Chiral sample classification method for mid-infrared spectroscopy

Abstract

The invention discloses a chiral sample classification method for mid-infrared spectroscopy, which comprises the following steps: acquiring sample data by using a Fourier infrared spectroscopy system, calculating the transmittance and absorption coefficient of a sample in an infrared band, and recording the data as experimental data of a chiral isomer; constructing a sample molecular structural formula; electron density and energy of molecules are calculated, energy of different excited states and corresponding wave functions are calculated by using a time-dependent density functional theory, the energy of the excited states is converted into frequency, and the absorption peak frequency of the material is obtained and serves as theoretical data of a chiral isomer; extracting a chiral isomer frequency band based on a density functional theory; and carrying out feature classification by using a support vector machine. According to the method, characterization analysis of sample vibration spectrum information and detection results can be achieved, effective vibration frequency bands in the molecular structure are extracted based on the density functional theory, theoretical support is provided for chiral molecule identification, and the accuracy and generalization ability of classification results are further improved.

Description

Technical Field

[0001] The invention belongs to the field of infrared spectrum analysis and classification and identification, and in particular relates to a chiral sample classification method for mid-infrared spectrum. Background Art

[0002] Mid-infrared spectroscopy (MIR) is an analytical technique that uses the absorption, emission or transmission characteristics of infrared light in the mid-infrared region (wavelength range is usually defined as 2.5 to 25 microns) to identify the structure and composition of compounds. Mid-infrared spectroscopy is an important branch of infrared spectroscopy because it covers the vibrational spectral information of most organic and inorganic compounds. Chiral drugs contain chiral centers in their molecular structures, so there are two stereoisomers, namely enantiomers. These two enantiomers may show significant differences in chemical and biological properties, including efficacy, toxicity and metabolic pathways. Given that these differences may result in one enantiomer having a therapeutic effect while the other may have side effects or even toxicity, it is crucial to accurately analyze and detect the enantiomers of chiral drugs in drug development, production and clinical applications.

[0003] There are various detection methods for chiral drugs, including high performance liquid chromatography, gas chromatography, capillary electrophoresis, etc., all of which can effectively separate and determine the enantiomers of chiral drugs. At the same time, spectroscopic methods such as infrared spectroscopy and nuclear magnetic resonance are also widely used in the structural characterization and analysis of chiral drugs. Typically, these analyses involve extracting key features from pre-processed spectral data, and then using these features to reflect the differences between different samples to achieve chiral classification. In general, the detection of chiral drugs requires the use of highly sensitive, highly selective and high-resolution analytical methods to ensure the safety and effectiveness of the drugs.

[0004] However, in terms of feature extraction methods, compared with in-depth research at the molecular level, the uniformity of existing models needs to be improved and the interpretability is not strong enough. From the perspective of spectral characteristics, existing methods have not fully considered the deep connection between the spectrum and sample molecules and lack solid theoretical support. Therefore, when targeting samples of the same type or chiral molecular samples, the spectral features are difficult to clearly distinguish, resulting in low accuracy of feature extraction. To address this problem, this paper proposes a new method for classifying chiral samples in the mid-infrared band. This method can effectively select the frequency band position in the infrared spectrum and achieve accurate identification of chiral molecules by classifying effective vibrational characteristic frequency bands.

[0005] Therefore, the infrared spectrum classification method for chiral samples proposed in the present invention includes the following steps: first, the absorption coefficient of the chiral isomer sample in the infrared band is collected by Fourier transform infrared spectroscopy system, and the frequency point position and spectral intensity are recorded as experimental data; secondly, the theoretical vibration absorption spectrum of the sample is calculated by density functional theory, and the theoretical and experimental analysis are combined to determine the main frequency point position and match; finally, based on the obtained frequency point information, the Euclidean distance algorithm is used to extract the data containing the main vibration characteristic frequency band, and classify and identify it. This series of steps constitutes a complete chiral sample classification method, covering multiple links such as experimental data collection, theoretical calculation, feature extraction and machine learning. Compared with the traditional feature extraction method, the present invention can not only accurately reveal the main vibration mode of the characteristic variable molecule, but also significantly improve the accuracy of sample classification, and has important application value in the field of chiral sample classification and identification. Summary of the invention

[0006] The purpose of the present invention is to propose a chiral sample classification method for mid-infrared spectroscopy to solve the problem that the spectral characteristics of chiral molecular samples are difficult to distinguish and lack theoretical support.

[0007] In order to achieve the above-mentioned invention object, the present invention adopts the following technical solutions:

[0008] A chiral sample classification method for mid-infrared spectroscopy comprises the following steps

[0009] S1. Collect sample data using the Fourier infrared spectroscopy system, calculate the transmittance and absorption coefficient of the sample in the infrared band, and record the above data as experimental data of chiral isomers;

[0010] S2. Construct sample molecular structure formula;

[0011] S3. Calculate the electron density and energy of the molecule, use time-dependent density functional theory to calculate the energy of different excited states and the corresponding wave functions, convert the excited state energy into frequency, and obtain the absorption peak frequency of the material as the theoretical data of the chiral isomer;

[0012] S4. Extraction of chiral isomer bands based on density functional theory;

[0013] S5. Use support vector machine for feature classification.

[0014] Furthermore, the step S1 comprises:

[0015] The Fourier infrared spectroscopy system is used to collect the light intensity I of the infrared light after passing through the sample and the light intensity I0 after passing through the background, and the transmittance T of the sample in the infrared band is calculated:

[0016]

[0017] Calculate the absorption coefficient A of the sample in the infrared band:

[0018]

[0019] Where T is the transmittance and d is the sample thickness.

[0020] Furthermore, the step S2 comprises:

[0021] According to the chemical structure formula of the molecule, carbon, hydrogen and oxygen atoms are selected as basic elements to place the corresponding number of atoms, and the atoms are connected to form the skeleton structure of the molecule. These parts are constructed one by one and connected to form a complete molecular structure. Finally, all molecular structures are structurally optimized and energy optimized.

[0022] Furthermore, the step S3 comprises:

[0023] First, the B3LYP / 6-31G(d) basis set and exchange-correlation functional are selected to calculate the electronic structure density and energy level of the molecule to obtain the electron density and energy of the system. Then, the time-dependent density functional theory is used to calculate the excited state properties of the system to obtain the energy and corresponding wave functions of different excited states. The excited state energy is converted into frequency to obtain the absorption peak frequency of the material, and this data is recorded as the theoretical data of the chiral isomer sample.

[0024] Furthermore, the step S4 comprises:

[0025] S41. Divide the frequency band boundaries based on the molecular vibration trend, divide the frequency bands, and assign different weights to different frequency bands;

[0026] S42. extract the frequency points corresponding to the absorption peaks of the experimental spectrum and the theoretical spectrum in each frequency band, calculate the difference ρ between the experimental spectrum and the theoretical spectrum at the corresponding frequency points to evaluate the similarity between them, and set a threshold t1 to eliminate the experimental spectrum frequency points whose difference ρ is less than the threshold;

[0027] S43. Perform weighted similarity measurement on each frequency band and calculate the weighted total spectral similarity And set the threshold t2 to eliminate the similarity Experimental spectral frequency data below the threshold;

[0028] S44. Record the maximum value λe(max) and the minimum value λe(min) of the experimental spectrum frequency point in each remaining frequency band, and update the interval S of each frequency band.

[0029] Furthermore, in step S41, based on the spectral results obtained theoretically, each frequency point with similar vibration trend is taken as an effective frequency band to obtain the main molecular vibration state of each frequency band, and the number of divided frequency bands is recorded as i; different weights w(i) are assigned to different frequency bands, and the sum of the weights of all frequency bands is 1.

[0030] Furthermore, in step S42, the frequency point of the experimental spectrum absorption peak is recorded as λe, the frequency point of the theoretical spectrum absorption peak is recorded as λc, and the difference ρ between the experimental spectrum and the theoretical spectrum at the corresponding frequency point is calculated:

[0031] ρ=|λe-λc|.

[0032] Furthermore, in step S43, the weighted total spectral similarity The calculation formula is:

[0033]

[0034] Where w(i) is the weight w(i) of the corresponding frequency band, i is the number of frequency bands, and ρ is the difference between the experimental spectrum and the theoretical spectrum at the corresponding frequency point.

[0035] Furthermore, the step S5 comprises:

[0036] S51. The characteristic intervals of the chiral isomers are respectively recorded as S1 and S2, the characteristic frequency band data S1 and S2 of the chiral molecule are read, and the imported data is divided into a training set P and a test set T according to a ratio of 7:3 for model training and verification;

[0037] S52. Use the svmtrain function to train the SVM model;

[0038] S53. The model performance is optimized by adjusting the penalty factor c and the radial basis function parameter g, and the characteristic frequency band data is used as the variable number to construct the SVM classification model.

[0039] The present invention has the following beneficial technical effects:

[0040] The present invention proposes a chiral sample classification method for mid-infrared spectroscopy, which can not only realize the characterization and analysis of sample vibration spectrum information and detection results, but also extract the effective vibration frequency band in the molecular structure based on density functional theory, provide theoretical support for chiral molecule identification, and further improve the accuracy and generalization ability of classification results. This method provides a basis for mid-infrared material identification and is of great significance in the fields of biology, material identification, food safety, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is an overall flow chart of a chiral sample classification method for mid-infrared spectroscopy according to the present invention;

[0042] Figure 2 is a flow chart of a specific method of an embodiment of the present invention;

[0043] Figure 3 The mid-infrared transmission spectra of Ibuprofen and S(+)-Ibuprofen samples in the embodiments of the present invention are shown;

[0044] Figure 4 The mid-infrared absorption spectra of Ibuprofen and S(+)-Ibuprofen samples in the embodiments of the present invention are shown;

[0045] Figure 5 The structural models of the molecules of Ibuprofen and S(+)-Ibuprofen samples in the embodiments of the present invention;

[0046] Figure 6 The experimental and theoretical absorption spectra of Ibuprofen and S(+)-Ibuprofen samples in the embodiments of the present invention are shown;

[0047] Figure 7 is the similarity between the experimental spectrum and the theoretical spectrum of the Ibuprofen and S(+)-Ibuprofen samples in the embodiments of the present invention at the corresponding frequency points Line chart of . DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0049] The present invention provides a chiral sample classification method for mid-infrared spectroscopy, comprising the following steps:

[0050] S1. Collect sample data using the Fourier infrared spectroscopy system, calculate the transmittance T and absorption coefficient A of the sample in the infrared band, and record the above data as the experimental data of the chiral isomer:

[0051] The Fourier transform infrared spectroscopy system is used to collect the light intensity I after the infrared light passes through the sample and the light intensity I0 after passing through the background, and the transmittance T of the sample in the infrared band is calculated by formula (1).

[0052]

[0053] The absorption coefficient A of the sample in the infrared band is obtained from Formula 2. The absorption coefficient A is the logarithm of the transmittance T multiplied by negative one, and finally divided by the sample thickness d.

[0054]

[0055] S2. Construct sample molecular structure formula:

[0056] According to the chemical structure formula of the molecule, carbon (C), hydrogen (H) and oxygen (O) atoms are selected as basic elements to place the corresponding number of atoms, and the atoms are connected to form the skeleton structure of the molecule. These parts are constructed one by one and connected to form a complete molecular structure. Finally, all molecular structures are structurally optimized and energy optimized.

[0057] S3. Calculate the electron density and energy of the molecule, use time-dependent density functional theory to calculate the energy of different excited states and the corresponding wave functions, convert the excited state energy into frequency, and obtain the absorption peak frequency of the material as the theoretical data of the chiral isomer:

[0058] Density Functional Theory (DFT) uses electron density as the basic variable to approximate the Schrödinger equation of a multi-electron system to a single-electron equation. First, the B3LYP / 6-31G(d) basis set and exchange-correlation functional are selected to calculate the electronic structure density and energy level of the molecule to obtain the electron density and energy of the system. Then, the time-dependent density functional theory (TDDFT) is used to calculate the excited state properties of the system, obtain the energy and corresponding wave functions of different excited states, convert the excited state energy into frequency, obtain the absorption peak frequency of the material, and record this data as the theoretical data of the chiral isomer sample.

[0059] S4. Extraction of chiral isomer bands based on density functional theory, such as Figure 2 As shown:

[0060] S41. Divide the frequency band boundaries based on the molecular vibration trend, divide the frequency bands, and assign different weights to different frequency bands:

[0061] Based on density functional theory, the characteristic peaks are attributed and analyzed, and then the frequency bands are divided. The determination of molecular vibration frequency bands involves multiple factors, mainly including molecular structure, type and strength of chemical bonds, atomic mass, etc. The molecular vibration state at each frequency point is different but the vibration trend is similar. According to the spectral results obtained theoretically, each frequency point with similar vibration trend is regarded as an effective frequency band, and the main molecular vibration state of each frequency band is preliminarily obtained. The number of main frequency bands divided is recorded as i. This method can flexibly adjust the boundaries of the frequency bands according to the characteristics of the spectral data to better reflect the subtle changes in the spectrum. Different weights w(i) are assigned to different frequency bands according to the importance and credibility of the frequency bands. i represents the position of the frequency band, and the sum of the weights of all frequency bands is 1, as shown in formula (3);

[0062]

[0063] Where: i represents the number of each frequency band, and n represents the total number of frequency bands.

[0064] S42. In each frequency band, the frequency points corresponding to the absorption peaks of the experimental spectrum and the theoretical spectrum are extracted respectively, where the frequency point of the absorption peak of the experimental spectrum is denoted as λe, and the frequency point of the absorption peak of the theoretical spectrum is denoted as λc. The difference ρ between the experimental spectrum and the theoretical spectrum at the corresponding frequency points is calculated to evaluate the similarity between them, as shown in formula (4), and a threshold t1 is set to eliminate the experimental spectrum frequency points whose difference ρ is less than the threshold.

[0065] ρ= |λe-λc| (4)

[0066] S43. Perform weighted similarity measurement on each frequency band. Multiply the difference ρ by the weight w(i) of its corresponding frequency band to calculate the weighted total spectral similarity And set the threshold t2 to eliminate the similarity The experimental spectrum frequency data below the threshold is shown in formula (5):

[0067]

[0068] Where: i represents the number of each frequency band, and n represents the total number of frequency bands.

[0069] S44. Record the maximum value λe(max) and the minimum value λe(min) of the experimental spectrum frequency points in each remaining frequency band, and update each frequency band interval S. The range of each frequency band interval S can be expressed as: λe(min)~λe(max).

[0070] S5. Feature classification using support vector machine (SVM):

[0071] S51. The characteristic intervals of the chiral isomers are respectively recorded as S1 and S2. First, the characteristic frequency band data S1 and S2 of the chiral molecule are read, and the imported data is divided into a training set P and a test set T in a ratio of 7:3 for model training and verification.

[0072] S52. Use the svmtrain function to train the SVM model. It optimizes the objective function of the SVM model in an iterative manner to find the optimal decision boundary. First, initialize the Lagrangian factor a to 0. In each iteration, select two Lagrangian factors a. i Update. The new factor value needs to satisfy the constraint to update the corresponding weight vector, as shown in Formula 6. Then check whether the conditions are met to determine whether the current solution is close to the optimal. If the conditions of all points are met or the maximum number of iterations is reached, stop the iteration. Otherwise, continue to select new factor values ​​for update. The final decision function is shown in Formula 7.

[0073] ∑a i y i =0; 0< ai <c (6)

[0074]

[0075] Where: y is the support vector x i The category label, K(x i , x j ) is the kernel function and b is the bias.

[0076] S53. In order to avoid overfitting and underfitting, the penalty factor c and radial basis function parameter g are adjusted to optimize the model performance. The characteristic frequency band data are used as the variable number to construct the SVM classification model.

[0077] Embodiment 1:

[0078] Ibuprofen, as an antipyretic and analgesic non-steroidal anti-inflammatory drug, has chiral properties. Specifically, the molecular structure of ibuprofen contains a chiral center, that is, four different groups are connected to a carbon atom, which makes the ibuprofen molecule have two enantiomers: S(+)-ibuprofen (dextrorotatory ibuprofen, S(+)-Ibuprofen) and R(-)-ibuprofen (levorotatory ibuprofen, R(-)-Ibuprofen). At present, the identification technology of chiral samples is limited by many conditions, such as: limitations of chiral stationary phases, selectors, instruments and equipment, and costs. Therefore, it is necessary to adopt a method for classifying and identifying chiral isomers as described in the present invention. The specific process is as follows. Figure 1 , Figure 2 shown.

[0079] This embodiment includes the following steps:

[0080] Step 1: Get the wave number (cm -1 ) is the horizontal axis and the amplitude is the vertical axis. The mid-infrared spectra of Ibuprofen and S(+)-Ibuprofen show the degree of attenuation of infrared radiation when passing through Ibuprofen and S(+)-Ibuprofen samples, as shown in Figure 3 As shown in the figure, the red solid line represents the experimentally obtained Ibuprofen isomer transmission spectrum, and the blue solid line represents the experimentally obtained S(+)-Ibuprofen isomer transmission spectrum. The infrared spectrum is subjected to absorbance calculation to obtain the Ibuprofen and S(+)-Ibuprofen absorption coefficient spectra with wave number as the horizontal axis and amplitude as the vertical axis, as shown in FIG. Figure 4As shown, it can be seen that there are multiple peaks and valleys in the mid-infrared spectral signal, representing the changes in molecular structure and chemical bond properties at different wavelengths. The red solid line in the figure represents the experimental absorption spectrum of the Ibuprofen isomer, and the blue solid line represents the experimental absorption spectrum of the S(+)-Ibuprofen isomer.

[0081] Step 2: Construct the structural model of the required molecules of Ibuprofen and S(+)-Ibuprofen in the chemical structure editing software, such as Figure 5 shown.

[0082] Step 3: Use quantum chemical calculation software to calculate the absorption spectra of Ibuprofen and S(+)-Ibuprofen molecules and the absorption spectra of Ibuprofen and S(+)-Ibuprofen obtained experimentally, respectively, as follows: Figure 6 As shown in (a) and (b), the absence of imaginary frequency indicates that the minimum energy structure of the molecule has been optimized. The red solid line in the figure represents the absorption spectrum of the ibuprofen isomer obtained by theoretical calculation, and the blue solid line represents the absorption spectrum of the ibuprofen isomer obtained by experiment. Since the experimental spectrum is based on the amorphous solid sample structure and there are molecular interactions, it causes certain deviations.

[0083] Step 4: Get the molecular effective vibration frequency band division results and weight distribution w(i), as shown in Table 1:

[0084] Table 1 Weights and vibration modes corresponding to characteristic frequency bands of Ibuprofen and S(+)-Ibuprofen samples

[0085]

[0086]

[0087] It can be seen that the main characterization results corresponding to the ibuprofen isomers have similar distributions of vibration modes in the corresponding peak frequency bands, which are used as the number of variables for preliminary feature band extraction. The difference value ρ of some frequency points is shown in the following formula:

[0088] ρ1=|402.7602-405|=2.2398

[0089] ρ2=|418.0481-418|=0.0481

[0090] ρ3=|422.2812-427|=4.7188

[0091] ρ4=|432.7688-436|=3.2312

[0092] ρ5=|443.6553-445|=1.3447

[0093] …

[0094] Threshold t1 is set to 3.5, and threshold t2 is set to 3.0. After threshold screening, Figure 7 Shows the similarity between the experimental spectrum and the theoretical spectrum at the corresponding frequency points The line graph of the chiral isomers finally extracted is as follows:

[0095] S1:443.0481~552.7385cm -1

[0096] S2:904.0638~940.6843cm -1

[0097] S3:1192.0972~1361.7501cm -1

[0098] S4:1649.9750~1658.1219cm -1

[0099] S5:3037.5506~3135.2990cm -1

[0100] S6:3666.8918~3702.7940cm-1

[0101] Step 5: Use support vector machine (SVM) for feature classification, and adjust the parameters to obtain the best performance of the model, including penalty coefficient c and kernel function parameter g. In order to avoid overfitting and underfitting, the penalty coefficient c ranges from 1 to 20, and the kernel function parameter g ranges from 0.01 to 10. The characteristic frequency band data is used as the number of variables to construct the SVM classification model. Table 2 shows the comparison results of the model output accuracy of the mid-infrared frequency band extraction method based on density functional theory compared with the SPA and CARS feature extraction methods.

[0102] Table 2 SVM feature classification results

[0103]

[0104] From the comparison, it can be seen that the method proposed in the present invention can effectively improve the accuracy of sample classification. Compared with the traditional SPA and CARS feature extraction methods, the mid-infrared frequency band extraction method based on density functional theory considers the difference between the two samples from the perspective of molecular vibration. It can not only obtain the main vibration modes of the corresponding characteristic variable molecules, but also improve the accuracy of sample classification.

[0105] The embodiments described above are only preferred embodiments of the present invention and are not intended to limit the scope of implementation of the present invention. Therefore, all changes made according to the shape and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A chiral sample classification method for mid-infrared spectroscopy, characterized in that: The following steps are included S1. Collect sample data using the Fourier infrared spectroscopy system, calculate the transmittance and absorption coefficient of the sample in the infrared band, and record the above data as experimental data of chiral isomers; S2. Construct sample molecular structure formula; S3. Calculate the electron density and energy of the molecule, use time-dependent density functional theory to calculate the energy of different excited states and the corresponding wave functions, convert the excited state energy into frequency, and obtain the absorption peak frequency of the material as the theoretical data of the chiral isomer; S4. Extraction of chiral isomer bands based on density functional theory; S5. Use support vector machine for feature classification.

2. A chiral sample classification method for mid-infrared spectroscopy according to claim 1, characterized in that: The step S1 comprises: The Fourier infrared spectroscopy system is used to collect the light intensity I of the infrared light after passing through the sample and the light intensity I0 after passing through the background, and the transmittance T of the sample in the infrared band is calculated: Calculate the absorption coefficient A of the sample in the infrared band: Where T is the transmittance and d is the sample thickness.

3. A chiral sample classification method for mid-infrared spectroscopy according to claim 1, characterized in that: The step S2 comprises: According to the chemical structure formula of the molecule, carbon, hydrogen and oxygen atoms are selected as basic elements to place the corresponding number of atoms, and the atoms are connected to form the skeleton structure of the molecule. These parts are constructed one by one and connected to form a complete molecular structure. Finally, all molecular structures are structurally optimized and energy optimized.

4. A chiral sample classification method for mid-infrared spectroscopy according to claim 1, characterized in that: The step S3 comprises: First, the B3LYP / 6-31G(d) basis set and exchange-correlation functional are selected to calculate the electronic structure density and energy level of the molecule to obtain the electron density and energy of the system. Then, the time-dependent density functional theory is used to calculate the excited state properties of the system to obtain the energy and corresponding wave functions of different excited states. The excited state energy is converted into frequency to obtain the absorption peak frequency of the material, and this data is recorded as the theoretical data of the chiral isomer sample.

5. A chiral sample classification method for mid-infrared spectroscopy according to claim 1, characterized in that: The step S4 comprises: S41. Divide the frequency band boundaries based on the molecular vibration trend, divide the frequency bands, and assign different weights to different frequency bands; S42. extract the frequency points corresponding to the absorption peaks of the experimental spectrum and the theoretical spectrum in each frequency band, calculate the difference ρ between the experimental spectrum and the theoretical spectrum at the corresponding frequency points to evaluate the similarity between them, and set a threshold t1 to eliminate the experimental spectrum frequency points whose difference ρ is less than the threshold; S43. Perform weighted similarity measurement on each frequency band, calculate the weighted total spectral similarity φ, and set a threshold t2 to remove experimental spectral frequency point data with similarity φ lower than the threshold; S44. Record the maximum value λe(max) and the minimum value λe(min) of the experimental spectrum frequency point in each remaining frequency band, and update the interval S of each frequency band.

6. A chiral sample classification method for mid-infrared spectroscopy according to claim 5, characterized in that: In step S41, based on the spectral results obtained theoretically, each frequency point with similar vibration trend is regarded as an effective frequency band to obtain the main molecular vibration state of each frequency band, and the number of divided frequency bands is recorded as i; different weights w(i) are assigned to different frequency bands, and the sum of the weights of all frequency bands is 1.

7. A chiral sample classification method for mid-infrared spectroscopy according to claim 5, characterized in that: In step S42, the frequency point of the experimental spectrum absorption peak is recorded as λe, the frequency point of the theoretical spectrum absorption peak is recorded as λc, and the difference ρ between the experimental spectrum and the theoretical spectrum at the corresponding frequency point is calculated: ρ=|λe-λc|.

8. The chiral sample classification method for mid-infrared spectroscopy according to claim 5, characterized in that: In step S43, the weighted total spectral similarity The calculation formula is: Where w(i) is the weight w(i) of the corresponding frequency band, i is the number of frequency bands, and ρ is the difference between the experimental spectrum and the theoretical spectrum at the corresponding frequency point.

9. A chiral sample classification method for mid-infrared spectroscopy according to claim 1, characterized in that: The step S5 comprises: S51. The characteristic intervals of the chiral isomers are respectively recorded as S1 and S2, the characteristic frequency band data S1 and S2 of the chiral molecule are read, and the imported data is divided into a training set P and a test set T according to a ratio of 7:3 for model training and verification; S52. Use the svmtrain function to train the SVM model; S53. The model performance is optimized by adjusting the penalty factor c and the radial basis function parameter g, and the characteristic frequency band data is used as the variable number to construct the SVM classification model.

Images