Brain electrical signal classification method, device, electronic equipment and medium

Abstract

The disclosed embodiments disclose an EEG signal classification method, device, electronic device and medium, the method comprising: obtaining feature information of an EEG signal to be classified; obtaining a trained EEG signal classification model; using a trained first EEG signal classification model to process time domain features to obtain a first classification result, using a trained second EEG signal classification model to process frequency domain features to obtain a second classification result, and using a trained third EEG signal classification model to process topological features to obtain a third classification result; using a trained fourth EEG signal classification model to process the classification accuracy and first classification result of the trained first EEG signal classification model, the classification accuracy and second classification result of the trained second EEG signal classification model, and the classification accuracy and third classification result of the trained third EEG signal classification model to obtain a target classification result of the EEG signal to be classified.

Description

Technical Field

[0001] The present disclosure relates to the technical field of electroencephalogram (EEG) signal processing, and more specifically, to an EEG signal classification method, device, electronic device and medium. Background Art

[0002] Electroencephalogram (EEG) is a bioelectric potential generated by electrical activity inside the human brain. In related technologies, EEG signals can be classified using a trained EEG signal classification model. However, the EEG signal classification model involves fewer features, resulting in low classification accuracy. Summary of the invention

[0003] The embodiments of the present disclosure provide a method, device, electronic device and medium for classifying electroencephalogram (EEG) signals.

[0004] According to a first aspect of the present disclosure, a method for classifying an electroencephalogram signal is provided, the method comprising:

[0005] Acquire characteristic information of the EEG signal to be classified; wherein the characteristic information is used to reflect the signal type of the EEG signal to be classified, and the characteristic information includes time domain characteristics, frequency domain characteristics and topological characteristics;

[0006] Obtaining a trained EEG signal classification model; wherein the trained EEG signal classification model includes a trained first EEG signal classification model, a trained second EEG signal classification model, a trained third EEG signal classification model, and a trained fourth EEG signal classification model;

[0007] Using the trained first EEG signal classification model to process the time domain features to obtain a first classification result of the EEG signal to be classified, using the trained second EEG signal classification model to process the frequency domain features to obtain a second classification result of the EEG signal to be classified, and using the trained third EEG signal classification model to process the topological features to obtain a third classification result of the EEG signal to be classified;

[0008] The trained fourth EEG signal classification model is used to process the classification accuracy of the trained first EEG signal classification model and the first classification result, the classification accuracy of the trained second EEG signal classification model and the second classification result, and the classification accuracy of the trained third EEG signal classification model and the third classification result to obtain the target classification result of the EEG signal to be classified.

[0009] Optionally, the method further comprises:

[0010] Obtaining an initial EEG signal to be classified;

[0011] Performing signal preprocessing on the initial EEG signal to be classified to obtain the EEG signal to be classified;

[0012] The signal preprocessing at least includes: signal reorganization processing, artifact removal processing, artifact category detection processing, and signal segmentation processing.

[0013] Optionally, the time domain feature includes at least one of mean, standard deviation, kurtosis and skewness;

[0014] The frequency domain feature includes at least one of a power spectrum of theta band, a power spectrum of alpha band and a power spectrum of beta band;

[0015] The topological feature includes at least one of an average clustering coefficient, density and global efficiency.

[0016] Optionally, the feature information includes topological features, and the step of obtaining the feature information of the EEG signal to be classified includes:

[0017] The graph structure of the EEG signal to be classified is constructed by taking the signal strength corresponding to each EEG signal in the EEG signal to be classified as a node, taking the connection between any two nodes as an edge, and taking the slope of the edge as the weight corresponding to the edge;

[0018] Obtaining topological features of the EEG signal to be classified according to the graph structure of the EEG signal to be classified;

[0019] Among them, there is a node c between node a and node b, and nodes a, node b and node c satisfy: x b -x a <T min In the case of , it is determined that there is an edge between node a and node b, where (x a ,y a ) a is the time when the EEG signal corresponding to node a is in the EEG signal to be classified, y a is the signal strength of the EEG signal, (x b ,y b ) b is the time when the EEG signal corresponding to node b is in the EEG signal to be classified, y b is the signal strength of the EEG signal, (x c ,y c ) c is the time when the EEG signal corresponding to node c is in the EEG signal to be classified, y c is the signal strength of the EEG signal, T min The preset time window.

[0020] Optionally, obtaining a trained EEG signal classification model includes:

[0021] Acquire a first sample set; wherein each sample in the first sample set includes a sample EEG signal and label information of the sample EEG signal;

[0022] Performing signal preprocessing on each of the samples in the first sample set to obtain a second sample set; wherein the second sample set includes a test sample set and a training sample set;

[0023] Using the training sample set, the current round of training is performed on the basic first EEG signal classification model, the basic second EEG signal classification model, the basic third EEG signal classification model and the basic fourth EEG signal classification model in the basic EEG signal classification model to obtain the EEG signal classification model after the current round of training based on the training sample set; wherein the EEG signal classification model after the current round of training based on the training sample set includes: the first EEG signal classification model after the current round of training based on the training sample set, the second EEG signal classification model after the current round of training based on the training sample set, the third EEG signal classification model after the current round of training based on the training sample set and the fourth EEG signal classification model after the current round of training based on the training sample set;

[0024] The test sample set is tested using the EEG signal classification model trained based on the training sample set in the current round to obtain the classification accuracy of the EEG signal classification model trained based on the training sample set in the current round;

[0025] Obtaining the average classification accuracy of the classification accuracy of the EEG signal classification model trained in the current round based on each training sample set in different training sample sets as the first average classification accuracy of the EEG signal classification model after the current round of training;

[0026] Obtain the average classification accuracy of the first average classification accuracy of the EEG signal classification model after multiple rounds of training as the second average classification accuracy;

[0027] When the second average classification accuracy rate does not reach the preset classification accuracy rate, the model parameters of the EEG signal classification model after the last round of training are adjusted.

[0028] Optionally, the using of the training sample set to perform a current round of training on a basic first EEG signal classification model, a basic second EEG signal classification model, a basic third EEG signal classification model, and a basic fourth EEG signal classification model in a basic EEG signal classification model to obtain a current round of EEG signal classification model trained based on the training sample set, comprises:

[0029] Performing feature selection on each training sample in the training sample set to obtain a feature matrix;

[0030] Using each training sample in the training sample set for each feature in the feature matrix, respectively perform a current round of training on the basic first EEG signal classification model, the basic second EEG signal classification model, and the basic third EEG signal classification model to obtain a current round of first EEG signal classification model trained based on the training sample set, a current round of second EEG signal classification model trained based on the training sample set, and a current round of third EEG signal classification model trained based on the training sample set;

[0031] For each training sample in the training sample set, the time domain features of the training sample are processed using the first EEG signal classification model trained based on the training sample set in the current round to obtain a first classification result of the training sample, the frequency domain features of the training sample are processed using the second EEG signal classification model trained based on the training sample set in the current round to obtain a second classification result of the training sample, and the topological features of the training sample are processed using the third EEG signal classification model trained based on the training sample set in the current round to obtain a third classification result of the training sample;

[0032] The basic fourth EEG signal classification model is trained by utilizing the first classification result of each training sample in the training sample set and the classification accuracy of the first EEG signal classification model trained based on the training sample set in the current round, the second classification result of the training sample and the classification accuracy of the second EEG signal classification model trained based on the training sample set in the current round, and the third classification result of the training sample and the classification accuracy of the third EEG signal classification model trained based on the training sample set in the current round, to obtain the fourth EEG signal classification model trained based on the training sample set in the current round.

[0033] Optionally, performing feature selection on each training sample in the training sample set to obtain a feature matrix includes:

[0034] Standardizing the feature information of each training sample in the training sample set to obtain first feature information;

[0035] Performing a preset transformation on the first characteristic information to obtain second characteristic information;

[0036] Obtaining a correlation score between the second feature information of each dimension and the label information of the sample EEG signal in the training sample set;

[0037] The feature matrix is ​​constructed according to the second feature information whose correlation score is greater than the preset correlation score.

[0038] According to a second aspect of the present disclosure, there is provided an electroencephalogram signal classification device, the device comprising:

[0039] A first acquisition module is used to acquire feature information of the EEG signal to be classified; wherein the feature information is used to reflect the signal type of the EEG signal to be classified, and the feature information includes time domain features, frequency domain features and topological features;

[0040] A second acquisition module is used to acquire a trained EEG signal classification model; wherein the trained EEG signal classification model includes a trained first EEG signal classification model, a trained second EEG signal classification model, a trained third EEG signal classification model and a trained fourth EEG signal classification model;

[0041] A processing module, used to process the time domain features using the trained first EEG signal classification model to obtain a first classification result of the EEG signal to be classified, to process the frequency domain features using the trained second EEG signal classification model to obtain a second classification result of the EEG signal to be classified, and to process the topological features using the trained third EEG signal classification model to obtain a third classification result of the EEG signal to be classified;

[0042] The processing module is also used to use the trained fourth EEG signal classification model to process the classification accuracy of the trained first EEG signal classification model and the first classification result, the classification accuracy of the trained second EEG signal classification model and the second classification result, and the classification accuracy of the trained third EEG signal classification model and the third classification result to obtain the target classification result of the EEG signal to be classified.

[0043] According to a third aspect of the present disclosure, an electronic device is provided, comprising a memory for storing executable computer instructions; and a processor for executing the EEG signal classification method according to the first aspect above under the control of the executable computer instructions.

[0044] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, on which computer instructions are stored. When the computer instructions are executed by a processor, the EEG signal classification method described in the first aspect above is executed.

[0045] According to the EEG signal classification method of the embodiment of the present disclosure, its trained EEG signal classification model includes a trained first EEG signal classification model, a trained second EEG signal classification model, a trained third EEG signal classification model and a trained fourth EEG signal classification model. In this way, the trained first EEG signal classification model can be used to process the time domain characteristics of the EEG signal to be classified to obtain the first classification result of the EEG signal to be classified, the trained second EEG signal classification model can be used to process the frequency domain characteristics of the EEG signal to be classified to obtain the second classification result of the EEG signal to be classified, and the trained third EEG signal classification model can be used to process the topological characteristics of the EEG signal to be classified to obtain the third classification result of the EEG signal to be classified. Finally, the trained fourth EEG signal classification model is used to process the classification accuracy and the first classification result of the trained first EEG signal classification model, the classification accuracy and the second classification result of the trained second EEG signal classification model, and the classification accuracy and the third classification result of the trained third EEG signal classification model to obtain the target classification result of the EEG signal to be classified. On the one hand, the trained EEG signal classification model involves the processing of the time domain features, frequency domain features and topological features of the EEG signal to be classified, and it uses the classification accuracy of different EEG signal classification models as weights to perform weighted summation of the classification results of different EEG signal classification models, thereby further enhancing the robustness of the model. In this way, the EEG signal to be classified can be accurately classified based on the trained EEG signal classification model.

[0046] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0048] Figure 1 is a schematic diagram of the hardware configuration of the EEG signal classification system according to an embodiment of the present disclosure;

[0049] Figure 2 is a flowchart of a method for classifying EEG signals according to an embodiment of the present disclosure;

[0050] Figure 3 is a schematic diagram of an impedance detection interface of a multi-channel EEG signal acquisition device according to an embodiment of the present disclosure;

[0051] Figure 4 is a schematic diagram of a flight mission according to an embodiment of the present disclosure;

[0052] Figure 5is a flowchart of an EEG signal classification method according to an example of the present disclosure;

[0053] Figure 6 is a schematic diagram of a second sample set according to an embodiment of the present disclosure;

[0054] Figure 7 is a schematic diagram of the principle of an EEG signal classification device according to an embodiment of the present disclosure;

[0055] Figure 8 Schematic diagram of the hardware configuration of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0056] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure unless otherwise specifically stated.

[0057] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.

[0058] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0059] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0060] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0061] <Hardware Configuration>

[0062] Figure 1 is a block diagram of the hardware configuration of the EEG signal classification system 100 according to an embodiment of the present disclosure.

[0063] according to Figure 1 As shown, the EEG signal classification system 100 includes an EEG signal acquisition device 1000 and an electronic device 2000. The EEG signal acquisition device 1000 may be a multi-channel EEG signal acquisition device, which is used to acquire the EEG signal to be classified and send the EEG signal to be classified to the electronic device 2000. The electronic device 2000 receives the EEG signal to be classified, and processes the EEG signal to be classified according to a pre-trained EEG signal classification model to obtain a target classification result of the EEG signal to be classified.

[0064] In one embodiment, the electronic device 2000 may be a server or a terminal device. The server may be an integrated server or a distributed server across multiple computers or computer data centers. The terminal device may be a portable computer, a desktop computer, a wearable device, or any other device having a computing device such as a processor and a storage device such as a memory, which is not limited in this embodiment.

[0065] like Figure 1 As shown, the electronic device 2000 may include a processor 2100, a memory 2200, an interface device 2300, a communication device 2400, a display device 2500, an input device 2600, a speaker 2700, a microphone 2800, and the like.

[0066] The processor 2100 may be a mobile version processor. The memory 2200 may include, for example, a ROM (read-only memory), a RAM (random access memory), a non-volatile memory such as a hard disk, etc. The interface device 2300 may include, for example, a USB interface, a headphone interface, etc. The communication device 2400 may, for example, be capable of wired or wireless communication. The communication device 2400 may include a short-range communication device, for example, any device for short-range wireless communication based on short-range wireless communication protocols such as Hilink protocol, WiFi (IEEE 802.11 protocol), Mesh, Bluetooth, ZigBee, Thread, Z-Wave, NFC, UWB, LiFi, etc. The communication device 2400 may also include a remote communication device, for example, any device for WLAN, GPRS, 2G / 3G / 4G / 5G remote communication. The display device 2500 may be, for example, a liquid crystal display, a touch display, etc. The input device 2600 may include, for example, a touch screen, a keyboard, etc. The terminal device 2000 may output audio information through a speaker 2700, and may collect audio information through a microphone 2800.

[0067] Despite Figure 1 In the figure, multiple devices are shown for the electronic device 2000, but the present disclosure may only involve some of the devices, for example, the electronic device 2000 only involves the memory 2200 and the processor 2100.

[0068] In the embodiment of the present disclosure, the memory 2200 of the electronic device 2000 is used to store instructions, and the instructions are used to control the processor 2100 to execute the EEG signal classification method provided by the embodiment of the present disclosure.

[0069] In the above description, the technician can design instructions according to the scheme disclosed in the present disclosure. How instructions control the processor to operate is well known in the art, so it will not be described in detail here.

[0070] The network 3000 may be a wireless communication network or a wired communication network, and may be a local area network or a wide area network. Figure 1 In the EEG signal classification system 100 shown, the EEG signal acquisition device 1000 and the electronic device 2000 can communicate through the network 3000.

[0071] It should be understood that although Figure 1 Only one EEG signal acquisition device 1000 and electronic device 2000 are shown, but this does not mean that the respective numbers are limited. The EEG signal classification system 100 may include multiple EEG signal acquisition devices 1000 and multiple electronic devices 2000.

[0072] <Method Example>

[0073] In this embodiment, a method for classifying an electroencephalogram signal is provided. The method for classifying an electroencephalogram signal can be implemented by an electronic device. The electronic device can be, for example, Figure 1 An electronic device 2000 is shown.

[0074] according to Figure 2 As shown, the EEG signal classification method of the embodiment of the present disclosure may include the following steps S2100 to S2400.

[0075] Step S2100, obtaining characteristic information of the EEG signal to be classified.

[0076] Among them, the EEG signal to be classified can be an EEG signal after the electronic device performs signal preprocessing on the initial EEG signal to be classified of any user collected by the multi-channel EEG signal acquisition device. At this time, the initial EEG signal to be classified can be called the initial multi-channel EEG signal to be classified, and the EEG signal to be classified can be called the multi-channel EEG signal to be classified. Exemplarily, the multi-channel EEG signal acquisition device is a 32-channel EEG signal acquisition device, and the 32 channels are Fp1, Fp2, F3, F4, C3, C4, P3, P4, O1, O2, F7, F8, T7, T8, P7, P8, Fz, Cz, Pz, CPZ, FC1, FC2, CP1, CP2, FC5, FC6, CP5, CP6, TP9, TP10, FT9, FT10. The 32-channel EEG signal acquisition device can collect the user's 32-channel EEG signals as the initial EEG signals to be classified.

[0077] It should be noted that before the data acquisition experiment, it is necessary to perform impedance detection on the multi-channel EEG signal acquisition equipment and reduce the impedance of all electrodes to below 30KΩ. Figure 3 The impedance detection interface diagram of a 32-channel EEG signal acquisition device is shown.

[0078] In an optional embodiment, the electronic device may obtain an initial EEG signal to be classified, and perform signal preprocessing on the initial EEG signal to be classified to obtain the EEG signal to be classified. The signal preprocessing at least includes: signal reorganization processing, artifact removal processing, artifact category detection processing, and signal segmentation processing.

[0079] Specifically, the electronic device can first obtain the initial EEG signal to be classified, and then sequentially remove the non-physiological artifacts and physiological artifacts in the initial EEG signal to be classified through signal reorganization processing, artifact removal processing, artifact category detection processing and signal segmentation processing. Among them, non-physiological artifacts include, but are not limited to, saturation artifacts, sudden slopes, outliers and power frequency artifacts. Physiological artifacts include, but are not limited to, electromyography, electrooculography and electrocardiography.

[0080] The characteristic information is used to reflect the signal type of the EEG signal to be classified, and the characteristic information includes time domain characteristics, frequency domain characteristics and topological characteristics.

[0081] The time domain feature includes at least one of mean, standard deviation, kurtosis and skewness.

[0082] The above kurtosis can be used to measure the steepness of the probability distribution of random variables. The kurtosis K refers to the following formula:

[0083]

[0084] The above skewness can be used to measure the asymmetry of the probability distribution of random variables. The skewness S is specifically referred to the following formula:

[0085]

[0086] The frequency domain feature includes at least one of a power spectrum estimation of a theta band, a power spectrum estimation of an alpha band, and a power spectrum estimation of a beta band.

[0087] Specifically, the electronic device can first convert the EEG signal to be classified into the frequency domain by Fourier transform, and then divide it into theta band, alpha band and beta band according to the frequency. For each band, a preset time window is used for division, N points are sampled in each time window and Fourier transform is performed, the square of the amplitude is taken and divided by N, and finally the results of all time windows are averaged to obtain an estimate of the power spectrum of the corresponding band.

[0088] The topological feature includes at least one of an average clustering coefficient, density and global efficiency.

[0089] In an optional embodiment, the feature information includes topological features, and the step S2100 of acquiring the feature information of the EEG signal to be classified may further include the following steps S2110 to S2120:

[0090] Step S2110, constructing a graph structure of the EEG signals to be classified by taking the signal strength corresponding to each EEG signal in the EEG signals to be classified as a node, taking the connection between any two nodes as an edge, and taking the slope of the edge as the weight corresponding to the edge.

[0091] Step S2120, obtaining the topological features of the EEG signal to be classified according to the graph structure of the EEG signal to be classified.

[0092] Among them, there is a node c between node a and node b, and nodes a, node b and node c satisfy: x b -x a <T min In the case of , it is determined that there is an edge between node a and node b, where (x a ,y a ) a is the moment when the EEG signal corresponding to node a is in the EEG signal to be classified, y a is the signal strength of the EEG signal, (x b ,y b ) b is the moment when the EEG signal corresponding to node b is in the EEG signal to be classified, y b is the signal strength of the EEG signal, (x c ,y c ) c is the time when the EEG signal corresponding to node c is in the EEG signal to be classified, y c is the signal strength of the EEG signal.

[0093] Specifically, the electronic device uses graph signal processing to convert a single-channel EEG signal into a graph structure, and extracts corresponding topological features based on the graph structure. Usually, the graph structure consists of nodes, edges, and weights, and the signal strength of the EEG signal at different times is used as the node of the graph structure.

[0094] The edge can be calculated by the weighted visibility graph (WVG) method. Assume that nodes a and b are any two points on the time series of the EEG signal to be classified, which are two nodes in the graph structure, and there is a node c between the two nodes. If the following conditions are met, there is an edge between nodes a and b:

[0095]

[0096] However, this method has certain problems. If the above formula is not constrained, then in the determined graph structure, there will be a problem that a node is connected to too many nodes, that is, the EEG signal at a certain moment will be connected to the EEG signal at a distant moment. However, there is not a strong connection between EEG signals at distant moments, so the graph structure constructed by this method has certain problems.

[0097] To solve this problem, it is necessary to constrain EEG signals at distant moments. This embodiment sets a time window for the node to be studied, and assumes that EEG signals outside the time window have no effect on the EEG signals at that moment. Therefore, when studying the topological structure of the node, only other EEG signal nodes within the time window need to be considered. The connection between node a and node b can be determined by whether the following formula is satisfied:

[0098]

[0099] After the electronic device considers the connection between all nodes and other nodes, the edges in the graph structure are obtained. After the edges of the graph structure are constructed, a weight can be set for each edge to obtain a weight matrix. Among them, the slope corresponding to the edge can be used as the weight.

[0100] After constructing the graph structure of the EEG signal to be classified, the electronic device can use the graph structure to calculate the topological features.

[0101] The above average clustering coefficient represents the average value of all local clustering coefficients calculated from each vertex of the network. The average clustering coefficient CC(n) is specifically referred to the following formula:

[0102]

[0103] Among them, T(n) represents the number of triangles formed around node n, and K(n) represents the degree of node n, that is, the number of edges connecting node n.

[0104] The above density represents the ratio of the number of edges that actually exist between nodes to the maximum number of possible edges. The density D is specifically calculated by the following formula:

[0105]

[0106] Among them, N represents the number of existing edges, and K represents the size of the graph structure.

[0107] The above global efficiency represents the efficiency of the network in the global scope. The global efficiency GE is specifically referred to the following formula:

[0108]

[0109] Among them, p ij Represents the shortest path length.

[0110] That is to say, this embodiment uses a graph signal processing method to convert the EEG signal of a single channel into a graph structure and extract the corresponding features. In terms of the graph structure construction method, based on the traditional weighted visible graph (WVG) method, the time window method is used to further optimize the EEG signals at distant moments and distinguish the correlation between EEG signals at close moments.

[0111] In this embodiment, the electronic device constructs a time domain feature matrix using the mean, standard deviation, kurtosis and skewness of different channels in the EEG signal to be classified, constructs a frequency domain feature matrix using the power spectrum of theta band, the power spectrum of alpha band and the power spectrum of beta band of different channels in the EEG signal to be classified, and constructs a topological feature matrix using the average clustering coefficient, density and global efficiency of different channels in the EEG signal to be classified.

[0112] After executing step S2100 to obtain the characteristic information of the EEG signal to be classified, proceed to:

[0113] Step S2200, obtaining the trained EEG signal classification model.

[0114] The trained EEG signal classification model includes a trained first EEG signal classification model, a trained second EEG signal classification model, a trained third EEG signal classification model and a trained fourth EEG signal classification model.

[0115] The above first EEG signal classification model is used to process the time domain features of the EEG signal to be classified to obtain the first classification result of the EEG signal to be classified, that is, the input of the first EEG signal classification model is the time domain features of the EEG signal to be classified, and the output is the first classification result of the EEG signal to be classified determined based on the time domain features. It should be noted that the first classification result of the EEG signal to be classified can be a low-load EEG signal or a high-load EEG signal.

[0116] The above second EEG signal classification model is used to process the frequency domain features of the EEG signal to be classified to obtain the second classification result of the EEG signal to be classified, that is, the input of the second EEG signal classification model is the frequency domain features of the EEG signal to be classified, and the output is the second classification result of the EEG signal to be classified determined based on the frequency domain features. It should be noted that the second classification result of the EEG signal to be classified can be a low-load EEG signal or a high-load EEG signal.

[0117] The above third EEG signal classification model is used to process the topological features of the EEG signal to be classified to obtain the third classification result of the EEG signal to be classified, that is, the input of the third EEG signal classification model is the topological features of the EEG signal to be classified, and the output is the third classification result of the EEG signal to be classified determined based on the topological features. It should be noted that the third classification result of the EEG signal to be classified can be a low-load EEG signal or a high-load EEG signal.

[0118] It should be noted that the first EEG signal classification model, the second EEG signal classification model, and the third EEG signal classification model can all be three trained support vector machine (SVM) classifiers. The support vector machine classifier is a very powerful and flexible supervised machine learning algorithm. By using kernel functions to map data to high-dimensional space, it can not only solve linearly inseparable problems in the original space, but also handle nonlinear situations.

[0119] The above fourth EEG signal classification model adopts a decision layer fusion method, which is used to take the classification accuracy of the first EEG signal classification model, the classification accuracy of the second EEG signal classification model and the classification accuracy of the third EEG signal as weights, and perform weighted summation on the first classification result output by the first EEG signal classification model, the second classification result output by the second EEG signal classification model and the third classification result output by the third EEG signal classification model to obtain the target classification result of the EEG signal to be classified. It should be noted that the target classification result of the EEG signal to be classified can be a low-load EEG signal or a high-load EEG signal.

[0120] After executing the above step S2200 to obtain the trained EEG signal classification model; wherein the trained EEG signal classification model includes the trained first EEG signal classification model, the trained second EEG signal classification model, the trained third EEG signal classification model and the trained fourth EEG signal classification model, enter:

[0121] Step S2300, using the trained first EEG signal classification model to process the time domain features to obtain a first classification result of the EEG signal to be classified, using the trained second EEG signal classification model to process the frequency domain features to obtain a second classification result of the EEG signal to be classified, and using the trained third EEG signal classification model to process the topological features to obtain a third classification result of the EEG signal to be classified.

[0122] In this embodiment, refer to Figure 5The electronic device inputs the time domain features of the EEG signal to be classified into the trained first EEG signal classification model to obtain a first classification result G1 of the EEG signal to be classified, inputs the frequency domain features of the EEG signal to be classified into the trained second EEG signal classification model to obtain a second classification result G2 of the EEG signal to be classified, and inputs the topological features of the EEG signal to be classified into the trained third EEG signal classification model to obtain a third classification result G3 of the EEG signal to be classified.

[0123] Step S2400, using the trained fourth EEG signal classification model to process the classification accuracy of the trained first EEG signal classification model and the first classification result, the classification accuracy of the trained second EEG signal classification model and the second classification result, and the classification accuracy of the trained third EEG signal classification model and the third classification result, to obtain the target classification result of the EEG signal to be classified.

[0124] In this embodiment, refer to Figure 5 The electronic device inputs the first classification result G1 of the EEG signal to be classified, the classification accuracy C1 of the first EEG signal classification model, the second classification result G2 of the EEG signal to be classified, the classification accuracy C2 of the second EEG signal classification model, the third classification result G3 of the EEG signal to be classified, and the classification accuracy C3 of the third EEG signal classification model into the fourth EEG signal classification model. The fourth EEG signal classification model uses the classification accuracy C1 of the first EEG signal classification model, the classification accuracy C2 of the second EEG signal classification model, and the classification accuracy C3 of the third EEG signal classification model as weights to perform weighted summation on the first classification result G1 of the EEG signal to be classified, the second classification result G2 of the EEG signal to be classified, and the third classification result G3 of the EEG signal to be classified to obtain a target classification result C of the EEG signal to be classified. That is, the target classification result satisfies: C=G1*C1+G2*C2+G3*C3.

[0125] According to the method of the embodiment of the present disclosure, the trained EEG signal classification model includes a trained first EEG signal classification model, a trained second EEG signal classification model, a trained third EEG signal classification model and a trained fourth EEG signal classification model. In this way, the trained first EEG signal classification model can be used to process the time domain characteristics of the EEG signal to be classified to obtain the first classification result of the EEG signal to be classified, the trained second EEG signal classification model can be used to process the frequency domain characteristics of the EEG signal to be classified to obtain the second classification result of the EEG signal to be classified, and the trained third EEG signal classification model can be used to process the topological characteristics of the EEG signal to be classified to obtain the third classification result of the EEG signal to be classified. Finally, the trained fourth EEG signal classification model is used to process the classification accuracy and the first classification result of the trained first EEG signal classification model, the classification accuracy and the second classification result of the trained second EEG signal classification model, and the classification accuracy and the third classification result of the trained third EEG signal classification model to obtain the target classification result of the EEG signal to be classified. On the one hand, the trained EEG signal classification model involves the processing of the time domain features, frequency domain features and topological features of the EEG signal to be classified, and it uses the classification accuracy of different EEG signal classification models as weights to perform weighted summation of the classification results of different EEG signal classification models, thereby further enhancing the robustness of the model. In this way, the EEG signal to be classified can be accurately classified based on the trained EEG signal classification model.

[0126] In one embodiment, the above step S2200 of obtaining the trained EEG signal classification model may further include the following steps S2210 to S2270:

[0127] Step S2210: Obtain a first sample set.

[0128] Each sample in the first sample set includes a sample EEG signal and label information of the sample EEG signal. The label information of the sample EEG signal is used to identify the true classification result of the sample EEG signal. For example, the label information of the sample EEG signal can be "0", indicating that the sample EEG signal is a low-load EEG signal, and the label information of the sample EEG signal can also be "1", indicating that the sample EEG signal is a high-load EEG signal.

[0129] Usually, when an individual performs a task, it takes a lot of mental effort. It can be said that every time an action is performed, every time an operation is performed, the individual must allocate the mental resources to perform the task. Performing multiple tasks requires multiple allocations. Therefore, the higher the complexity of the task, the more resources the individual needs to perform the task, and the greater the brain load. There is a positive relationship between task complexity and brain load.

[0130] The degree of brain load can be measured from the following two aspects: 1. The number of information processed at the same time; 2. The total information demand of the task on the person at a certain moment (the difficulty of the task). Therefore, this embodiment designs two flight tasks of different difficulty. The EEG signals collected when the subjects complete the low-difficulty flight task are used as low-load EEG signals, labeled as 0; the EEG signals collected when the subjects complete the high-difficulty flight task are high-load EEG signals, labeled as 1. The low-difficulty and high-difficulty flight tasks respectively represent the subjects performing the same maneuvers (such as level flight, circling, etc.) according to the instructor's commands under different climatic conditions (the subjects have different visibility). In order to prevent the subjects from memorizing the rules, the order of the maneuvers completed by the subjects will be adjusted in different test trials, but if Figure 4 As shown in the figure, under different climatic conditions in the same test, the maneuvers of the low-difficulty classification task and the high-difficulty flight task are exactly the same. In other words, this embodiment designs a reasonable and practical experimental paradigm for the simulated cockpit environment and existing flight control tasks, which can better extract the high-load and low-load states of the subjects.

[0131] It should be noted that, as mentioned above, the sample EEG signal can be collected based on a multi-channel EEG signal acquisition device, such as a 32-channel EEG signal acquisition device, and the sample EEG signal is a multi-channel EEG signal, such as a 32-channel EEG signal. In addition, it is necessary to perform impedance detection on the multi-channel EEG signal acquisition device and reduce the impedance of all electrodes to below 30KΩ.

[0132] Step S2220: performing signal preprocessing on each of the samples in the first sample set to obtain a second sample set.

[0133] In this step S2220, the electronic device performs signal reconstruction processing, artifact removal processing, artifact category detection processing, and signal segmentation processing on the sample EEG signal of each sample in the first sample set in sequence to obtain the processed sample EEG signal of the sample, and then obtain the second sample set.

[0134] Each sample in the second sample set includes a processed sample EEG signal and label information of the processed sample EEG signal.

[0135] The second sample set includes a test sample set and a training sample set.

[0136] In this step S2220, in order to better select the model parameters of the EEG signal classification model and ensure the training effect of the EEG signal classification model, K-fold cross validation is used for processing. At the same time, considering the problem of sample size, it is finally decided that K is 5. Figure 6, 5-fold cross validation is as follows: for each training in any round, the second sample set is first shuffled and then divided into 5 parts, one of which is used as the test sample set each time to verify the EEG signal classification model, and the other 4 are used as the training sample sets for training the EEG signal classification model. Each round of training is repeated 5 times. In other words, each round of training involves Figure 5 The first division, second division, third division, fourth division and fifth division are shown.

[0137] Step S2230, using the training sample set, performs the current round of training on the basic first EEG signal classification model, the basic second EEG signal classification model, the basic third EEG signal classification model and the basic fourth EEG signal classification model in the basic EEG signal classification model to obtain the current round of EEG signal classification model trained based on the training sample set.

[0138] Among them, the EEG signal classification model after the current round of training based on the training sample set includes: the first EEG signal classification model after the current round of training based on the training sample set, the second EEG signal classification model after the current round of training based on the training sample set, the third EEG signal classification model after the current round of training based on the training sample set and the fourth EEG signal classification model after the current round of training based on the training sample set.

[0139] Optionally, this step S2230 uses the training sample set to perform a current round of training on the basic first EEG signal classification model, the basic second EEG signal classification model, the basic third EEG signal classification model, and the basic fourth EEG signal classification model in the basic EEG signal classification model, and obtaining the current round of EEG signal classification model trained based on the training sample set may further include the following steps S2231 to S2234:

[0140] Step S2231, performing feature selection on each training sample in the training sample set to obtain a feature matrix.

[0141] In this step S2231, feature selection is performed on each training sample in the training sample set, and obtaining the feature matrix can further include: standardizing the feature information of each training sample in the training sample set to obtain first feature information; performing a preset transformation on the first feature information to obtain second feature information; obtaining the correlation score between the second feature information of each dimension and the label information of the sample EEG signal in the training sample set; and constructing the feature matrix according to the second feature information whose correlation score is greater than the preset correlation score.

[0142] Specifically, the electronic device uses the Z-score standardization method to standardize the time domain feature matrix, frequency domain feature matrix, and topological feature matrix of each training sample in the training sample set, and calculates the mean and standard deviation of each feature. Then, each feature is subtracted from the mean and divided by the standard deviation. The data after such processing conforms to the standard normal distribution, that is, the mean is 0 and the standard deviation is 1. Then, using the 3α criterion, the training samples with a standard deviation exceeding 3 times the mean are eliminated. Finally, the Pearson correlation is calculated for the features of each dimension and the label information of the training sample, and the features with significant correlation (p < 0.01) are retained to form a feature matrix.

[0143] Next, we will show an example of the calculation process of Pearson correlation. For variables X and Y, there are n samples (x_1, y_1), (x_2, y_2), ..., (x_n, y_n), where x_i represents the features of each dimension and y_i is the label of the sample. The calculation formula of Pearson correlation coefficient r is as follows:

[0144]

[0145] in, and is the average value.

[0146] Next, construct the statistic t, and the calculation formula is as follows:

[0147]

[0148] Finally, by querying the t distribution table, you can find out the size of the p value.

[0149] Step S2232, using each training sample in the training sample set for each feature in the feature matrix, respectively perform the current round of training on the basic first EEG signal classification model, the basic second EEG signal classification model and the basic third EEG signal classification model, to obtain the current round of the first EEG signal classification model trained based on the training sample set, the current round of the second EEG signal classification model trained based on the training sample set, and the current round of the third EEG signal classification model trained based on the training sample set.

[0150] Step S2233, for each training sample in the training sample set, use the first EEG signal classification model trained based on the training sample set in the current round to process the time domain features of the training sample to obtain the first classification result of the training sample, use the second EEG signal classification model trained based on the training sample set in the current round to process the frequency domain features of the training sample to obtain the second classification result of the training sample, and use the third EEG signal classification model trained based on the training sample set in the current round to process the topological features of the training sample to obtain the third classification result of the training sample.

[0151] Step S2234, using the first classification result of each training sample in the training sample set and the classification accuracy of the first EEG signal classification model trained in the current round based on the training sample set, the second classification result of the training sample and the classification accuracy of the second EEG signal classification model trained in the current round based on the training sample set, and the third classification result of the training sample and the classification accuracy of the third EEG signal classification model trained in the current round based on the training sample set, train the basic fourth EEG signal classification model to obtain the fourth EEG signal classification model trained in the current round based on the training sample set.

[0152] According to the above steps S2231 to S2234, the electronic device divides different types of EEG signals and designs corresponding EEG signal classification models respectively. Finally, the classification accuracy of the EEG signal classification model is used as the weight to perform weighted summation on the classification labels of different EEG signal classification models, thereby further enhancing the robustness of the trained EEG signal classification model.

[0153] Step S2240, testing the test sample set using the EEG signal classification model trained based on the training sample set in the current round to obtain the classification accuracy of the EEG signal classification model trained based on the training sample set in the current round.

[0154] Step S2250, obtaining the average classification accuracy of the classification accuracy of the EEG signal classification model trained in the current round based on each training sample set in different training sample sets as the first average classification accuracy of the EEG signal classification model after the current round of training.

[0155] Exemplarily, the current round of training is repeated 5 times to complete, and 5 trained EEG signal classification models are obtained. Then, the classification accuracy of each trained EEG signal classification model is calculated and the average is calculated to obtain the classification accuracy of the current round of 5-fold cross-validation.

[0156] Step S2260, obtaining the average classification accuracy of the first average classification accuracy of the EEG signal classification model after multiple rounds of training as the second average classification accuracy.

[0157] Exemplarily, the electronic device can perform 20 rounds of training, each round of training is repeated 5 times based on the divided different test sample sets and multiple training sample sets, and the 5 trained EEG signal classification models after the round of training are calculated, and then the classification accuracy of each trained EEG signal classification model is calculated and averaged to obtain the classification accuracy of the 5-fold cross validation of the round. Finally, the average classification accuracy obtained from the 20 rounds of training is averaged to obtain the final average classification accuracy.

[0158] Step S2270: When the second average classification accuracy rate does not reach a preset classification accuracy rate, adjust the model parameters of the EEG signal classification model after the last round of training.

[0159] Specifically, when the final average classification accuracy does not reach the preset classification accuracy, the electronic device will adjust the model parameters of the EEG signal classification model after the last round of training, such as adjusting the penalty coefficient c and the linear kernel function t, where t is 1, indicating that the linear kernel function selects a polynomial kernel function, t is 2, indicating that the linear kernel function selects a Gaussian kernel function, t is 3, indicating that the linear kernel function selects a sigmoid function, and t is 4, indicating that the linear kernel function selects a predefined kernel function.

[0160] In this embodiment, the training results show that the accuracy is relatively high when c=0.9 and t=2. Therefore, a Gaussian kernel function is used. The Gaussian kernel function formula is as follows:

[0161]

[0162] Here, σ is the default value.

[0163] According to the method of this embodiment, the trained EEG signal classification model includes a trained first EEG signal classification model, a trained second EEG signal classification model, a trained third EEG signal classification model and a trained fourth EEG signal classification model. In this way, the trained first EEG signal classification model can be used to process the time domain characteristics of the EEG signal to be classified to obtain the first classification result of the EEG signal to be classified, the trained second EEG signal classification model can be used to process the frequency domain characteristics of the EEG signal to be classified to obtain the second classification result of the EEG signal to be classified, and the trained third EEG signal classification model can be used to process the topological characteristics of the EEG signal to be classified to obtain the third classification result of the EEG signal to be classified. Finally, the trained fourth EEG signal classification model is used to process the classification accuracy and the first classification result of the trained first EEG signal classification model, the classification accuracy and the second classification result of the trained second EEG signal classification model, and the classification accuracy and the third classification result of the trained third EEG signal classification model to obtain the target classification result of the EEG signal to be classified. On the one hand, the trained EEG signal classification model involves the processing of the time domain features, frequency domain features and topological features of the EEG signal to be classified, and it uses the classification accuracy of different EEG signal classification models as weights to perform weighted summation of the classification results of different EEG signal classification models, thereby further enhancing the robustness of the model. In this way, the EEG signal to be classified can be accurately classified based on the trained EEG signal classification model.

[0164] <Device Example>

[0165] In this embodiment, a device 7000 for classifying electroencephalogram signals is provided. Figure 7 As shown, the EEG signal classification device 7000 may include a first acquisition module 7100 , a second acquisition module 7200 , and a processing module 7300 .

[0166] The first acquisition module 7100 is used to acquire feature information of the EEG signal to be classified; wherein the feature information is used to reflect the signal type of the EEG signal to be classified, and the feature information includes time domain features, frequency domain features and topological features;

[0167] The second acquisition module 7200 is used to acquire a trained EEG signal classification model; wherein the trained EEG signal classification model includes a trained first EEG signal classification model, a trained second EEG signal classification model, a trained third EEG signal classification model and a trained fourth EEG signal classification model;

[0168] The processing module 7300 is used to process the time domain features using the trained first EEG signal classification model to obtain a first classification result of the EEG signal to be classified, process the frequency domain features using the trained second EEG signal classification model to obtain a second classification result of the EEG signal to be classified, and process the topological features using the trained third EEG signal classification model to obtain a third classification result of the EEG signal to be classified;

[0169] The processing module 7300 is also used to use the trained fourth EEG signal classification model to process the classification accuracy of the trained first EEG signal classification model and the first classification result, the classification accuracy of the trained second EEG signal classification model and the second classification result, and the classification accuracy of the trained third EEG signal classification model and the third classification result to obtain the target classification result of the EEG signal to be classified.

[0170] In one embodiment, the first acquisition module 7100 is also used to acquire an initial EEG signal to be classified; perform signal preprocessing on the initial EEG signal to be classified to obtain the EEG signal to be classified; wherein the signal preprocessing at least includes in sequence: signal reconstruction processing, artifact removal processing, artifact category detection processing, and signal segmentation processing.

[0171] In one embodiment, the time domain feature includes at least one of a mean, a standard deviation, a kurtosis, and a skewness;

[0172] The frequency domain feature includes at least one of a power spectrum of theta band, a power spectrum of alpha band and a power spectrum of beta band;

[0173] The topological feature includes at least one of an average clustering coefficient, density and global efficiency.

[0174] In one embodiment, the feature information includes topological features, and the first acquisition module 7100 is specifically used to: construct a graph structure of the EEG signal to be classified using the signal strengths corresponding to each EEG signal in the EEG signal to be classified as nodes, the connection between any two nodes as edges, and the slope of the edges as the weights corresponding to the edges; obtain the topological features of the EEG signal to be classified according to the graph structure of the EEG signal to be classified;

[0175] Among them, there is a node c between node a and node b, and nodes a, node b and node c satisfy: x b -x a <T min In the case of , it is determined that there is an edge between node a and node b, where (x a ,ya ) a is the time when the EEG signal corresponding to node a is in the EEG signal to be classified, y a is the signal strength of the EEG signal, (x b ,y b ) b is the time when the EEG signal corresponding to node b is in the EEG signal to be classified, y b is the signal strength of the EEG signal, (x c ,y c ) c is the time when the EEG signal corresponding to node c is in the EEG signal to be classified, y c is the signal strength of the EEG signal, T min The preset time window.

[0176] In one embodiment, the second acquisition module 7200 is specifically used to: acquire a first sample set; wherein each sample in the first sample set includes a sample EEG signal and label information of the sample EEG signal; perform signal preprocessing on each of the samples in the first sample set to obtain a second sample set; wherein the second sample set includes a test sample set and a training sample set; use the training sample set to perform a current round of training on a basic first EEG signal classification model, a basic second EEG signal classification model, a basic third EEG signal classification model, and a basic fourth EEG signal classification model in a basic EEG signal classification model to obtain an EEG signal classification model trained based on the training sample set in the current round; wherein the EEG signal classification model trained based on the training sample set in the current round includes: the first EEG signal classification model trained based on the training sample set in the current round, the second EEG signal classification model trained based on the training sample set in the current round, and the fourth EEG signal classification model trained based on the training sample set in the current round. an EEG signal classification model, a third EEG signal classification model after training based on the training sample set in the current round, and a fourth EEG signal classification model after training based on the training sample set in the current round; using the EEG signal classification model after training based on the training sample set in the current round to test the test sample set, and obtaining the classification accuracy of the EEG signal classification model after training based on the training sample set in the current round; obtaining the average classification accuracy of the classification accuracy of the EEG signal classification model after training based on each training sample set in different training sample sets in the current round, as the first average classification accuracy of the EEG signal classification model after training in the current round; obtaining the average classification accuracy of the first average classification accuracy of the EEG signal classification model after multiple rounds of training, as the second average classification accuracy; when the second average classification accuracy does not reach the preset classification accuracy, adjusting the model parameters of the EEG signal classification model after the last round of training.

[0177] In one embodiment, the second acquisition module 7200 is specifically used to: perform feature selection on each training sample in the training sample set to obtain a feature matrix; use each training sample in the training sample set to perform the current round of training on the basic first EEG signal classification model, the basic second EEG signal classification model and the basic third EEG signal classification model for each feature in the feature matrix, and obtain the first EEG signal classification model trained based on the training sample set in the current round, the second EEG signal classification model trained based on the training sample set in the current round, and the third EEG signal classification model trained based on the training sample set in the current round; for each training sample in the training sample set, use the first EEG signal classification model trained based on the training sample set in the current round to process the time domain features of the training sample to obtain the first classification result of the training sample, and use the first EEG signal classification model trained based on the training sample set in the current round to obtain the first classification result of the training sample. The second EEG signal classification model after training based on the training sample set is used to process the frequency domain features of the training samples to obtain the second classification results of the training samples, and the third EEG signal classification model after training based on the training sample set in the current round is used to process the topological features of the training samples to obtain the third classification results of the training samples; the basic fourth EEG signal classification model is trained using the first classification result of each training sample in the training sample set and the classification accuracy of the first EEG signal classification model after training based on the training sample set in the current round, the second classification result of the training sample and the classification accuracy of the second EEG signal classification model after training based on the training sample set in the current round, and the third classification result of the training sample and the classification accuracy of the third EEG signal classification model after training based on the training sample set in the current round to obtain the fourth EEG signal classification model after training based on the training sample set in the current round.

[0178] In one embodiment, the second acquisition module 7200 is specifically used to: standardize the feature information of each training sample in the training sample set to obtain first feature information; perform a preset transformation on the first feature information to obtain second feature information; obtain the correlation score between the second feature information of each dimension and the label information of the sample EEG signal in the training sample set; and construct the feature matrix based on the second feature information whose correlation score is greater than the preset correlation score.

[0179] According to this embodiment, its trained EEG signal classification model includes a trained first EEG signal classification model, a trained second EEG signal classification model, a trained third EEG signal classification model and a trained fourth EEG signal classification model. In this way, the trained first EEG signal classification model can be used to process the time domain characteristics of the EEG signal to be classified to obtain the first classification result of the EEG signal to be classified, the trained second EEG signal classification model can be used to process the frequency domain characteristics of the EEG signal to be classified to obtain the second classification result of the EEG signal to be classified, and the trained third EEG signal classification model can be used to process the topological characteristics of the EEG signal to be classified to obtain the third classification result of the EEG signal to be classified. Finally, the trained fourth EEG signal classification model is used to process the classification accuracy and the first classification result of the trained first EEG signal classification model, the classification accuracy and the second classification result of the trained second EEG signal classification model, and the classification accuracy and the third classification result of the trained third EEG signal classification model to obtain the target classification result of the EEG signal to be classified. On the one hand, the trained EEG signal classification model involves the processing of time domain features, frequency domain features and topological features of the EEG signal to be classified. Moreover, it uses the classification accuracy of different EEG signal classification models as weights, and performs weighted summation of the classification results of different EEG signal classification models, which further strengthens the robustness of the model.

[0180] <Equipment Embodiment>

[0181] Figure 8 FIG. 1 is a schematic diagram of the hardware structure of an electronic device according to an embodiment. Figure 8 As shown, the electronic device 8000 includes a processor 8100 and a memory 8200 .

[0182] The memory 8200 can be used to store executable computer instructions.

[0183] The processor 8100 can be used to execute the EEG signal classification method described in the method embodiment of the present disclosure under the control of the executable computer instructions.

[0184] The electronic device 8000 may be Figure 1 The electronic device 2000 shown may also be a device with other hardware structures, which is not limited here.

[0185] In another embodiment, the electronic device 8000 may include the above EEG signal classification device 7000 .

[0186] In one embodiment, each module of the above EEG signal classification device 7000 can be implemented by the processor 8100 running computer instructions stored in the memory 8200 .

[0187] <Computer Readable Storage Medium>

[0188] The embodiment of the present disclosure further provides a computer-readable storage medium on which computer instructions are stored. When the computer instructions are executed by a processor, the EEG signal classification method provided by the embodiment of the present disclosure is executed.

[0189] The present disclosure 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 disclosure.

[0190] Computer readable storage medium can be a tangible device that can hold and store instructions used by an 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 of the above. More specific examples (non-exhaustive list) of computer readable storage medium include: portable computer disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanical encoding device, such as a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination of the above. The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (e.g., a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0191] 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, optical fiber transmissions, wireless transmissions, 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 for storage in the computer-readable storage medium in each computing / processing device.

[0192] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related 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++, etc., and conventional procedural programming languages-such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a 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 may 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 may 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) may be personalized by utilizing the state information of a computer-readable program instruction, and the electronic circuit may execute a computer-readable program instruction, thereby realizing various aspects of the present disclosure.

[0193] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0194] 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 that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0195] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating 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 boxes in the flowchart and / or block diagram.

[0196] The flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of an instruction, and the module, a program segment or a part of an instruction contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a different order from the order 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 the flowchart, and the combination of the boxes in the block diagram and / or the 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 it is equivalent to implement it by hardware, implement it by software, and implement it by combining software and hardware.

[0197] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the marketplace, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein. The scope of the present disclosure is defined by the appended claims.

Claims

1. A method for classifying electroencephalogram signals, characterized in that: The method comprises: Acquire characteristic information of the EEG signal to be classified; wherein the characteristic information is used to reflect the signal type of the EEG signal to be classified, and the characteristic information includes time domain characteristics, frequency domain characteristics and topological characteristics; Obtaining a trained EEG signal classification model; wherein the trained EEG signal classification model includes a trained first EEG signal classification model, a trained second EEG signal classification model, a trained third EEG signal classification model, and a trained fourth EEG signal classification model; Using the trained first EEG signal classification model to process the time domain features to obtain a first classification result of the EEG signal to be classified, using the trained second EEG signal classification model to process the frequency domain features to obtain a second classification result of the EEG signal to be classified, and using the trained third EEG signal classification model to process the topological features to obtain a third classification result of the EEG signal to be classified; Using the trained fourth EEG signal classification model to process the classification accuracy of the trained first EEG signal classification model and the first classification result, the classification accuracy of the trained second EEG signal classification model and the second classification result, and the classification accuracy of the trained third EEG signal classification model and the third classification result, to obtain a target classification result of the EEG signal to be classified; The step of obtaining characteristic information of the EEG signal to be classified includes: The graph structure of the EEG signal to be classified is constructed by taking the signal strength corresponding to each EEG signal in the EEG signal to be classified as a node, taking the connection between any two nodes as an edge, and taking the slope of the edge as the weight corresponding to the edge; Obtaining topological features of the EEG signal to be classified according to the graph structure of the EEG signal to be classified; Among them, there is a node c between node a and node b, and nodes a, node b and node c satisfy: , In the case of , it is determined that there is an edge between node a and node b, where ( )middle is the moment when the EEG signal corresponding to node a is in the EEG signal to be classified, is the signal strength of the EEG signal, middle is the moment when the EEG signal corresponding to node b is in the EEG signal to be classified, is the signal strength of the EEG signal, middle is the moment when the EEG signal corresponding to node c is in the EEG signal to be classified, is the signal strength of the EEG signal, The preset time window.

2. The method according to claim 1, characterized in that The method further comprises: Obtaining an initial EEG signal to be classified; Performing signal preprocessing on the initial EEG signal to be classified to obtain the EEG signal to be classified; The signal preprocessing at least includes: signal reorganization processing, artifact removal processing, artifact category detection processing, and signal segmentation processing.

3. The method according to claim 1, characterized in that The time domain feature includes at least one of mean, standard deviation, kurtosis and skewness; The frequency domain feature includes at least one of a power spectrum of theta band, a power spectrum of alpha band and a power spectrum of beta band; The topological feature includes at least one of an average clustering coefficient, density and global efficiency.

4. The method according to claim 1, characterized in that: The step of obtaining the trained EEG signal classification model comprises: Acquire a first sample set; wherein each sample in the first sample set includes a sample EEG signal and label information of the sample EEG signal; Performing signal preprocessing on each of the samples in the first sample set to obtain a second sample set; wherein the second sample set includes a test sample set and a training sample set; Using the training sample set, the current round of training is performed on the basic first EEG signal classification model, the basic second EEG signal classification model, the basic third EEG signal classification model and the basic fourth EEG signal classification model in the basic EEG signal classification model to obtain the EEG signal classification model after the current round of training based on the training sample set; wherein the EEG signal classification model after the current round of training based on the training sample set includes: the first EEG signal classification model after the current round of training based on the training sample set, the second EEG signal classification model after the current round of training based on the training sample set, the third EEG signal classification model after the current round of training based on the training sample set and the fourth EEG signal classification model after the current round of training based on the training sample set; The test sample set is tested using the EEG signal classification model trained based on the training sample set in the current round to obtain the classification accuracy of the EEG signal classification model trained based on the training sample set in the current round; Obtaining the average classification accuracy of the classification accuracy of the EEG signal classification model trained in the current round based on each training sample set in different training sample sets as the first average classification accuracy of the EEG signal classification model after the current round of training; Obtain the average classification accuracy of the first average classification accuracy of the EEG signal classification model after multiple rounds of training as the second average classification accuracy; When the second average classification accuracy rate does not reach the preset classification accuracy rate, the model parameters of the EEG signal classification model after the last round of training are adjusted.

5. The method according to claim 4, characterized in that The method of using the training sample set to perform a current round of training on a basic first EEG signal classification model, a basic second EEG signal classification model, a basic third EEG signal classification model, and a basic fourth EEG signal classification model in the basic EEG signal classification model to obtain the current round of EEG signal classification model trained based on the training sample set includes: Performing feature selection on each training sample in the training sample set to obtain a feature matrix; Using each training sample in the training sample set for each feature in the feature matrix, respectively perform a current round of training on the basic first EEG signal classification model, the basic second EEG signal classification model, and the basic third EEG signal classification model to obtain a current round of first EEG signal classification model trained based on the training sample set, a current round of second EEG signal classification model trained based on the training sample set, and a current round of third EEG signal classification model trained based on the training sample set; For each training sample in the training sample set, the time domain features of the training sample are processed using the first EEG signal classification model trained based on the training sample set in the current round to obtain a first classification result of the training sample, the frequency domain features of the training sample are processed using the second EEG signal classification model trained based on the training sample set in the current round to obtain a second classification result of the training sample, and the topological features of the training sample are processed using the third EEG signal classification model trained based on the training sample set in the current round to obtain a third classification result of the training sample; The basic fourth EEG signal classification model is trained by utilizing the first classification result of each training sample in the training sample set and the classification accuracy of the first EEG signal classification model trained based on the training sample set in the current round, the second classification result of the training sample and the classification accuracy of the second EEG signal classification model trained based on the training sample set in the current round, and the third classification result of the training sample and the classification accuracy of the third EEG signal classification model trained based on the training sample set in the current round, to obtain the fourth EEG signal classification model trained based on the training sample set in the current round.

6. The method according to claim 5, characterized in that The step of performing feature selection on each training sample in the training sample set to obtain a feature matrix includes: Standardizing the feature information of each training sample in the training sample set to obtain first feature information; Performing a preset transformation on the first characteristic information to obtain second characteristic information; Obtaining a correlation score between the second feature information of each dimension and the label information of the sample EEG signal in the training sample set; The feature matrix is ​​constructed according to the second feature information whose correlation score is greater than the preset correlation score.

7. An electroencephalogram signal classification device, characterized in that: The device comprises: A first acquisition module is used to acquire feature information of the EEG signal to be classified; wherein the feature information is used to reflect the signal type of the EEG signal to be classified, and the feature information includes time domain features, frequency domain features and topological features; A second acquisition module is used to acquire a trained EEG signal classification model; wherein the trained EEG signal classification model includes a trained first EEG signal classification model, a trained second EEG signal classification model, a trained third EEG signal classification model and a trained fourth EEG signal classification model; A processing module, used to process the time domain features using the trained first EEG signal classification model to obtain a first classification result of the EEG signal to be classified, to process the frequency domain features using the trained second EEG signal classification model to obtain a second classification result of the EEG signal to be classified, and to process the topological features using the trained third EEG signal classification model to obtain a third classification result of the EEG signal to be classified; The processing module is further used to use the trained fourth EEG signal classification model to process the classification accuracy of the trained first EEG signal classification model and the first classification result, the classification accuracy of the trained second EEG signal classification model and the second classification result, and the classification accuracy of the trained third EEG signal classification model and the third classification result to obtain the target classification result of the EEG signal to be classified; The first acquisition module is used to construct a graph structure of the EEG signal to be classified by taking the signal strength corresponding to each EEG signal in the EEG signal to be classified as a node, taking the connection between any two nodes as an edge, and taking the slope of the edge as the weight corresponding to the edge; and obtaining the topological features of the EEG signal to be classified according to the graph structure of the EEG signal to be classified; wherein there is a node c between node a and node b, and node a, node b and node c satisfy: , In the case of , it is determined that there is an edge between node a and node b, where ( )middle is the moment when the EEG signal corresponding to node a is in the EEG signal to be classified, is the signal strength of the EEG signal, middle is the moment when the EEG signal corresponding to node b is in the EEG signal to be classified, is the signal strength of the EEG signal, middle is the moment when the EEG signal corresponding to node c is in the EEG signal to be classified, is the signal strength of the EEG signal, The preset time window.

8. An electronic device, characterized in that: include: Memory for storing executable computer instructions; A processor is used to execute the EEG signal classification method according to any one of claims 1 to 6 under the control of the executable computer instructions.

9. A computer-readable storage medium having computer instructions stored thereon, wherein the computer instructions are executed by a processor to execute the EEG signal classification method according to any one of claims 1 to 6.

Images