Method and system for spectrogram analysis

Abstract

The invention discloses a method and a system for spectrogram analysis. A method for spectrogram analysis includes: acquiring an electron paramagnetic resonance (EPR) spectrogram signal, a category corresponding to the EPR spectrogram signal, and a microwave frequency utilized to generate the EPR spectrogram signal; the EPR spectrogram signal is converted into a two-dimensional image signal; and generating an interval value of a fitting parameter based on the two-dimensional image signal and the microwave frequency using a deep learning prediction module corresponding to the category.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and more specifically, to a method and system for spectrogram analysis. Background Art

[0002] Electron paramagnetic resonance (EPR) is a magnetic resonance technique derived from the magnetic moment of unpaired electrons. It can be used to qualitatively and quantitatively detect unpaired electrons in atoms or molecules and explore the structural properties of their surroundings. For free radicals, orbital magnetic moments play little role; the majority (over 99%) of the total magnetic moment comes from electron spin. Therefore, electron paramagnetic resonance is also called electron spin resonance (ESR). EPR is primarily used to detect free radicals and paramagnetic metal ions and their compounds, obtaining structural and compositional information. It has wide applications in materials science, chemistry, physics, semiconductors, biology, food, health protection, environmental protection, and pharmaceuticals. Analysis of EPR spectra allows for the identification of species. However, existing EPR spectrum analysis methods rely heavily on arduous manual labor and data fitting. This not only requires the analyst to possess extensive expertise to accurately determine species identity, but also suffers from low efficiency and limited accuracy due to human factors, making it difficult to meet the demands of efficient and precise scientific research. Therefore, a fast and accurate spectrum analysis method and system are urgently needed. Summary of the Invention

[0003] It is to be understood that both the foregoing general description and the following detailed description of the present invention are exemplary and explanatory and are intended to provide further explanation of the invention as claimed.

[0004] According to a first aspect of the present invention, a method for spectrum analysis is provided, comprising: acquiring an electron paramagnetic resonance (EPR) spectrum signal, a category corresponding to the EPR spectrum signal, and a microwave frequency utilized to generate the EPR spectrum signal; converting the EPR spectrum signal into a two-dimensional image signal; and generating interval values ​​of fitting parameters based on the two-dimensional image signal and the microwave frequency using a deep learning prediction module corresponding to the category.

[0005] The method as described in any one of the above items, wherein the category corresponding to the EPR spectrum signal comprises one of the following: spin-trapped free radicals, spin-free free radicals, metal species, spin-trapped compounds, and vacancies.

[0006] The method as described in any one of the above items, wherein the fitting parameters include one or more of the following: hyperfine coupling constant, g factor, line width, nuclear spin angular momentum quantum number.

[0007] The method as described in any one of the above items, wherein the method further comprises: performing an adjustment based on the interval value of the fitting parameter to obtain an adjusted value of the fitting parameter.

[0008] A method as described in any of the above items, wherein adjusting based on the interval value of the fitting parameter to obtain the adjusted value of the fitting parameter includes: generating multiple groups of candidate values ​​based on the interval value of the fitting parameter; generating a fitting spectrum for each group of candidate values ​​in the multiple groups of candidate values ​​to obtain multiple fitting spectra; calculating the similarity between the EPR spectrum signal and each of the multiple fitting spectra; and selecting a group of candidate values ​​corresponding to the fitting spectrum with the highest similarity as the adjusted value of the fitting parameter.

[0009] The method as described in any one of the above items, wherein the method further comprises: performing a finer-grained adjustment based on the adjusted value of the fitting parameter to obtain a fine-tuned value of the fitting parameter.

[0010] The method as described in any one of the above items, wherein the method further comprises: obtaining a capture agent type; searching a spin species database for a species corresponding to the capture agent type and the adjusted value of the fitting parameter; and outputting species information associated with the species.

[0011] The method as described in any one of the above items, wherein converting the EPR spectrum signal into a two-dimensional image signal includes: denoising the EPR spectrum signal.

[0012] A method as described in any of the above items, wherein denoising the EPR spectrum signal comprises: filtering the EPR spectrum signal using a threshold; or receiving user input to determine the position of a peak or a trough in the EPR spectrum signal.

[0013] A method as described in any of the above items, wherein the deep learning prediction module is trained by the following steps: obtaining a data set, the data set including a training EPR spectrum signal, a category corresponding to the training EPR spectrum signal, and a microwave frequency used to generate the training EPR spectrum signal as input information, and including a true value of a fitting parameter as output information; converting the training EPR spectrum signal into a training two-dimensional image signal; using a deep learning prediction module corresponding to the category, generating a prediction interval of the fitting parameter based on the training two-dimensional image signal and the microwave frequency; calculating a loss function between the prediction interval of the fitting parameter and the true value of the fitting parameter; and updating the parameters of the deep learning prediction module based on the loss function to obtain a trained deep learning prediction module.

[0014] According to a second aspect of the present invention, a system for spectrum analysis is provided, comprising an apparatus for executing any of the above methods.

[0015] According to a third aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, wherein the computer program implements any of the above methods when executed by a processor.

[0016] According to a fourth aspect of the present invention, a computer device is provided, comprising a memory and a processor, wherein a computer program that can be run on the processor is stored in the memory, wherein the processor implements any of the above methods when executing the computer program.

[0017] According to a fifth aspect of the present invention, a computer program product is provided, comprising a computer program, wherein the computer program implements any of the above methods when executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are included to provide a further understanding of the present invention, are incorporated into and constitute a part of this application, illustrate embodiments of the present invention, and together with the specification serve to explain the principles of the present invention. In the drawings:

[0019] Figure 1 is a flow chart of a method for spectrogram analysis according to one embodiment of the present invention;

[0020] Figure 2 is a schematic diagram of a process of spectrum analysis according to one embodiment of the present invention;

[0021] Figure 3 is a flow chart of a method for spectrogram analysis according to another embodiment of the present invention;

[0022] Figure 4 is a flow chart of a process for parameter adjustment according to one embodiment of the present invention;

[0023] Figure 5 is a flow chart of a method for spectrum analysis according to yet another embodiment of the present invention;

[0024] Figure 6 is a flowchart of a method for training a deep learning prediction module according to one embodiment of the present invention; and

[0025] Figure 7 is a block diagram of a system for spectrum analysis according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] Embodiments of the present invention will now be described in detail with reference to the accompanying drawings, but the invention is not limited thereto but only by the claims. In the drawings, for illustrative purposes, the dimensions of some of the elements may be exaggerated and not drawn to scale. Wherever possible, the same reference numerals will be used throughout the drawings to refer to the same or similar parts.

[0027] Although the terms used in the present invention are selected from commonly known and commonly used terms, some of the terms mentioned in the present invention specification may be selected by the applicant at his or her discretion, and their detailed meanings are explained in the relevant parts of the description herein. In addition, it is required to understand the present invention not only by the actual terms used, but also by the meaning implied by each term.

[0028] In the description provided herein, numerous specific details are set forth. However, it should be understood that embodiments of the present invention may be practiced without these specific details. In other instances, well-known methods, structures, and techniques are not shown in detail to avoid obscuring an understanding of the present invention.

[0029] Existing EPR spectrum analysis methods rely heavily on the deep knowledge and extensive theoretical experience of experts. Manual analysis is also extremely time-consuming, particularly for complex mixture spectra, which can take hours or even days to complete, resulting in low analytical efficiency. Furthermore, the manual nature of the analysis process limits the accuracy of the results, making it difficult to meet the demands of high-precision research.

[0030] In view of the above problems, the present application proposes a method and system for spectrum analysis, which is used to realize automated, rapid and accurate spectrum analysis. The scheme of the present application can be extended to electron paramagnetic resonance spectroscopy analysis of different bands, including pulsed electron paramagnetic resonance spectroscopy, transient electron paramagnetic resonance spectroscopy, L-band electron paramagnetic resonance spectroscopy, S-band electron paramagnetic resonance spectroscopy, X-band electron paramagnetic resonance spectroscopy, Q-band electron paramagnetic resonance spectroscopy and K-band electron paramagnetic resonance spectroscopy, electron nuclear double resonance spectroscopy (ENDR), double electron electron resonance spectroscopy (DEER), etc.

[0031] Figure 1 is a flow chart of a method 100 for spectrum analysis according to one embodiment of the present invention. Figure 2 4 is a flow chart of spectrum analysis according to an embodiment of the present invention.

[0032] At step 102, the electron paramagnetic resonance (EPR) spectrum signal, the category corresponding to the EPR spectrum signal, and the microwave frequency used to generate the EPR spectrum signal can be obtained. The EPR spectrum signal can be divided into categories such as spin-trapped free radicals, non-spin-trapped free radicals, metal species, spin-trapped mixtures, and vacancies. Therefore, the user can determine the category corresponding to the EPR spectrum based on the EPR spectrum. The microwave frequency is the frequency of the microwave radiation generated by the EPR instrument in the process of obtaining the EPR spectrum. Figure 2 As shown, an EPR spectrum signal can be obtained, the signal type of which is a spin-trapped free radical, and the corresponding microwave frequency can be obtained.

[0033] At step 104, the EPR spectrum signal can be converted into a two-dimensional image signal. The EPR spectrum signal is a one-dimensional signal, and it is converted into a two-dimensional image signal to facilitate deep learning model processing. Signal conversion can be achieved by using methods such as Gram angle field, Markov transition field, recursive graph, etc. Figure 2 As shown, the EPR spectrum signal can be converted into a Gram angle sum field characteristic map as a two-dimensional image signal. The EPR spectrum signal can also be converted into a Gram angle difference field characteristic map as a two-dimensional image signal.

[0034] At step 106, a deep learning prediction module corresponding to the category can be used to generate interval values ​​for the fitting parameters based on the two-dimensional image signal and the microwave frequency. For different categories of the spectrogram, predictions can be made using the corresponding deep learning prediction module. The deep learning prediction module can utilize a deep learning algorithm, such as a convolutional neural network, a visual transformer, a densely connected convolutional network, or the like.

[0035] In one embodiment, the fitting parameters may include one or more of the following: hyperfine coupling constant, g-factor, linewidth, and nuclear spin angular momentum quantum number. The hyperfine coupling constant, also known as the A value, describes the interaction between unpaired electrons and nuclear spins. The g-factor, also known as the g-value, describes the interaction between electron spins and external magnetic fields and is used to determine the magnetic properties of the sample. Linewidth refers to the width of the EPR spectral line.

[0036] For spin-trapped radicals, the predicted fitting parameters include hyperfine coupling constants and line widths. Hyperfine coupling constants include AN, AH, A0.5, AN2, AH3, and AH4.

[0037] For spin-free trapped radicals, the predicted fitting parameters include hyperfine coupling constants and line widths. Hyperfine coupling constants include AN, AH, A0.5, AN2, AH3, and AH4.

[0038] For metallic species, the predicted fitting parameters include hyperfine coupling constants (A1, A2, and A3), g-factors (g1, g2, and g3), linewidth, and nuclear spin angular momentum quantum numbers.

[0039] For spin-trapping mixtures, the predicted fitting parameters include the hyperfine coupling constant and linewidth for each species. Hyperfine coupling constants include AN, AH, A0.5, AN2, AH3, and AH4. In one embodiment, after the deep learning prediction module generates fitting parameters for each species, linear regression can be used to predict the proportion of each species.

[0040] For vacancies, the predicted fitting parameters include g-factor and line width. The g-factor includes g1, g2, and g3.

[0041] The deep learning prediction module can divide the potential range of each fitting parameter into multiple intervals as multiple classes. During the prediction process of the deep learning prediction module, each fitting parameter can be classified to predict that the fitting parameter belongs to one of the multiple classes, and thus the predicted class is used as the interval value of the fitting parameter. For example, the numerical range of 0-1 can be divided into 10 intervals, including 0-0.1, 0.1-0.2, etc., as 10 classes, and then the fitting parameter is predicted to belong to one of the 10 classes, and the interval corresponding to the class is used as the interval value of the fitting parameter. It should be understood that the number of intervals and the range of intervals for different fitting parameters may be different.

[0042] In one embodiment, based on the interval values ​​of the fitting parameters of the EPR spectrum signal, the user can determine the species information based on experience.

[0043] In one embodiment, converting the EPR spectrum signal into a two-dimensional image signal may include denoising the EPR spectrum signal. In one embodiment, the EPR spectrum signal may be filtered using a threshold. For example, signal values ​​whose absolute values ​​are lower than a threshold may be filtered out. The threshold may be a default setting or may be selected by the user as needed. In one embodiment, user input may be received to determine the location of a peak or trough in the EPR spectrum signal. For example, Figure 2 As shown, the user can click on the position of the peak or trough in the spectrum signal through the input device to help determine the exact position of the peak or trough through human-computer interaction.

[0044] Therefore, by using the deep learning prediction module to perform automated spectral analysis, the spectral signal can be analyzed quickly and accurately, greatly shortening the data processing time while ensuring the accuracy of the analysis results.

[0045] Figure 3FIG. 3 is a flow chart of a method 300 for spectrum analysis according to another embodiment of the present invention. Steps 302 to 306 are similar to those in the reference Figure 1 Steps 102 to 106 are the same as those in the method 100. To avoid redundancy, the specific details of steps 302 to 306 are not described here.

[0046] After completing step 306, at step 308, adjustments may be made based on the interval values ​​of the fitting parameters to obtain adjusted values ​​of the fitting parameters. An optimization method such as a multi-level optimization algorithm, a genetic algorithm, or a particle swarm optimization algorithm may be used to make the fitting spectrum generated based on the adjusted values ​​of the fitting parameters more similar to the original spectrum signal.

[0047] Figure 4 FIG. 4 is a flow chart of a process 400 for parameter adjustment according to one embodiment of the present invention. The process 400 may be Figure 3 The specific implementation of step 308 in the method 300 is described in detail, but the scope of the present invention is not limited thereto.

[0048] At step 402, multiple sets of candidate values ​​can be generated based on the interval values ​​of the fitting parameters. For example, using spin-trapped free radicals as an example, for each of the fitting parameters AN, AH, A0.5, AN2, AH3, AH4, and linewidth, an expansion can be performed based on the predicted interval values. For example, 10 candidate values ​​from the predicted interval value and / or adjacent intervals of AN are selected, 10 candidate values ​​from the predicted interval value and / or adjacent intervals of AH are selected, and so on, thereby obtaining 10 candidate values ​​for each fitting parameter. The minimum candidate value of each fitting parameter collectively forms the first set of candidate values, the second minimum candidate value of each fitting parameter collectively forms the second set of candidate values, and so on, until the maximum candidate value of each fitting parameter collectively forms the tenth set of candidate values, thereby obtaining ten sets of candidate values ​​for the fitting parameters. It should be understood that the 10 candidate values ​​selected above are merely examples, and any number of values ​​can be selected as needed.

[0049] At step 404, a fitted spectrum can be generated for each of the multiple sets of candidate values ​​to obtain multiple fitted spectra. A spectrum fitting tool such as Easyspin can be used to draw the fitted spectra based on the candidate values ​​of the fitting parameters. For example, a first fitted spectrum can be generated based on the first set of candidate values, a second fitted spectrum can be generated based on the second set of candidate values, and so on, thereby generating 10 fitted spectra.

[0050] At step 406, the similarity between the EPR spectrum signal and each of the plurality of fitted spectra can be calculated. For example, the number, location, and size of peaks / troughs of the original EPR spectrum signal and the fitted spectra can be compared, or a method such as dynamic time warping (DTW) can be used to calculate the similarity between the original EPR spectrum signal and the fitted spectra. Thus, the similarity between the original EPR spectrum signal and each of the 10 previously generated fitted spectra can be calculated.

[0051] At step 408, a set of candidate values ​​corresponding to the fitted spectrum with the highest similarity may be selected as the adjusted values ​​of the fitting parameters. For example, if, among the ten sets of candidate values, the fitted spectrum generated using the third set of candidate values ​​is most similar to the original EPR spectrum signal, the third set of candidate values ​​may be selected as the adjusted values ​​of the fitting parameters.

[0052] return Figure 3 In one embodiment, method 300 may further include step 310. At step 310, a finer-grained adjustment may be performed based on the adjusted value of the fitting parameter to obtain a fine-tuned value of the fitting parameter. For example, assuming that the difference between the different sets of candidate values ​​selected at step 308 is 0.01, then at step 310, multiple sets of candidate values ​​with a difference less than 0.01, such as a difference of 0.002, may be selected based on the adjusted value of the fitting parameter, and the fine-tuned value of the fitting parameter may be determined based on the similarity between the original EPR spectrum signal and the fitted spectra corresponding to the multiple sets of candidate values, thereby achieving finer-grained adjustment.

[0053] Figure 5 FIG. 5 is a flow chart of a method 500 for spectrum analysis according to another embodiment of the present invention. Steps 502 to 508 are similar to those in the reference Figure 3 Steps 302 to 308 are the same as those in the method 300. To avoid redundancy, the specific details of steps 502 to 508 are not described here.

[0054] After completing step 508, at step 510, the capture agent type can be obtained. The user can enter the capture agent type used during the experiment to obtain the EPR spectrum. The capture agent type can include DMPO (5,5-dimethyl-1-pyrroline-N-oxide), PBN (N-tert-butyl-α-phenylnitrone), MNP (2-methyl-2-nitrosopropane), DBNBS (3,5-dibromo-4-nitrosobenzenesulfonic acid sodium salt), POBN (α-(4-pyridyl-1-oxy)-N-tert-butyl nitroketone), TEMP (2,2,6,6-tetramethylpiperidinium oxide), etc.

[0055] At step 512, a species corresponding to the trapping agent type and the adjusted values ​​of the fitting parameters can be searched in a spin species database. For example, a spin species database can be constructed, covering multiple categories such as spin-trapped free radicals, non-spin-trapped free radicals, metal species, spin-trapped mixtures, vacancies, etc., which contains spin species data and related parameters, such as real test data and its parameters, as well as related trapping agents, configuration images, and corresponding literature. Alternatively, the corresponding species can be searched in an established spin species database. Therefore, the corresponding species can be searched in the spin species database based on the acquired trapping agent type and the determined fitting parameters.

[0056] At step 514, species information associated with the species may be output. The species information may include one or more of the following: species affiliation, capture agent, species configuration, relevant literature, adjusted values ​​of fitting parameters, and fitted spectra. The output species information may include information about one or more species that are closest to the adjusted values ​​of the fitting parameters.

[0057] By searching for species information in the spin species database and outputting the species information, users can quickly obtain relevant knowledge about the species and make full use of the EPR spectrum information.

[0058] Figure 6 is a flowchart of a method 600 for training a deep learning prediction module according to one embodiment of the present invention.

[0059] At step 602, a data set may be acquired. The data set may include a training EPR spectrum signal, a category corresponding to the training EPR spectrum signal, and a microwave frequency used to generate the training EPR spectrum signal as input information, and may include true values ​​of fitting parameters as output information.

[0060] At step 604 , the training EPR spectrum signal may be converted into a training two-dimensional image signal.

[0061] At step 606 , a deep learning prediction module corresponding to the category may be used to generate a prediction interval of the fitting parameter based on the training two-dimensional image signal and the microwave frequency.

[0062] At step 608 , a loss function between the prediction interval of the fitted parameter and the true value of the fitted parameter may be calculated.

[0063] At step 610 , parameters of the deep learning prediction module may be updated based on the loss function to obtain a trained deep learning prediction module.

[0064] Figure 7is a block diagram of a system 700 for spectrum analysis according to one embodiment of the present invention. System 700 may include an acquisition module 702, a signal conversion module 704, and a deep learning prediction module 706. Acquisition module 702 may be configured to acquire an electron paramagnetic resonance (EPR) spectrum signal, the category corresponding to the EPR spectrum signal, and the microwave frequency used to generate the EPR spectrum signal. Signal conversion module 704 may be configured to convert the EPR spectrum signal into a two-dimensional image signal. Deep learning prediction module 706, corresponding to the category, may be configured to generate interval values ​​for fitting parameters based on the two-dimensional image signal and the microwave frequency.

[0065] In one embodiment, the category corresponding to the EPR spectrum signal may include one of the following: spin-trapped radicals, spin-free radicals, metal species, spin-trapped compounds, and vacancies.

[0066] In one embodiment, the fitting parameters may include one or more of the following: hyperfine coupling constant, g-factor, linewidth, nuclear spin angular momentum quantum number.

[0067] In one embodiment, the signal conversion module 704 may also be configured to perform denoising on the EPR spectrum signal. In one embodiment, the signal conversion module 704 may perform denoising on the EPR spectrum signal by filtering the EPR spectrum signal using a threshold value or receiving user input to determine the location of a peak or trough in the EPR spectrum signal.

[0068] In one embodiment, the deep learning prediction module 706 can be trained by the following steps: obtaining a data set, the data set including a training EPR spectrum signal, a category corresponding to the training EPR spectrum signal, and a microwave frequency used to generate the training EPR spectrum signal as input information, and including the true value of the fitting parameter as output information; converting the training EPR spectrum signal into a training two-dimensional image signal; using a deep learning prediction module corresponding to the category to generate a prediction interval of the fitting parameter based on the training two-dimensional image signal and the microwave frequency; calculating a loss function between the prediction interval of the fitting parameter and the true value of the fitting parameter; and updating the parameters of the deep learning prediction module based on the loss function to obtain a trained deep learning prediction module.

[0069] In one embodiment, the system 700 may further include an adjustment module 708. The adjustment module 708 may be configured to perform an adjustment based on the interval value of the fitting parameter to obtain an adjusted value of the fitting parameter.

[0070] In one embodiment, the adjustment module 708 performs adjustment based on the interval value of the fitting parameter, which may include: generating multiple groups of candidate values ​​based on the interval value of the fitting parameter; generating a fitting spectrum for each group of candidate values ​​in the multiple groups of candidate values ​​to obtain multiple fitting spectra; calculating the similarity between the EPR spectrum signal and each of the multiple fitting spectra; and selecting a group of candidate values ​​corresponding to the fitting spectrum with the highest similarity as the adjusted value of the fitting parameter.

[0071] In one embodiment, the adjustment module 708 may be further configured to perform finer-grained adjustments based on the adjusted values ​​of the fitting parameters to obtain fine-tuned values ​​of the fitting parameters.

[0072] In one embodiment, system 700 may further include a search module 710 and an output module 712. The acquisition module 702 may further be configured to acquire a capture agent type. The search module 710 may be configured to search a spin species database for a species corresponding to the capture agent type and the adjusted values ​​of the fitting parameters. The output module 712 may be configured to output species information associated with the species. The species information may include one or more of the following: species identity, capture agent, species configuration, relevant literature, adjusted values ​​of the fitting parameters, and a fitted spectrum.

[0073] By building a model for analyzing electron paramagnetic resonance spectra based on artificial intelligence based on the method and system of this application, experiments have found that the model can achieve rapid analysis of spin species such as spin-trapped free radicals, non-spin-trapped free radicals, metal species, spin-trapped mixtures, and vacancies, with an accuracy of more than 99% within an error range of 0.1G. When tested with an image resolution of 224×224, the average time after testing 100 spin captures was 6 minutes, the average time after testing 100 spin-trapped mixtures was 6 minutes, the average time after testing 100 free radicals was 30 seconds, and the average time after testing 100 metals was 10 minutes, all of which achieved rapid analysis.

[0074] In summary, this application provides an artificial intelligence-based EPR spectrum analysis system and method that can achieve multi-category analysis of EPR spectra. For spin capture spectra, it can achieve rapid and accurate analysis. Addressing the complex challenges of mixture spectrum analysis, it can accurately analyze each species and their proportions in the mixture, improving analysis efficiency and accuracy. For metal species, it can capture the unique characteristic parameters of metal species spectra for precise attribution.

[0075] This application also constructs a spin species database covering various categories such as spin-trapped free radicals, non-spin-trapped free radicals, metal species, spin-trapped mixtures, vacancies, etc., realizing the universal analysis, identification and attribution of spin species, and providing a convenient resource platform for scientific research.

[0076] In addition, this application has strong generalization capabilities by converting EPR spectrum signals into two-dimensional image signals and combining them with a deep learning prediction module to extract the relevant features required for spectrum interpretation in the spectrum. It can cope with the arbitrary scaling and extended recognition requirements of electron paramagnetic resonance spectra, ensuring stable analytical performance.

[0077] Therefore, the EPR spectrum analysis system and method of the present application realize the automated and intelligent analysis of EPR spectra, lower the analysis threshold, improve the analysis efficiency and accuracy, and promote the application and development of EPR analysis technology in a wider range of fields.

[0078] The present invention may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present invention.

[0079] Computer-readable storage medium can be a tangible device that can keep and store the instructions used by the instruction execution device.Computer-readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device or any suitable combination thereof.More specific examples (non-exhaustive list) of computer-readable storage medium include: 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), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove having instructions stored thereon, and any suitable combination thereof.Computer-readable storage medium used herein is not interpreted as a transient signal itself, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagated by waveguides or other transmission media (for example, light pulses by fiber optic cables), or electrical signals transmitted by wires.

[0080] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to be stored in the computer-readable storage medium in each computing / processing device.

[0081] The computer program instructions for performing the operation of the present invention can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, Python, and conventional procedural programming languages ​​such as "C" language or similar programming languages. The computer readable program instructions can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), is personalized by utilizing the state information of the computer readable program instructions, and the electronic circuit can execute the computer readable program instructions, thereby realizing various aspects of the present invention.

[0082] Various aspects of the present invention are described herein with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present invention. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable program instructions.

[0083] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device is generated that implements the functions / actions specified in one or more blocks in the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, where these instructions cause the computer, programmable data processing device, and / or other device to operate in a specific manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0084] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0085] The flowcharts and block diagrams in the accompanying drawings show the possible implementation architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment or part of an instruction, and the module, program segment or part of the instruction contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions. It is well known to those skilled in the art that implementation by hardware, implementation by software, and implementation by a combination of software and hardware are all equivalent.

[0086] While various embodiments of the present invention have been described above, the foregoing description is intended to be illustrative, non-exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technological improvements in the marketplace, or to enable others skilled in the art to understand the embodiments disclosed herein. The scope of the present invention is defined by the appended claims.

Claims

1. A method for spectrum analysis, comprising: Acquire an electron paramagnetic resonance (EPR) spectrum signal, a category corresponding to the EPR spectrum signal, and a microwave frequency used to generate the EPR spectrum signal; Converting the EPR spectrum signal into a two-dimensional image signal; and An interval value of a fitting parameter is generated based on the two-dimensional image signal and the microwave frequency using a deep learning prediction module corresponding to the category.

2. The method of claim 1, wherein: The category corresponding to the EPR spectrum signal includes one of the following: spin-trapped free radicals, spin-free free radicals, metal species, spin-trapped mixtures, and vacancies.

3. The method of claim 1, wherein: The fitting parameters include one or more of the following: hyperfine coupling constant, g-factor, line width, nuclear spin angular momentum quantum number.

4. The method of claim 1, wherein: The method further comprises: An adjustment is performed based on the interval value of the fitting parameter to obtain an adjusted value of the fitting parameter.

5. The method of claim 4, wherein: Adjusting based on the interval value of the fitting parameter to obtain an adjusted value of the fitting parameter includes: generating a plurality of sets of candidate values ​​based on interval values ​​of the fitting parameters; Generating a fitting spectrum for each set of candidate values ​​in the plurality of sets of candidate values ​​to obtain a plurality of fitting spectrums; Calculating the similarity between the EPR spectrum signal and each of the plurality of fitted spectra; and A set of candidate values ​​corresponding to the fitted spectra with the highest similarity are selected as adjusted values ​​of the fitting parameters.

6. The method of claim 4, wherein: The method further comprises: A finer-grained adjustment is performed based on the adjusted value of the fitting parameter to obtain a fine-tuned value of the fitting parameter.

7. The method of claim 4, wherein: The method further comprises: Get the capture agent type; searching a spin species database for species corresponding to the capture agent type and the adjusted value of the fit parameter; and Output species information associated with the species.

8. The method of claim 1, wherein: Converting the EPR spectrum signal into a two-dimensional image signal comprises: The EPR spectrum signal is denoised.

9. The method of claim 8, wherein: Denoising the EPR spectrum signal includes: Filtering the EPR spectrum signal using a threshold; or User input is received to determine the location of a peak or a trough in the EPR spectrum signal.

10. The method of claim 1, wherein: The deep learning prediction module is trained by the following steps: Acquire a data set, the data set including a training EPR spectrum signal, a category corresponding to the training EPR spectrum signal, and a microwave frequency used to generate the training EPR spectrum signal as input information, and including a true value of a fitting parameter as output information; Converting the training EPR spectrum signal into a training two-dimensional image signal; generating a prediction interval of a fitting parameter based on the training two-dimensional image signal and the microwave frequency using a deep learning prediction module corresponding to the category; Calculating a loss function between a prediction interval of the fitting parameter and a true value of the fitting parameter; and The parameters of the deep learning prediction module are updated based on the loss function to obtain a trained deep learning prediction module.

11. A system for spectrogram analysis, comprising means for performing the method according to any one of claims 1 to 10.

12. A computer-readable storage medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

13. A computer device comprising a memory and a processor, wherein a computer program capable of being run on the processor is stored in the memory, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 10 is implemented.

14. A computer program product comprising a computer program, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

Images