Feature classification method, system and equipment for electroencephalogram signals and medium

Abstract

The invention discloses an electroencephalogram signal feature classification method, system and device and a medium, and the method comprises the following steps: obtaining an original electroencephalogram signal, and decomposing the original electroencephalogram signal into a plurality of sub-signals with different center frequencies by adopting an SVMD strategy based on energy evaluation; adopting a GASP-based image conversion strategy to convert the plurality of sub-signals into a group of two-dimensional picture data retaining periodic and local features of the signals; performing feature extraction classification on the two-dimensional picture data by adopting a CNN model to obtain a classification result about the original electroencephalogram signals; according to the method, a complex electroencephalogram signal can be decomposed into components with different frequency characteristics at different center frequencies by using an SVMD adaptive signal decomposition method, time sequence data is mapped to a two-dimensional image through a GASF method, and the classification precision and the classification efficiency are improved and the pertinence is enhanced by combining with the CNN automatic feature extraction capability in a deep learning technology.

Description

Technical Field

[0001] The present invention relates to the field of neural signal processing technology, and in particular to a method, system, device and medium for feature classification of electroencephalogram (EEG) signals. Background Art

[0002] In the field of electroencephalogram (EEG) signal processing and analysis, signal decomposition and feature extraction are key steps to achieve efficient classification and recognition. Traditional methods often use the empirical mode decomposition (EMD) algorithm to decompose EEG signals to obtain a series of intrinsic mode functions (IMFs). These IMFs can reflect the characteristics of the signal at different time scales and provide a rich information basis for subsequent feature classification. However, the EMD algorithm has a significant aliasing effect in the decomposition process, that is, there are overlapping parts between different IMFs, which may lead to overfitting in feature classification, thereby affecting the accuracy and stability of the classification results.

[0003] In addition, EEG signals have a multi-channel effect, that is, EEG signals of different channels are correlated in time and space, and together reflect the complex patterns of brain activity. In order to make full use of this feature, it is usually necessary to increase the EEG signal data from two dimensions (time-amplitude) to three dimensions (time-amplitude-channel) in order to more comprehensively capture the characteristic information in the signal. On this basis, combined with the existing convolutional neural network (CNN) model for classification, efficient recognition of EEG signals can be achieved. However, since EMD decomposition does not depend on frequency, and different frequency bands of EEG signals have significant distinguishing significance for feature classification, directly using IMFs after EMD decomposition for CNN classification may not fully utilize all the characteristic information of EEG signals.

[0004] To address the above issues, another common approach is to use neural networks for time series signals, such as time series convolutional networks (TCNs) or recurrent neural networks (RNNs). These network structures can directly process the time series characteristics of EEG signals without the need for dimensionality increase in advance, thereby simplifying the data processing process. However, time series networks also face some challenges when processing EEG signals. First, EEG signals usually contain a lot of noise and redundant information, and time series networks perform better when processing signals with smaller data volumes and relatively purer data, but when processing EEG signals with more noise and larger signal volumes, it is often difficult to obtain accurate results. Secondly, time series networks have high requirements for parameters, and different parameter settings have a greater impact on the results, and are not suitable for transfer learning using pre-trained models. In addition, the universality of time series networks is not strong, and the neural network structure needs to be adjusted on a large scale for different problems, which increases the difficulty of model design and optimization.

[0005] In summary, although the EMD algorithm combined with CNN and temporal neural network has certain potential in processing EEG signals, there are still many challenges and limitations. Therefore, how to effectively overcome these limitations and improve the accuracy and stability of EEG signal classification is an important direction of current research. Summary of the invention

[0006] The purpose of the present invention is to provide a method, system, device and medium for feature classification of EEG signals, thereby solving all or one of the above-mentioned problems existing in the prior art.

[0007] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0008] In one aspect, the present invention provides a method for feature classification of an electroencephalogram signal, comprising the following steps:

[0009] Signal decomposition steps:

[0010] Acquire an original EEG signal, and decompose the original EEG signal into a plurality of sub-signals with different central frequencies by using an SVMD strategy based on energy evaluation;

[0011] Two-dimensional image conversion steps:

[0012] Using a GASP-based image conversion strategy to convert the plurality of sub-signals into a set of two-dimensional image data that retains signal periodicity and local features;

[0013] Image classification steps:

[0014] The CNN model is used to perform feature extraction and classification on the two-dimensional image data to obtain classification results of the original EEG signal.

[0015] As an improved solution, the SVMD strategy based on energy evaluation includes:

[0016] Setting constraint conditions, and performing cyclic decomposition on the original EEG signal based on the constraint conditions;

[0017] In the cycle decomposition process, a filter is used to separate background noise;

[0018] A loop end condition based on an energy evaluation method is set, and the loop decomposition process is terminated based on the loop end condition to obtain a loop decomposition result composed of a plurality of the sub-signals.

[0019] As an improved solution, the setting of constraint conditions and the cyclic decomposition of the original EEG signal based on the constraint conditions include:

[0020] Setting a first constraint condition, a second constraint condition and a third constraint condition based on a center frequency condition;

[0021] The decomposition step is cyclically performed based on the first constraint condition, the second constraint condition and the third constraint condition, and the decomposition step includes:

[0022] Preliminarily decomposing the original EEG signal according to the first constraint condition, the second constraint condition and the third constraint condition to obtain a first decomposition relational expression;

[0023] According to the combination of quadratic penalty terms and Lagrange multipliers, the augmented Lagrangian function is established to obtain the second decomposition relation;

[0024] The first decomposition relational expression and the second decomposition relational expression are processed by using the Parseval identity to obtain a primary decomposition result.

[0025] As an improved solution, the loop end condition includes:

[0026]

[0027] Among them, P is the energy of the signal, P u is the energy of the signal reconstructed by SVMD, P nosie is the energy of the background noise signal separated by the filter, P r is the energy of the original signal, f(n) is the signal whose energy is to be calculated, N is the total number of sample points, and n is the current sample point position.

[0028] As an improved solution, the filter includes: a Savitzky-Golay filter;

[0029] The filter formula of the Savitzky-Golay filter is:

[0030]

[0031] m j =(A T A) -1 A T e 0 ;

[0032] Among them, f r (t) is the original signal, m j is the parameter matrix of the filter; A is a (2k+1)×(n+1) matrix, whose element A ij =j i ;e 0 is a (2k+1)×1 unit vector, the first element of which is 1 and the rest are 0.

[0033] As an improved solution, the GASP-based image conversion strategy includes:

[0034] Performing standardization processing on the signal segments of the sub-signals;

[0035] Map the normalized time series data onto the unit circle and calculate the corresponding phase angle;

[0036] Calculating a Gramian matrix based on the calculated phase angle;

[0037] Setting the Gramian matrix to the grayscale value of the two-dimensional image data;

[0038] Each element of the Gramian matrix is ​​mapped to a range of grayscale values ​​according to a value range to obtain the two-dimensional image data.

[0039] As an improved solution, the CNN model includes: Inception-V3, AlexNet, ResNet-101 and VGG-16.

[0040] On the other hand, the present invention also provides a feature classification system for EEG signals, comprising:

[0041] The signal decomposition module is used to: obtain the original EEG signal, and decompose the original EEG signal into a plurality of sub-signals with different central frequencies by using the SVMD strategy based on energy evaluation;

[0042] A two-dimensional image conversion module, used to: adopt an image conversion strategy based on GASP to convert the plurality of sub-signals into a set of two-dimensional image data that retains the periodicity and local features of the signal;

[0043] The image classification module is used to: use the CNN model to perform feature extraction and classification on the two-dimensional image data to obtain classification results for the original EEG signal.

[0044] On the other hand, the present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the feature classification method of the electroencephalogram signal are implemented.

[0045] On the other hand, the present invention further provides a computer device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; wherein:

[0046] The memory is used to store computer programs;

[0047] The processor is used to execute the steps of the feature classification method of the electroencephalogram signal by running the program stored in the memory.

[0048] The beneficial effects of the technical solution of the present invention are:

[0049] 1. The feature classification method of EEG signals described in the present invention can utilize the SVMD adaptive signal decomposition method to decompose complex EEG signals into components with different frequency characteristics at different center frequencies, effectively retaining their frequency characteristics, and setting the SVMD termination condition by adding an energy evaluation part, which not only speeds up the algorithm speed, but also filters out high-frequency noise and improves the classification accuracy.

[0050] 2. The feature classification method of EEG signals described in the present invention can map time series data to two-dimensional images through the GASF method, retaining its local, global patterns and periodic characteristics. After converting one-dimensional data into two-dimensional images, the method uses a pre-trained CNN model for classification processing, which significantly improves the classification speed, and different pre-trained models can be selected according to specific problems to enhance pertinence.

[0051] 3. The feature classification method of EEG signals described in the present invention can be combined with the CNN automatic feature extraction capability in deep learning technology and the signal processing technology of SVMD and GASF to more accurately identify and classify EEG signal features, significantly improve the accuracy of EEG feature classification, and solve the problems of poor universality and poor accuracy of existing EEG signal classification models.

[0052] 4. The feature classification system of the EEG signal described in the present invention can realize the feature classification method of the EEG signal described in the present invention through the mutual cooperation of system modules.

[0053] 5. The computer-readable storage medium described in the present invention can realize the cooperation of the guidance system module, thereby realizing the characteristic classification method of the EEG signal described in the present invention, and the computer-readable storage medium described in the present invention also effectively improves the operability of the characteristic classification method of the EEG signal.

[0054] 6. The computer device described in the present invention can store and execute the computer-readable storage medium, thereby realizing the feature classification method of the EEG signal described in the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 is a schematic flow chart of the feature classification method of EEG signals described in Example 1 of the present invention;

[0057] Figure 2 Schematic diagram of the logic architecture of the feature classification method of EEG signals described in Example 1 of the present invention;

[0058] Figure 3 is a schematic diagram of the architecture of the feature classification system for EEG signals according to Embodiment 2 of the present invention;

[0059] Figure 4 is a schematic diagram of the structure of the computer device described in Example 4 of the present invention;

[0060] The symbols in the accompanying drawings are explained as follows:

[0061] 1501, processor; 1502, communication interface; 1503, memory; 1504, communication bus. DETAILED DESCRIPTION

[0062] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0063] In the description of the present invention, it should be noted that the embodiments described in the present invention are only part of the embodiments of the present invention, rather than all of the embodiments; based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.

[0064] The terms "first", "second", etc. in the specification and claims of this article and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of this article described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0065] In the description of the present invention, it should be noted that: SVMD is successive variational mode decomposition; GASF is Gram angular sum field; CNN is convolutional neural network; EMD is empirical mode decomposition.

[0066] Example 1

[0067] This embodiment provides a feature classification method for EEG signals, such as Figure 1 and Figure 2 As shown, the following steps are included:

[0068] S100, a signal decomposition step based on SVMD, including:

[0069] S101, setting the constraint condition of signal decomposition according to the central frequency condition when decomposing the signal 1 ,J 2 ,J 3 , as follows:

[0070]

[0071] In the above formula, β i (t) is the impulse response of the filter, is the filter, The frequency is ω L A sine wave; For u L The derivative of (t), for u L The Hilbert transform of (t), To find the derivative of the formula with respect to t.

[0072] S102, decomposing the original EEG signal f(t) according to the constraint conditions, as follows:

[0073] f(t)=u L (t)+f u (t)+∑ i＝1:L-1 ui (t);

[0074] min{αJ 1 +J 2 +J 3}.

[0075] S103, based on the combination of the quadratic penalty term and the Lagrangian multiplier, an augmented Lagrangian function is established, as follows:

[0076]

[0077] S104. After processing the above formula using the Parseval identity, we finally get:

[0078]

[0079] In the above formula, λ is the Lagrange multiplier, is the Fourier transform of the original signal, is the decomposed signal of the next iteration in the frequency domain, is the Lagrange multiplier for the next iteration.

[0080] S105, loop through (2) to (4), and finally decompose the EEG signal into multiple signal components u i (t) and the remaining part u i (t), as follows:

[0081] f r (t)=∑ i＝1:L-1 u i (t)+f u (t);

[0082] In the above formula, f r (t) is the original signal, u i (t) is the decomposed signal of different frequency levels, f u (t) is the signal component that has not been decomposed, L is the number of decomposed signals, ω is the center frequency to be calculated, ω L is the center frequency of the previous classification, α is a custom parameter, r represents that the signal is the original signal, and i represents the serial number of the signal component.

[0083] S106. Since EEG signals often carry more high-frequency components and stronger background noise, in order to exclude the above-mentioned influencing factors from the reconstructed signal, the end condition of the algorithm loop is set based on the energy evaluation method in this step. The high-frequency components in the signal are restricted during signal reconstruction. The loop is terminated when the sum of the energy of the reconstructed signal and the background noise energy is approximately equal to the energy of the original signal. At the same time, a filter is used to separate the background noise, as follows:

[0084]

[0085] In the above formula, P is the energy of the signal, P u is the energy of the signal reconstructed by SVMD, P nosie is the energy of the background noise signal separated by the filter, P r is the energy of the original signal; f(n) is the signal whose energy needs to be calculated, N is the total number of sample points, and n is the current sample point position.

[0086] S107, using Savitzky-Golay filter to separate background noise. Savitzky-Golay filter can arbitrarily select different window widths at any position on the same curve to meet the needs of different smoothing filters, as follows:

[0087]

[0088] m j =(A T A) -1 A T e 0 ;

[0089] In the above formula, m j is the parameter matrix of the filter; A is a (2k+1)×(n+1) matrix, whose element A ij =j i ;e 0 is a (2k+1)×1 unit vector, the first element of which is 1 and the rest are 0.

[0090] S108. Finally, the original single-channel one-dimensional EEG signal is decomposed into signal components with different center frequencies, as follows:

[0091] f r (t) = f(t) + f noise (t);

[0092]

[0093] In the above formula, f noise (t) is the noise signal, k is the number of component signals, and j is the sequence number of the component signal.

[0094] S200, two-dimensional image conversion steps based on GASP, specifically:

[0095] S201, performing standardization processing on the decomposed single signal segment, as follows:

[0096]

[0097] In the above formula, μ is the mean of the signal and σ is the standard deviation of the signal.

[0098] S202, mapping the normalized time series data onto a unit circle, and calculating the phase angle, the calculation formula is as follows:

[0099]

[0100] S203, calculating the Gramian matrix based on the calculated phase angle, the element calculation formula of the Gramian matrix is ​​as follows:

[0101] G i,j =cos(θ i +θ j ).

[0102] S204, setting the gray value of the two-dimensional image to the Gramian matrix obtained by the above calculation, and mapping each element of the Gramian matrix to a range of gray values ​​according to a certain value range, as follows:

[0103]

[0104] In the above formula, U is the component signal after SVMD decomposition, θ i and θ j are the phase angles corresponding to the timing signal on the polar coordinates, θ i is the angle of the signal at point i, θ j is the angle of the signal at point j; I(x,y) is the value at point (x,y) after grayscale processing, max(G) is the maximum value of the G matrix, min(G) is the minimum value of the G matrix, and G(x,y) is the value of point (x,y) in the G matrix; finally, based on the GASP method, each decomposed signal with different center frequencies is converted into a set of two-dimensional images that retain the periodicity and local features of the signal, supporting subsequent feature extraction and classification.

[0105] S300, image classification step based on CNN, specifically includes:

[0106] S301, using multiple CNN models to perform classification processing on the two-dimensional image data based on feature extraction, as follows:

[0107] (1) Use a CNN model that is suitable for the classification problem and data volume, or use a pre-trained deep model to reduce the time spent on retraining the model, improve operational efficiency, and ensure a certain classification effect;

[0108] (2) Conventional pre-trained deep CNN networks include but are not limited to: Inception-V3, AlexNet, ResNet-101 and VGG-16, etc.

[0109] It should be noted that the above examples are only for explaining the present invention and cannot limit the protection scope of the present invention.

[0110] Example 2

[0111] This embodiment is based on the same inventive concept as the method for characteristic classification of EEG signals described in Embodiment 1, and provides a characteristic classification system for EEG signals, such as Figure 3 As shown, including:

[0112] The signal decomposition module is used to: obtain the original EEG signal, and decompose the original EEG signal into a plurality of sub-signals with different central frequencies by using the SVMD strategy based on energy evaluation;

[0113] A two-dimensional image conversion module, used to: adopt an image conversion strategy based on GASP to convert the plurality of sub-signals into a set of two-dimensional image data that retains the periodicity and local features of the signal;

[0114] The image classification module is used to: use the CNN model to perform feature extraction and classification on the two-dimensional image data to obtain classification results for the original EEG signal.

[0115] Example 3

[0116] This embodiment provides a computer-readable storage medium, including:

[0117] The storage medium is used to store computer software instructions used to implement the feature classification method for EEG signals described in the above-mentioned embodiment 1, which includes a program for executing the above-mentioned program set for the feature classification method for EEG signals; specifically, the executable program can be built into the feature classification system for EEG signals described in embodiment 2, so that the feature classification system for EEG signals can implement the feature classification method for EEG signals described in embodiment 1 by executing the built-in executable program.

[0118] In addition, the computer-readable storage medium of this embodiment may adopt any combination of one or more computer-readable storage media, wherein the computer-readable storage medium includes electrical, optical, electromagnetic, infrared or semiconductor systems, devices or components, or any combination thereof.

[0119] Example 4

[0120] This embodiment provides an electronic device, such as Figure 4As shown, the electronic device may include: a processor 1501 , a communication interface 1502 , a memory 1503 and a communication bus 1504 , wherein the processor 1501 , the communication interface 1502 , and the memory 1503 communicate with each other via the communication bus 1504 .

[0121] Memory 1503, used for storing computer programs;

[0122] The processor 1501 is used to implement the steps of the feature classification method of the EEG signal described in the above-mentioned embodiment 1 when executing the computer program stored in the memory 1503.

[0123] As an embodiment of the present invention, the communication bus mentioned in the above terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0124] As an implementation mode of the present invention, the communication interface is used for communication between the above-mentioned terminal and other devices.

[0125] As an embodiment of the present invention, the memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0126] As an embodiment of the present invention, the above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0127] Different from the prior art, the feature classification method, system, device and medium of EEG signals of the present application can utilize the SVMD adaptive signal decomposition method to decompose complex EEG signals into components with different frequency characteristics at different center frequencies, map the time series data to a two-dimensional image through the GASF method, and combine the CNN automatic feature extraction capability in deep learning technology to more accurately identify and classify EEG signal features, thereby improving classification accuracy and efficiency and enhancing targetedness.

[0128] It should be understood that in the various embodiments of this document, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this document.

[0129] It should also be understood that in the embodiments of this article, the term "and / or" is only a description of the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0130] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this article.

[0131] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0132] In the several embodiments provided herein, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or can be electrical, mechanical or other forms of connection.

[0133] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of this article.

[0134] In addition, each functional unit in each embodiment of this invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional unit.

[0135] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this article is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of this article. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0136] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A feature classification method for EEG signals, characterized in that: The following steps are involved: Signal decomposition steps: Acquire an original EEG signal, and decompose the original EEG signal into a plurality of sub-signals with different central frequencies by using an SVMD strategy based on energy evaluation; Two-dimensional image conversion steps: Using a GASP-based image conversion strategy to convert the plurality of sub-signals into a set of two-dimensional image data that retains signal periodicity and local features; Image classification steps: The CNN model is used to perform feature extraction and classification on the two-dimensional image data to obtain classification results of the original EEG signal.

2. The feature classification method of EEG signals according to claim 1, characterized in that: The SVMD strategy based on energy evaluation includes: Setting constraint conditions, and performing cyclic decomposition on the original EEG signal based on the constraint conditions; In the cycle decomposition process, a filter is used to separate background noise; A loop end condition based on an energy evaluation method is set, and the loop decomposition process is terminated based on the loop end condition to obtain a loop decomposition result composed of a plurality of the sub-signals.

3. The feature classification method of EEG signals according to claim 2, characterized in that: The setting of constraint conditions and performing cyclic decomposition on the original EEG signal based on the constraint conditions include: Setting a first constraint condition, a second constraint condition and a third constraint condition based on a center frequency condition; The decomposition step is cyclically performed based on the first constraint condition, the second constraint condition and the third constraint condition, and the decomposition step includes: Preliminarily decomposing the original EEG signal according to the first constraint condition, the second constraint condition and the third constraint condition to obtain a first decomposition relational expression; According to the combination of quadratic penalty terms and Lagrangian multipliers, the augmented Lagrangian function is established to obtain the second decomposition relation; The first decomposition relational expression and the second decomposition relational expression are processed by using the Parseval identity to obtain a primary decomposition result.

4. The feature classification method of EEG signals according to claim 2, characterized in that: The cycle end condition includes: Among them, P is the energy of the signal, P u is the energy of the signal reconstructed by SVMD, P nosie is the energy of the background noise signal separated by the filter, P r is the energy of the original signal, f(n) is the signal whose energy is to be calculated, N is the total number of sample points, and n is the current sample point position.

5. The feature classification method of EEG signals according to claim 2, characterized in that: The filter comprises: a Savitzky-Golay filter; The filter formula of the Savitzky-Golay filter is: m j =(A T AND) -1 AND T e0; Among them, f r (t) is the original signal, m j is the parameter matrix of the filter; A is a (2k+1)×(n+1) matrix, whose element A ij =j i ; e0 is a (2k+1)×1 unit vector, the first element is 1 and the rest are 0.

6. The feature classification method of EEG signals according to claim 1, characterized in that: The GASP-based image conversion strategy includes: Performing standardization processing on the signal segments of the sub-signals; Map the normalized time series data onto the unit circle and calculate the corresponding phase angle; Calculating a Gramian matrix based on the calculated phase angle; Setting the Gramian matrix to the grayscale value of the two-dimensional image data; Each element of the Gramian matrix is ​​mapped to a range of grayscale values ​​according to a value range to obtain the two-dimensional image data.

7. The feature classification method of EEG signals according to claim 1, characterized in that: The CNN models include: Inception-V3, AlexNet, ResNet-101 and VGG-16.

8. A feature classification system for EEG signals, characterized in that: include: The signal decomposition module is used to: obtain the original EEG signal, and decompose the original EEG signal into a plurality of sub-signals with different central frequencies by using the SVMD strategy based on energy evaluation; A two-dimensional image conversion module, used to: adopt an image conversion strategy based on GASP to convert the plurality of sub-signals into a set of two-dimensional image data that retains the periodicity and local features of the signal; The image classification module is used to: use the CNN model to perform feature extraction and classification on the two-dimensional image data to obtain classification results for the original EEG signal.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the feature classification method of EEG signals described in any one of claims 1 to 7 are implemented.

10. A computer device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; wherein: The memory is used to store computer programs; The processor is used to execute the steps of the feature classification method of EEG signals according to any one of claims 1 to 7 by running the program stored in the memory.

Images