under complex spatial structure.

[0065] Based on this, preferably, in an example of the present application, the characteristic peak information is combined with the representation vector through a decoder-based target substance concentration detection module to obtain a detection result, including:

[0066] Calculating the sum of the absolute values ​​of each eigenvalue of the characteristic peak information joint representation vector to obtain a first characteristic peak information joint representation space structure value, and calculating the square root of the sum of the squares of each eigenvalue of the characteristic peak information joint representation vector to obtain a second characteristic peak information joint representation space structure value;

[0067] Multiplying each eigenvalue of the characteristic peak information joint representation vector by the first characteristic peak information joint representation space structure value and the second characteristic peak information joint representation space structure value respectively to obtain a first characteristic peak information joint representation structure reference value and a second characteristic peak information joint representation structure reference value corresponding to each eigenvalue;

[0068] Multiplying each eigenvalue of the characteristic peak information joint representation vector by the length of the characteristic peak information joint representation vector and the square root of the length to obtain a first characteristic peak information joint representation scale transformation value and a second characteristic peak information joint representation scale transformation value corresponding to each eigenvalue;

[0069] Dividing the first characteristic peak information joint representation structure reference value by the difference between the first characteristic peak information joint representation space structure value and the first characteristic peak information joint representation scale transformation value to obtain a first characteristic peak information joint representation transformation adjustment value;

[0070] The second characteristic peak information joint representation structure reference value is divided by the difference between the second characteristic peak information joint representation space structure value and the second characteristic peak information joint representation scale transformation value to obtain a second characteristic peak information joint representation transformation adjustment value;

[0071] Calculating a weighted sum of the first characteristic peak information joint representation transformation adjustment value and the second characteristic peak information joint representation transformation adjustment value to obtain each eigenvalue of the optimized characteristic peak information joint representation vector; and

[0072] The optimized characteristic peak information is combined with a representation vector and passed through a target substance concentration detection module based on a decoder to obtain a detection result.

[0073] Here, the characteristic peak information is jointly represented by a vector, denoted as V, which is optimally represented as:

[0074] V′=V1⊕(ω⊙V2)

[0075]

[0076] v 1i=(α×v i ) / (α-L×v i )

[0077]

[0078] Wherein, V is the joint representation vector of the characteristic peak information, L is the length of the joint representation vector of the characteristic peak information, and v i is the eigenvalue of the ith position in the characteristic peak information joint representation vector, α is the spatial structure value of the first characteristic peak information joint representation, β is the spatial structure value of the second characteristic peak information joint representation, v 1i is the first characteristic peak information joint representation transformation adjustment value at the ith position in the first characteristic peak information joint representation transformation adjustment vector, v 2i is the second characteristic peak information jointly represented transformation adjustment value of the ith position in the second characteristic peak information jointly represented transformation adjustment vector, V1 is the first characteristic peak information jointly represented transformation adjustment vector, V2 is the second characteristic peak information jointly represented transformation adjustment vector, ω represents a weighted hyperparameter, ⊕ represents position point addition, ⊙ represents position point multiplication, and V′ is the optimized characteristic peak information jointly represented vector.

[0079] Therefore, in the preferred example, for the spatial structure information of the feature set of the characteristic peak information joint representation vector in the high-dimensional space, the scale-based frame transformation of each eigenvalue of the characteristic peak information joint representation vector is performed by using the class norm space structured representation of the characteristic peak information joint representation vector as a reference window, and the frame attention weight adjustment based on the spatial structure of each eigenvalue of the characteristic peak information joint representation vector is realized to ensure the spatial transformation (translation, scaling and rotation) invariance of the characteristic peak information joint representation vector under the interaction of the feature space, thereby improving the convergence and generalization effect of the decoding regression of the feature set of the characteristic peak information joint representation vector under the complex spatial structure representation, and improving the accuracy of the detection result obtained by the target substance concentration detection module based on the decoder. In this way, by extracting characteristic peaks and intensity values ​​by automated means, the analysis time of the operator can be greatly shortened and the detection efficiency is improved. At the same time, characteristic peaks and measured intensity values ​​can be more accurately identified, especially when processing complex samples, errors caused by human subjectivity are avoided, thereby achieving more reliable and intelligent Raman spectroscopy quantitative analysis.

[0080] In summary, the quantitative detection method based on Raman spectroscopy based on the embodiment of the present application is explained, which obtains the Raman spectrum of the sample to be tested, and uses data analysis and processing technology based on artificial intelligence to extract each characteristic peak and intensity value from the Raman spectrum, and each characteristic peak is uniquely encoded, and the intensity value is added at the end of each characteristic peak uniquely encoded to form a characteristic peak low-dimensional embedded coding feature, and then, each characteristic peak low-dimensional embedded coding feature is significantly enhanced, and the characteristic peak information obtained by linearly aggregating the low-dimensional embedded coding features of each enhanced characteristic peak is combined to intelligently obtain the concentration value of the target substance. In this way, by extracting characteristic peaks and intensity values ​​by automated means, the analysis time of the operator can be greatly shortened, and the detection efficiency is improved. At the same time, characteristic peaks and measured intensity values ​​can be more accurately identified, especially when processing complex samples, errors caused by human subjectivity are avoided, thereby achieving more reliable and intelligent Raman spectroscopy quantitative analysis.

[0081] Figure 5 FIG. 1 is a block diagram of a quantitative detection system based on Raman spectroscopy according to an embodiment of the present application. Figure 5As shown, according to the embodiment of the present application, the quantitative detection system 100 based on Raman spectroscopy includes: a Raman spectrum acquisition module 110 for the sample to be tested, which is used to obtain the Raman spectrum of the sample to be tested; a characteristic peak intensity value extraction module 120, which is used to extract the intensity value of each characteristic peak from the Raman spectrum; a characteristic peak one-hot encoding module 130, which is used to perform one-hot encoding on the each characteristic peak to obtain a set of characteristic peak one-hot encoding vectors; a characteristic peak low-dimensional embedded coding module 140, which is used to add the intensity value of each characteristic peak at the end of each characteristic peak one-hot encoding vector in the set of characteristic peak one-hot encoding vectors to obtain a set of characteristic peak low-dimensional embedded coding vectors; a characteristic peak sequence enhancement module 150, which is used to perform characteristic peak low-dimensional sequence feature enhancement on the set of characteristic peak low-dimensional embedded coding vectors to obtain a set of characteristic peak low-dimensional embedded enhanced coding vectors, wherein the characteristic peak sequence enhancement module includes: a characteristic peak low-dimensional mapping unit, which is used to encode the characteristic peak low-dimensional embedded coding vector Each characteristic peak low-dimensional embedded coding vector in the set of vectors is mapped to a hyperbolic space to obtain a set of characteristic peak low-dimensional embedded coding mapping feature vectors; a characteristic peak significance description factor calculation unit is used to calculate the feature significance description factor of each characteristic peak low-dimensional embedded coding mapping feature vector in the set of characteristic peak low-dimensional embedded coding mapping feature vectors to obtain a set of characteristic peak low-dimensional embedded coding significance description factors; a characteristic peak low-dimensional embedding enhancement unit is used to perform feature significance enhancement on the set of characteristic peak low-dimensional embedded coding vectors based on the set of characteristic peak low-dimensional embedded coding significance description factors to obtain the set of characteristic peak low-dimensional embedded enhanced coding vectors; a characteristic peak information combination module 160 is used to aggregate the set of characteristic peak low-dimensional embedded enhanced coding vectors to obtain a characteristic peak information joint representation; a target substance concentration decoding module 170 is used to obtain a detection result based on the characteristic peak information joint representation, and the detection result is a decoded value of the concentration of the target substance.

[0082] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned quantitative detection system based on Raman spectroscopy have been described in detail above. Figures 1 to 5 The description of the quantitative detection method based on Raman spectroscopy has been described in detail, and therefore, its repeated description will be omitted.

[0083] As described above, the quantitative detection system 100 based on Raman spectroscopy according to the embodiment of the present disclosure can be implemented in various wireless terminals, such as a server with a quantitative detection algorithm based on Raman spectroscopy. In a possible implementation, the quantitative detection system 100 based on Raman spectroscopy according to the embodiment of the present disclosure can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the quantitative detection system 100 based on Raman spectroscopy can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the quantitative detection system 100 based on Raman spectroscopy can also be one of the many hardware modules of the wireless terminal.

[0084] Alternatively, in another example, the Raman spectroscopy-based quantitative detection system 100 and the wireless terminal may be separate devices, and the Raman spectroscopy-based quantitative detection system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in an agreed data format.

[0085] The above is merely an example of the principles of the present disclosure, and various modifications may be made by those skilled in the art without departing from the scope of the present disclosure. The above embodiments are presented for the purpose of illustration rather than limitation. The present disclosure may also take many forms other than those explicitly described herein.

Claims

1. A quantitative detection method based on Raman spectroscopy, characterized in that: include: Obtaining a Raman spectrum of the sample to be tested; Extracting the intensity value of each characteristic peak from the Raman spectrum; Performing one-hot encoding on each of the characteristic peaks to obtain a set of characteristic peak one-hot encoding vectors; Adding the intensity value of each characteristic peak at the end of each characteristic peak one-hot encoding vector in the set of characteristic peak one-hot encoding vectors to obtain a set of characteristic peak low-dimensional embedded encoding vectors; The set of the feature peak low-dimensional embedded coding vectors is subjected to feature peak low-dimensional sequence feature enhancement to obtain a set of feature peak low-dimensional embedded coding vectors, including: mapping each feature peak low-dimensional embedded coding vector in the set of the feature peak low-dimensional embedded coding vectors to a hyperbolic space to obtain a set of feature peak low-dimensional embedded coding mapping feature vectors; calculating a feature significance description factor of each feature peak low-dimensional embedded coding mapping feature vector in the set of the feature peak low-dimensional embedded coding mapping feature vectors to obtain a set of feature peak low-dimensional embedded coding significance description factors; based on the set of feature peak low-dimensional embedded coding significance description factors, the set of the feature peak low-dimensional embedded coding vectors is subjected to feature significance enhancement to obtain a set of feature peak low-dimensional embedded coding enhancement coding vectors; Aggregating the set of the feature peak low-dimensional embedded enhanced coding vectors to obtain a joint representation of feature peak information; Based on the joint representation of the characteristic peak information, a detection result is obtained, and the detection result is a decoded value of the concentration of the target substance.

2. The quantitative detection method based on Raman spectroscopy according to claim 1, characterized in that: Each feature peak low-dimensional embedded coding vector in the set of feature peak low-dimensional embedded coding vectors is mapped to a hyperbolic space to obtain a set of feature peak low-dimensional embedded coding mapping feature vectors, including: after matrix-vector multiplication of a first weight matrix and the feature peak low-dimensional embedded coding vector, vector-matrix multiplication of the obtained first feature peak low-dimensional embedded coding weight vector and a second weight matrix to obtain the feature peak low-dimensional embedded coding mapping feature vector.

3. The quantitative detection method based on Raman spectroscopy according to claim 2, characterized in that: Calculating the feature significance description factor of each feature peak low-dimensional embedded coding mapping feature vector in the set of feature peak low-dimensional embedded coding mapping feature vectors to obtain a set of feature peak low-dimensional embedded coding significance description factors, including: Calculating the square of a norm of the feature peak low-dimensional embedded coding mapping feature vector to obtain a feature peak low-dimensional embedded coding mapping feature norm value; Adding and subtracting a constant one from the characteristic peak low-dimensional embedded coding mapping feature norm value to obtain a first characteristic peak low-dimensional embedded coding mapping feature norm modulation value and a second characteristic peak low-dimensional embedded coding mapping feature norm modulation value, and dividing the first characteristic peak low-dimensional embedded coding mapping feature norm modulation value by the second characteristic peak low-dimensional embedded coding mapping feature norm modulation value to obtain a characteristic peak low-dimensional embedded coding mapping feature norm modulation value; The logarithmic function value of the logarithm of the modulation value of the feature norm of the feature peak low-dimensional embedded coding mapping is calculated with base two to obtain the significance description factor of the feature peak low-dimensional embedded coding.

4. The quantitative detection method based on Raman spectroscopy according to claim 3, characterized in that: Based on the set of feature peak low-dimensional embedded coding significance description factors, the set of feature peak low-dimensional embedded coding vectors is significantly enhanced to obtain the set of feature peak low-dimensional embedded enhanced coding vectors, including: Performing hierarchical mask modulation on the set of characteristic peak low-dimensional embedded coding significant description factors to obtain a set of characteristic peak low-dimensional embedded coding significant modulation weights; Taking each characteristic peak low-dimensional embedded coding significant modulation weight in the set of characteristic peak low-dimensional embedded coding significant modulation weights as a weight, the set of characteristic peak low-dimensional embedded coding vectors is weighted enhanced to obtain the set of characteristic peak low-dimensional embedded enhanced coding vectors.

5. The quantitative detection method based on Raman spectroscopy according to claim 4, characterized in that: The set of the characteristic peak low-dimensional embedded coding significant description factors is subjected to hierarchical mask modulation to obtain a set of characteristic peak low-dimensional embedded coding significant modulation weights, including: Normalizing the set of characteristic peak low-dimensional embedded coding significant description factors to obtain a set of normalized characteristic peak low-dimensional embedded coding significant description factors; The set of normalized feature peak low-dimensional embedded coding significant description factors is hierarchically modulated to obtain a set of feature peak low-dimensional embedded coding significant modulation weights.

6. The quantitative detection method based on Raman spectroscopy according to claim 5, characterized in that: The set of the characteristic peak low-dimensional embedded coding significant description factors is normalized to obtain a set of normalized characteristic peak low-dimensional embedded coding significant description factors, including: Calculate an exponential function value with the natural constant e as the base and each characteristic peak low-dimensional embedded coding significance description factor in the set of characteristic peak low-dimensional embedded coding significance description factors as the exponent to obtain a set of characteristic peak low-dimensional embedded coding exponential significance description factors; Calculating the position-wise sum of the set of the characteristic peak low-dimensional embedded coding index significance description factors to obtain the characteristic peak low-dimensional embedded coding index total sum significance description factor; The position-by-position division between the set of the feature peak low-dimensional embedded coding index significance description factors and the total sum of the feature peak low-dimensional embedded coding index significance description factors is calculated to obtain the set of the feature peak low-dimensional embedded coding index significance description factors.

7. The quantitative detection method based on Raman spectroscopy according to claim 6, characterized in that: The set of the normalized feature peak low-dimensional embedded coding significant description factors is hierarchically modulated to obtain a set of the feature peak low-dimensional embedded coding significant modulation weights, including: Comparing each normalized feature peak low-dimensional embedded coding significant description factor in the set of normalized feature peak low-dimensional embedded coding significant description factors with a preset threshold to obtain a set of feature peak low-dimensional embedded coding significant modulation weights; In which, in response to the normalized feature peak low-dimensional embedded coding significance description factor being greater than a second preset threshold, the product of the normalized feature peak low-dimensional embedded coding significance description factor and the first feature scaling factor is used as the feature peak low-dimensional embedded coding significance modulation weight; In response to the normalized feature peak low-dimensional embedded coding significance description factor being greater than a first preset threshold and less than or equal to a second preset threshold, using the normalized feature peak low-dimensional embedded coding significance description factor as the feature peak low-dimensional embedded coding significance modulation weight; In response to the normalized feature peak low-dimensional embedded coding significance description factor being less than or equal to a first preset threshold, the product of the normalized feature peak low-dimensional embedded coding significance description factor and the second feature scaling factor is used as the feature peak low-dimensional embedded coding significance modulation weight.

8. The quantitative detection method based on Raman spectroscopy according to claim 7, characterized in that: The set of the feature peak low-dimensional embedded enhanced coding vectors is aggregated to obtain a joint representation of feature peak information, including: inputting the set of the feature peak low-dimensional embedded enhanced coding vectors into a linear aggregation network to obtain a feature peak information joint representation vector as the feature peak information joint representation.

9. The quantitative detection method based on Raman spectroscopy according to claim 8, characterized in that: Based on the joint representation of the characteristic peak information, a detection result is obtained, including: inputting the joint representation vector of the characteristic peak information into a target substance concentration detection module based on a decoder to obtain the detection result.

10. A quantitative detection system based on Raman spectroscopy, characterized in that: include: A Raman spectrum acquisition module for the sample to be tested, used to obtain the Raman spectrum of the sample to be tested; A characteristic peak intensity value extraction module, used to extract the intensity value of each characteristic peak from the Raman spectrum; A characteristic peak one-hot encoding module, used for performing one-hot encoding on each characteristic peak to obtain a set of characteristic peak one-hot encoding vectors; A characteristic peak low-dimensional embedding coding module, used for adding the intensity value of each characteristic peak at the end of each characteristic peak one-hot encoding vector in the set of characteristic peak one-hot encoding vectors to obtain a set of characteristic peak low-dimensional embedding coding vectors; A feature peak sequence enhancement module is used to perform feature peak low-dimensional sequence feature enhancement on the set of feature peak low-dimensional embedded coding vectors to obtain a set of feature peak low-dimensional embedded enhanced coding vectors, wherein the feature peak sequence enhancement module includes: a feature peak low-dimensional mapping unit, used to map each feature peak low-dimensional embedded coding vector in the set of feature peak low-dimensional embedded coding vectors to a hyperbolic space to obtain a set of feature peak low-dimensional embedded coding mapping feature vectors; a feature peak significance description factor calculation unit, used to calculate the feature significance description factor of each feature peak low-dimensional embedded coding mapping feature vector in the set of feature peak low-dimensional embedded coding mapping feature vectors to obtain a set of feature peak low-dimensional embedded coding significance description factors; a feature peak low-dimensional embedding enhancement unit, used to perform feature significance enhancement on the set of feature peak low-dimensional embedded coding vectors based on the set of feature peak low-dimensional embedded coding significance description factors to obtain the set of feature peak low-dimensional embedded enhanced coding vectors; A feature peak information joint module, used for aggregating the set of feature peak low-dimensional embedded enhanced coding vectors to obtain a feature peak information joint representation; The target substance concentration decoding module is used to obtain a detection result based on the joint representation of the characteristic peak information, and the detection result is a decoded value of the concentration of the target substance.

Images