Raman spectrum fluorescence baseline correction method and device

Abstract

The invention discloses a Raman spectrum fluorescence baseline correction method and device, and the method comprises the steps: collecting a time domain signal of a to-be-detected sample, and obtaining an IRF function of a to-be-detected sample collection system; pre-processing the time domain signal, and carrying out deconvolution on the pre-processed time domain signal and an IRF function to obtain a deconvolution result; fitting the second half data of the deconvolution result by using a support vector regression model to obtain a fitted curve; wherein a radial basis function is adopted as a kernel function, and hyper-parameters of the support vector regression model are optimized through a grid search and cross validation method; performing difference operation on the deconvolution result and the fitting curve to obtain a Raman attenuation curve; and windowing the Raman attenuation curve and averaging to obtain a Raman spectrum. According to the method, a fluorescence attenuation model and parameters thereof do not need to be preset, the IRF deconvolution result is processed and the time domain baseline is predicted by utilizing the characteristics of the time domain signal data and the support vector regression model, the signal-to-noise ratio of the extracted Raman signal can be greatly improved, a good effect is achieved, and the method is suitable for practical application.

Description

Technical Field

[0001] The present invention relates to the technical field of Raman signal processing, and in particular to a Raman spectrum fluorescence baseline correction method, device, terminal equipment and computer-readable storage medium. Background Art

[0002] With the continuous advancement of spectral analysis technology, Raman spectroscopy, as a non-destructive and non-contact analysis method, has gradually been widely used in materials science, biomedicine, chemistry and other fields. Raman spectroscopy can detect the molecular vibration mode of the material by detecting the inelastic scattering effect between sample photons, obtain the molecular vibration information of the sample, thereby providing in-depth information about the chemical composition and structure of the material, and realizing the composition analysis and structural identification of the sample. Compared with traditional infrared spectroscopy, Raman spectroscopy is not only insensitive to moisture, but also can detect asymmetric vibration modes, so it has significant advantages in the analysis of liquid samples and biological tissues. In recent years, with the continuous development of laser, spectrometer and detector technology, the application scope of Raman spectroscopy has been continuously expanded, especially in the fields of drug testing, food safety and environmental monitoring, showing great potential. However, Raman signals are usually much weaker than other scattering signals and are easily interfered by background noise, so highly sensitive detectors and signal enhancement technology are required. At the same time, the fluorescence interference of the sample and environmental noise will also affect the accuracy of the signal, limiting the application of Raman spectroscopy technology.

[0003] A major challenge in Raman spectroscopy applications is removing interference from the fluorescence background. Since Raman scattering signals are usually several orders of magnitude weaker than the excitation light and fluorescence signals, the fluorescence background can significantly interfere with the interpretation of the Raman signal and affect its accuracy. Therefore, effectively removing the fluorescence background is a key task in Raman data processing. There are many data processing methods for baseline calibration of conventional Raman equipment. The literature (Wold S, Martens H, Wold H, The multivariate calibration problem in chemistry solved by the PLS method, Lecture Notes in Mathematics. Springer, 1983) proposed the partial least squares (PLS) method; the literature (Zhang ZM, Chen S, Liang YZ, Baseline correction using adaptive iteratively reweighted penalized least squares, Analyst. 2010 May; 135(5): 1138-46) improved the PLS method and proposed the penalized least squares (Adaptive Iterative Re-weighted Penalized Least Squares, airPLS); the literature (Eilers, Paul and Boelens, Hans, Baseline Correction with Asymmetric Least Squares Smoothing. Unpubl. Manuscr[J]. 2005) proposed the symmetric least squares method (Asymmetric Least Squares, ALS), and there are also methods that use deep learning and other methods to perform baseline calibration on conventional Raman data. The above methods perform fluorescence baseline calibration on conventional Raman data, which cannot effectively suppress strong fluorescence background noise and has very limited effects. Based on the temporal difference between Raman scattering and fluorescence response, time-resolved technology is increasingly used to suppress fluorescence interference. Time-resolved Raman spectroscopy (TRRS) technology is usually implemented by combining pulsed lasers with high-time-resolution detectors (such as photomultiplier tubes PMT, high-speed optical shutters based on Kerr units, enhanced charge-coupled devices ICCD, and single-photon avalanche diodes (SPAD)). For example, the combination of picosecond pulsed lasers and ICCD detectors can be used to penetrate polymers or plastic packaging to detect explosives and identify minerals at different depths; SPAD-based TCSPC technology can achieve rapid acquisition of TRFS.Among many time-resolved techniques, time-correlated single photon counting (TCSPC) has become an ideal choice for suppressing fluorescence background due to its low noise, high detection efficiency and compact size. Many studies have shown that the time-domain Raman signal obtained by TCSPC technology contains rich Raman information. If the fluorescence baseline can be corrected for the time-domain Raman signal, the sensitivity and accuracy of spectral detection will be significantly improved. However, the intensity of Raman signals is usually very low compared to fluorescence signals (about one percent), and due to system noise, instrument response function (IRF) and nonlinear effects, it brings considerable challenges to designing a suitable fluorescence baseline correction algorithm.

[0004] At present, the signal processing methods commonly used to correct fluorescence curves for time domain data to obtain Raman spectra are mainly direct methods and inversion methods based on exponential decay model fitting. The direct method is to perform IRF deconvolution on the time domain data and then directly take a window on the deconvolution result to obtain the Raman spectrum. This method is simple and easy to operate, but it ideally assumes that the Raman signal intensity is obvious, ignoring that the Raman intensity usually only accounts for one percent of the fluorescence intensity or even less. The Raman signal is very easy to be submerged in the fluorescence background noise. This method is often difficult to obtain a high signal-to-noise ratio Raman spectrum. The inversion method based on exponential decay model fitting is to fit the fluorescence curve using the exponential decay model and then remove the fluorescence signal. It is the mainstream method for obtaining Raman spectra. However, there are still many problems. For example, although the traditional multi-exponential fitting method can fit the fluorescence decay curve, it is easy to cause inaccurate fitting results when facing a high noise environment or when the component is unknown and the exponent is uncertain. The setting of the initial parameters will also bring unexpected results. For example, the literature (J. Kekkonen, T. Talala and I. Nissinen, Time-and spectrally-resolved mesoscopic Raman and fluorescence imaging of carious enamel by a CMOS SPAD-based spectrometer, 2023 IEEE International Instrumentation and Measurement Technology Conference (I2MTC), Kuala Lumpur, Malaysia, 2023) uses a single exponential decay model to fit the fluorescence baseline of the tooth sample data. The literature states that the fluorescence decay curve of the tooth biological sample is assumed to be a single exponential decay, and then the deconvolution data is exponentially fitted to correct the fluorescence baseline and extract the Raman signal of the tooth biological sample. This approach is an idealized simplification of the fluorescence decay mode and lacks rigorous theoretical support. Similarly, patent number CN202110290121.4 uses the least squares method to fit the fluorescence lifetime curve and establish the measurement relationship between multi-photons and single photons. This method also presets the fluorescence decay model and its parameters.The literature (N. Krstajić et al., A256×8 SPAD line sensor for time resolvedfluorescence and raman sensing, ESSCIRC 2014-40th European Solid StateCircuitsConference(ESSCIRC), Venice Lido, Italy, 2014, pp143-146) uses the nonlinear least squares method to extract Raman signals, and also uses the exponential decay model. Similarly, patent CN201810940171.0 also uses the fluorescence decay model to process the time domain signal to obtain the Raman spectrum. It can be seen that the existing extraction method that uses the exponential decay model to fit the fluorescence curve (whether it is a single exponential model or a multi-exponential model) often requires a lot of prior knowledge and presets a large number of parameters. The processing complexity is high and it is impossible to achieve efficient processing and high generalization while ensuring accuracy. The literature (Joseph R, Lakowicz, Principles of Fluorescence Spectroscop, Springer, 2006) points out that the fitting results of different exponential models can be very close, and the same exponential model can have completely different results due to different preset parameters. Therefore, even if the exponential decay model is used to fit the fluorescence lifetime curve, the obtained parameters may be far from the actual fluorescence lifetime. The composition of fluorophores in reality is often more complex, making it more difficult and unpredictable to remove the fluorescence baseline using the fluorescence decay model. Summary of the invention

[0005] In order to solve the deficiencies of the above-mentioned prior art, the present invention provides a Raman spectroscopy fluorescence baseline correction method, device, terminal device and computer-readable storage medium, which do not preset the fluorescence attenuation model and its parameters, utilize the characteristics of the time domain signal data itself, and use support vector regression (SVR) to process the IRF deconvolution results to predict the time domain baseline, which can greatly improve the signal-to-noise ratio of the extracted Raman signal, achieve good results, and have practical significance.

[0006] The first object of the present invention is to provide a Raman spectrum fluorescence baseline correction method.

[0007] The second object of the present invention is to provide a Raman spectrum fluorescence baseline correction device.

[0008] The third object of the present invention is to provide a terminal device.

[0009] A fourth object of the present invention is to provide a computer-readable storage medium.

[0010] The first object of the present invention can be achieved by adopting the following technical solutions: A method for Raman spectroscopy fluorescence baseline correction, the method comprising: Collecting the time domain signal of the sample to be tested, and obtaining the IRF function of the acquisition system of the sample to be tested; Preprocess the time domain signal, deconvolve the preprocessed time domain signal with the IRF function to obtain a deconvolution result; The support vector regression model is used to fit the second half of the deconvolution result data to obtain a fitting curve; wherein the radial basis function is used as the kernel function, and the hyperparameters of the support vector regression model are optimized through grid search and cross-validation methods; Perform a difference operation between the deconvolution result and the fitting curve to obtain the Raman attenuation curve; The Raman attenuation curve is windowed and averaged to obtain the Raman spectrum.

[0011] Furthermore, the support vector regression model finds the optimal regression function on the input data set so that most of the data points are within the tolerance of the function; at the same time, slack variables are introduced to handle inevitable errors.

[0012] Furthermore, let the introduced slack variable be and , then the objective function is: And satisfy the constraints: Where ω and b are the parameters of the model; C is the regularization parameter, which controls the trade-off between the complexity and error of the model; y i is the i-th data sample point x i The corresponding actual output, n is the total number of data sample points, and ϵ is the tolerance.

[0013] Furthermore, the regression function of the support vector regression model is: Among them, the Gaussian radial basis function is: Where m is the number of support vectors, x i is the i-th data sample point; and is the Lagrange multiplier, which means x i The weight of is a kernel function used to calculate the input data sample points x and x iThe kernel function determines the mapping method to the high-dimensional feature space; b is the parameter of the model; γ is the parameter of the kernel function, which is used to control the width of the Gaussian function; x j is the jth data sample point.

[0014] Furthermore, the preprocessing includes noise reduction and outlier removal.

[0015] Furthermore, the sample acquisition system to be tested is a time-correlated single-photon counting system, which is composed of a pulsed laser, an optical path module, a photodetector, a single-photon avalanche diode detector and a TCSPC system. The pulsed laser is divided into a first light beam and a second light beam through the optical path module; after the first light beam passes through the photodetector, the obtained electrical signal is transmitted to the TCSPC system as a reference synchronization signal to record the collected time domain signal; the second light beam passes through the optical path module to couple the laser into the microscope objective and focus it on the sample to be tested to obtain a time domain signal; the time domain signal is coupled to the optical path module after being collimated by the microscope objective, and is coupled to the single-photon avalanche diode detector through a spectrometer for signal amplification and transmitted to the TCSPC system for data recording.

[0016] Furthermore, the second light beam is coupled into a 25 um optical fiber for transmission, and white light is added as illumination light to assist the laser in focusing on the surface of the sample to be measured so as to obtain the best Raman signal.

[0017] The time-correlated single photon counting system is a coupled system that uses a microscope as the excitation light and the signal light.

[0018] The second object of the present invention can be achieved by adopting the following technical solutions: A Raman spectrum fluorescence baseline correction device, the device comprising: A signal acquisition unit, used to acquire the time domain signal of the sample to be tested and obtain the IRF function of the acquisition system of the sample to be tested; A deconvolution unit is used to preprocess the time domain signal, deconvolve the preprocessed time domain signal with the IRF function, and obtain a deconvolution result; A data fitting unit is used to fit the second half of the deconvolution result data using a support vector regression model to obtain a fitting curve; wherein a radial basis function is used as a kernel function, and the hyperparameters of the support vector regression model are optimized by grid search and cross-validation methods; A calculation unit, used for performing a difference operation between the deconvolution result and the fitting curve to obtain a Raman attenuation curve; The calculation unit is used to window the Raman attenuation curve and take the average value to obtain the Raman spectrum.

[0019] The third object of the present invention can be achieved by adopting the following technical solutions: A terminal device comprises a processor and a memory for storing a program executable by the processor. When the processor executes the program stored in the memory, the above-mentioned Raman spectrum fluorescence baseline correction method is implemented.

[0020] The fourth object of the present invention can be achieved by adopting the following technical solutions: A computer-readable storage medium stores a program, and when the program is executed by a processor, the Raman spectrum fluorescence baseline correction method is implemented.

[0021] The present invention has the following beneficial effects compared with the prior art: By introducing a post-processing algorithm based on the SVR model, the present invention can maintain a high fitting accuracy under high noise conditions and obtain the Raman signal of the fluorescent sample without requiring a large amount of prior knowledge; at the same time, the present invention has high computational efficiency and is suitable for real-time processing of large-scale data. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0023] Figure 1 This is a flow chart of the Raman spectroscopy fluorescence baseline correction method according to Example 1 of the present invention.

[0024] Figure 2 This is a schematic diagram of the Raman spectroscopy fluorescence baseline correction method according to Example 1 of the present invention.

[0025] Figure 3 This is a schematic structural diagram of the time-correlated single-photon counting system of Example 1 of the present invention.

[0026] Figure 4 It is a schematic diagram comparing the fitted fluorescence lifetime curve and the original fluorescence lifetime in Example 1 of the present invention.

[0027] Figure 5 This is a comparison diagram of the effects of Example 1 of the present invention, wherein (a) is a Raman signal obtained by the method provided by the present invention, and (b) is a Raman signal obtained by a multi-exponential decay method.

[0028] Figure 6 This is a structural block diagram of a Raman spectroscopy fluorescence baseline correction device according to Example 2 of the present invention.

[0029] Figure 7This is a structural block diagram of the terminal device of embodiment 3 of the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. It should be understood that the specific embodiments described are only used to explain the present application and are not used to limit the present application.

[0031] Embodiment 1: like Figure 1 , 2 As shown, this embodiment provides a Raman spectroscopy fluorescence baseline correction method, comprising the following steps: S101, collecting the time domain signal of the sample to be tested, and obtaining the IRF function of the acquisition system of the sample to be tested.

[0032] The IRF function of the time-correlated single photon counting system is obtained, and the time-domain signal of the sample is collected using the time-correlated single photon counting system.

[0033] The time domain signal acquisition of this embodiment mainly relies on the time-correlated single photon counting system, such as Figure 3 As shown, the system consists of a laser source, an optical path module, a spectrometer, a single photon avalanche diode (SPAD) detector, and a data acquisition and processing system (TCSPC system). The optical path module includes a spectroscope 1, an optical fiber 2, an optical fiber 3, a narrow-band filter 4, a dichroic mirror 6, a bandpass filter 7, a collection optical fiber 8, and a spectrometer.

[0034] Choosing a suitable laser source is the key. Pulsed lasers are usually used. This system uses a picosecond laser with a repetition rate of 20MHz to ensure the high temporal resolution of the light source. In addition, in order to ensure the stability of the experimental conditions, the laser source must have stable output power and good beam quality.

[0035] After the pulse laser passes through the spectroscope 1, the first beam and the second beam are obtained; after the first beam passes through the photodetector, the obtained electrical signal is transmitted to the time-correlated single photon counter as a reference synchronization signal to record the collected time domain signal; the second beam passes through the optical fibers 2 and 3 and is collimated by the fiber collimator, and the narrowband filter 4 filters out the bands other than the 532nm excitation light. After passing through multiple reflectors, the laser is coupled into the microscope objective by the dichroic mirror 6 and focused on the sample to obtain the time domain signal; the time domain signal is collimated by the microscope objective and coupled to the 50um optical fiber 8, coupled to the single photon avalanche diode detector (SPAD) through the spectrometer for signal amplification and transmitted to the time-correlated single photon counter (TCSPC) for data recording. In order to reduce the energy loss of the laser in the optical path, the reflector coating selects a high reflectivity coating corresponding to the laser band. In order to ensure that the laser is accurately irradiated on the sample surface, white light is added as illumination light in the microscopic coaxial system to assist the laser in focusing on the surface of the sample to be tested to obtain the best Raman signal acquisition. The beam splitting is due to the need for the system to record accurate time-related data.

[0036] In order to ensure the size of the system and facilitate coupling with the signal collection system, the laser source is coupled into 25um optical fibers 2 and 3. The time-correlated single photon counting system is a coupling system that uses a microscope as the excitation light and the signal light.

[0037] In the optical path design, this embodiment also considers the suppression of spatial light and stray light, including using suppression materials to absorb stray light and structural closure to suppress spatial light. In addition, by using a low-noise fiber coupling system, the transmission efficiency of Raman signals can be effectively improved and the loss in the fiber connection process can be reduced.

[0038] (1) Collection of IRF (Instrument Response Function).

[0039] Accurate measurement of the instrument response function (IRF) is an integral part of the TCSPC technique, as the width and shape of the IRF directly affect the accuracy of fluorescence lifetime measurements.

[0040] The collection of IRF is usually completed by measuring the fluorescence signal of a standard sample with a known very short lifetime (e.g., picoseconds). In this embodiment, a fluorescent dye with an extremely short fluorescence lifetime is selected as a standard sample, and the same laser excitation is used, and its fluorescence decay curve is collected by the same optical path configuration as above. The shape of this curve is the IRF of the system, and by fitting this curve, the time resolution of the IRF can be accurately obtained.

[0041] (2) Acquisition of time domain signals.

[0042] After completing the optical path design and IRF acquisition, collect the device signal of the sample. In actual operation, the selection of samples should try to avoid materials with strong absorption or strong scattering to reduce the influence of background noise. After the laser excites the sample, the generated time domain signal is received by the SPAD detector and transmitted to the data acquisition system. In order to obtain high-quality fluorescence lifetime data, long-term signal acquisition is required to ensure sufficient photon counting rate. The collected time domain signal usually contains system noise, dark noise and background noise.

[0043] S102, deconvolve the preprocessed time domain signal with the IRF function to obtain a deconvolution result.

[0044] The time domain signal is mainly preprocessed to remove abnormal points.

[0045] The preprocessing in this embodiment includes noise reduction and outlier removal.

[0046] The preprocessed time domain signal is deconvolved with the collected IRF function to obtain a deconvolved time domain signal as a deconvolution result.

[0047] S103, fitting the deconvolution result using a support vector regression model to obtain a fitting curve.

[0048] The deconvolved time domain signal is input into the support vector regression (SVR) model for fitting. The selection and parameter optimization of the SVR model are the core of the entire data post-processing. In this embodiment, the radial basis function (RBF) is used as the kernel function, and the hyperparameters of the SVR model (including the penalty coefficient C and the kernel parameter γ) are optimized by grid search and cross-validation methods to ensure that the model has good fitting performance. The fitting effect is shown in Figure 2. Figure 4 .

[0049] The SVR model fits a smooth fluorescence lifetime curve that conforms to physical properties by minimizing the balance between prediction error and model complexity. Compared with the traditional multi-exponential fitting method, the SVR model can still maintain high fitting accuracy under high noise conditions and does not need to rely on a priori exponential model assumptions. The fitted fluorescence lifetime curve can be further used to extract Raman signals. The specific method is to find the characteristic peak of the Raman signal through derivative analysis of the fitting curve and calculate its intensity.

[0050] The SVR model finds an optimal regression function on the input data set so that most data points are within a certain distance (i.e., tolerance) of the function. At the same time, SVR also allows some data points to deviate from the function by introducing slack variables, but tries to minimize these deviations.

[0051] The goal of the SVR model is to find a regression function: So that: Among them, y i Represents the i-th data sample point x i The corresponding actual output, ϵ is the tolerance, ω and b are the parameters of the model.

[0052] In order to deal with unavoidable errors, slack variables are introduced and , the objective function becomes: Where n represents the total number of data sample points. Usually the value of i ranges from 1 to n.

[0053] And satisfy the constraints: This embodiment uses the kernel function Gaussian radial basis function RBF for nonlinear regression, and the regression function of the SVR model can be expressed as: Where m is the number of support vectors; and is the Lagrange multiplier, used to represent x i The weight of is the kernel function, which is used to calculate the input data sample points x and x i The choice of kernel function determines the way of mapping to high-dimensional feature space.

[0054] The Gaussian radial basis function is expressed as: Among them, γ is the parameter of the kernel function, which is used to control the width of the Gaussian function; x j represents the jth data sample point.

[0055] Since the duration of Raman is extremely short (usually hundreds of picoseconds), while the fluorescence decay usually ranges from a few nanoseconds to more than ten nanoseconds, the second half of the photon decay curve can be considered as the fluorescence decay curve. This embodiment uses the last 50% of the data as samples to fit the trajectory of the entire fluorescence background curve.

[0056] S104, performing a difference operation between the deconvolution result and the fitting curve to obtain a Raman attenuation curve.

[0057] The deconvolution result obtained in step S102 is subtracted from the fitting curve obtained in step S103 to obtain a Raman attenuation curve.

[0058] S105, adding a window to the Raman attenuation curve and taking an average value to obtain a Raman spectrum.

[0059] At the initial excitation position, the Raman attenuation curve is windowed and averaged to obtain the Raman spectrum data. By analyzing the change rate and peak position of the Raman spectrum data, the signal peak corresponding to Raman scattering is identified. Due to the good fitting performance of the SVR model, the extracted Raman signal has a high signal-to-noise ratio and can accurately reflect the molecular vibration information of the sample.

[0060] In the final stage of data processing, the extracted Raman signal will be compared with the known standard Raman spectrum to determine the composition and structure information of the sample. If the Raman signal of the sample is weak, the resolution and accuracy of the signal can be further improved by accumulating the signal of multiple acquisition cycles. The final comparison effect is as follows Figure 5 .

[0061] This embodiment uses a post-processing algorithm based on a support vector regression (SVR) model to process the photon attenuation curve collected by the time domain device to obtain a Raman spectrum. During the processing, first, the data is preprocessed to remove obvious noise and outliers; then, the preprocessed data is deconvolved with the measured IRF function, and the data of the last 50% length of the result curve is input into the SVR model, and the good generalization ability of the model is used to fit an accurate fluorescence background curve; the fitted background curve will be subtracted from the deconvolution result, and finally a high signal-to-noise ratio Raman attenuation curve will be obtained; finally, on this basis, the initial position of the excitation is windowed to obtain a high signal-to-noise ratio Raman spectrum.

[0062] Those skilled in the art will appreciate that all or part of the steps in the method for implementing the above embodiments may be completed by instructing related hardware through a program, and the corresponding program may be stored in a computer-readable storage medium.

[0063] It should be noted that although the method operations of the above embodiments are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. On the contrary, the steps depicted can be performed in a different order. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step, and / or one step can be decomposed into multiple steps.

[0064] Embodiment 2: like Figure 6 As shown, this embodiment provides a Raman spectrum fluorescence baseline correction device, which includes a signal acquisition unit 601, a deconvolution unit 602, a data fitting unit 603, an operation unit 604 and a calculation unit 605, wherein: The signal acquisition unit 601 is used to acquire the time domain signal of the sample to be tested and obtain the IRF function of the acquisition system of the sample to be tested.

[0065] The deconvolution unit 602 is used to preprocess the time domain signal and deconvolve the preprocessed time domain signal with the IRF function to obtain a deconvolution result.

[0066] The data fitting unit 603 is used to fit the second half of the deconvolution result data using the support vector regression model to obtain a fitting curve, wherein the radial basis function is used as the kernel function, and the hyperparameters of the support vector regression model are optimized by grid search and cross validation methods.

[0067] The operation unit 604 is used to perform a difference operation between the deconvolution result and the fitting curve to obtain a Raman attenuation curve.

[0068] The calculation unit 605 is used to perform windowing on the Raman attenuation curve and take an average value to obtain a Raman spectrum.

[0069] The specific implementation of each unit in this embodiment can refer to the above-mentioned embodiment 1, and will not be described one by one here. It should be noted that the device provided in this embodiment is only illustrated by the division of the above-mentioned functional units. In practical applications, the above-mentioned functions can be assigned to different functional units as needed, that is, the internal structure can be divided into different functional units to complete all or part of the functions described above.

[0070] Embodiment 3: This embodiment provides a computer device, which may be a computer, such as Figure 7 As shown, a processor 702, a memory, an input device 703, a display 704 and a network interface 705 connected via a system bus 701 are provided. The processor is used to provide computing and control capabilities. The memory includes a non-volatile storage medium 706 and an internal memory 707. The non-volatile storage medium 706 stores an operating system, a computer program and a database. The internal memory 707 provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. When the processor 702 executes the computer program stored in the memory, the Raman spectrum fluorescence baseline correction method of the above-mentioned embodiment 1 is implemented as follows: The time domain signal of the sample to be tested is collected, and the IRF function of the acquisition system of the sample to be tested is obtained.

[0071] The time domain signal is preprocessed, and the preprocessed time domain signal is deconvolved with the IRF function to obtain a deconvolution result.

[0072] The support vector regression model is used to fit the second half of the deconvolution result data to obtain the fitting curve. The radial basis function is used as the kernel function, and the hyperparameters of the support vector regression model are optimized through grid search and cross-validation methods.

[0073] The deconvolution result is subtracted from the fitting curve to obtain the Raman attenuation curve.

[0074] The Raman attenuation curve is windowed and averaged to obtain the Raman spectrum.

[0075] Embodiment 4: This embodiment provides a storage medium, which is a computer-readable storage medium, storing a computer program. When the computer program is executed by a processor, the Raman spectrum fluorescence baseline correction method of the above embodiment 1 is implemented as follows: The time domain signal of the sample to be tested is collected, and the IRF function of the acquisition system of the sample to be tested is obtained.

[0076] The time domain signal is preprocessed, and the preprocessed time domain signal is deconvolved with the IRF function to obtain a deconvolution result.

[0077] The support vector regression model is used to fit the second half of the deconvolution result data to obtain the fitting curve. The radial basis function is used as the kernel function, and the hyperparameters of the support vector regression model are optimized through grid search and cross-validation methods.

[0078] The deconvolution result is subtracted from the fitting curve to obtain the Raman attenuation curve.

[0079] The Raman attenuation curve is windowed and averaged to obtain the Raman spectrum.

[0080] It should be noted that the computer-readable storage medium of the present embodiment may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0081] In summary, the present invention proposes for the first time a data post-processing technology based on the SVR model to handle the baseline correction problem of the time-domain signal in the Raman spectroscopy device, improves the signal-to-noise ratio of the data processing of the Raman spectrum, and solves the problems of high noise sensitivity and processing complexity in the traditional method. By introducing the support vector regression (SVR) model, a powerful regression model, to fit the fluorescence background curve of the time-domain signal in the Raman spectrum, instead of using the traditional exponential decay curve fitting, the fluorescence lifetime curve fitting based on the fluorescence lifetime model is avoided, and the dependence on prior knowledge is reduced.

[0082] The above is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solution and inventive concept of the present invention within the scope disclosed by the present invention, which shall fall within the protection scope of the present invention.

Claims

1. A method for Raman spectroscopy fluorescence baseline correction, characterized in that: The method comprises: Collecting the time domain signal of the sample to be tested, and obtaining the IRF function of the acquisition system of the sample to be tested; Preprocess the time domain signal, deconvolve the preprocessed time domain signal with the IRF function to obtain a deconvolution result; The support vector regression model is used to fit the second half of the deconvolution result data to obtain a fitting curve; wherein the radial basis function is used as the kernel function, and the hyperparameters of the support vector regression model are optimized through grid search and cross-validation methods; Perform a difference operation between the deconvolution result and the fitting curve to obtain the Raman attenuation curve; The Raman attenuation curve is windowed and averaged to obtain the Raman spectrum.

2. The Raman spectroscopy fluorescence baseline correction method according to claim 1, characterized in that: The support vector regression model finds the optimal regression function on the input data set so that most data points are within the tolerance of the function; at the same time, slack variables are introduced to handle inevitable errors.

3. The Raman spectroscopy fluorescence baseline correction method according to claim 2, characterized in that: Assume that the introduced slack variable is and , then the objective function is: And satisfy the constraints: Where ω and b are the parameters of the model; C is the regularization parameter, which controls the trade-off between the complexity and error of the model; y i is the i-th data sample point x i The corresponding actual output, n is the total number of data sample points, and ϵ is the tolerance.

4. The Raman spectroscopy fluorescence baseline correction method according to claim 2, characterized in that: The regression function of the support vector regression model is: Among them, the Gaussian radial basis function is: Where m is the number of support vectors, x i is the i-th data sample point; and is the Lagrange multiplier, representing x i The weight of is a kernel function used to calculate the input data sample points x and x i The kernel function determines the mapping method to the high-dimensional feature space; b is the parameter of the model; γ is the parameter of the kernel function, which is used to control the width of the Gaussian function; x j is the jth data sample point.

5. The Raman spectroscopy fluorescence baseline correction method according to any one of claims 1 to 4, characterized in that: The preprocessing includes noise reduction and outlier removal.

6. The Raman spectroscopy fluorescence baseline correction method according to any one of claims 1 to 4, characterized in that: The sample acquisition system to be tested is a time-correlated single-photon counting system, which consists of a pulsed laser, an optical path module, a photodetector, a single-photon avalanche diode detector and a TCSPC system. The pulsed laser is divided into a first light beam and a second light beam through the optical path module; after the first light beam passes through the photodetector, the obtained electrical signal is transmitted to the TCSPC system as a reference synchronization signal to record the collected time domain signal; the second light beam passes through the optical path module to couple the laser into the microscope objective and focus it on the sample to be tested to obtain a time domain signal; the time domain signal is coupled to the optical path module after being collimated by the microscope objective, and is coupled to the single-photon avalanche diode detector through a spectrometer for signal amplification and transmission to the TCSPC system for data recording.

7. The Raman spectrum fluorescence baseline correction method according to claim 6, characterized in that: The second light beam is coupled into a 25um optical fiber for transmission, and white light is added as illumination light to assist the laser in focusing on the surface of the sample to be tested in order to obtain the best Raman signal. The time-correlated single photon counting system is a coupled system that uses a microscope as the excitation light and the signal light.

8. A Raman spectrum fluorescence baseline correction device, characterized in that: The device comprises: A signal acquisition unit, used to acquire the time domain signal of the sample to be tested and obtain the IRF function of the acquisition system of the sample to be tested; A deconvolution unit is used to preprocess the time domain signal, deconvolve the preprocessed time domain signal with the IRF function, and obtain a deconvolution result; A data fitting unit is used to fit the second half of the deconvolution result data using a support vector regression model to obtain a fitting curve; wherein a radial basis function is used as a kernel function, and the hyperparameters of the support vector regression model are optimized by grid search and cross-validation methods; A calculation unit, used for performing a difference operation between the deconvolution result and the fitting curve to obtain a Raman attenuation curve; The calculation unit is used to window the Raman attenuation curve and take the average value to obtain the Raman spectrum.

9. A computer device comprising a processor and a memory for storing a program executable by the processor, characterized in that: When the processor executes the program stored in the memory, the Raman spectroscopy fluorescence baseline correction method according to any one of claims 1 to 7 is implemented.

10. A storage medium storing a program, characterized in that: When the program is executed by a processor, the Raman spectroscopy fluorescence baseline correction method according to any one of claims 1 to 7 is implemented.

Images