A method and system for EEG identity recognition based on graph neural network

Abstract

The present invention discloses an EEG identity recognition method and system based on a graph neural network. The method of the present invention comprises collecting EEG signals when a user performs an identity recognition operation; extracting graph feature data from the EEG signals; and inputting the graph feature data into a pre-trained graph neural network to obtain an identity recognition result. The EEG identity authentication method based on a graph neural network of the present invention can combine the temporal and spatial features of EEG signals, and by constructing the form of user graph data, while retaining the single-channel features of EEG, introduces rich EEG data graph topological structure features, can better model the EEG signals of different individuals, reduce the overfitting problem of traditional classification, improve classification accuracy, and have better interpretability.

Description

Technical Field

[0001] The present invention relates to identity recognition technology, and in particular to an EEG identity recognition method and system based on graph neural network. Background Art

[0002] As people's demand for security increases, biometric-based identity authentication systems have attracted much attention. EEG signals have been widely studied by researchers due to their many advantages such as high reliability, confidentiality and non-replicability. At present, many EEG-based identity authentication methods have been proposed. However, most of the classification methods used in these methods process EEG signals as a multi-channel time series, and do not utilize the spatial features and topological relationships that naturally exist between EEG signal channels. Therefore, it is easy for the classifier to learn false features and thus produce overfitting problems. Summary of the invention

[0003] Technical problem to be solved by the present invention: In view of the above-mentioned problems of the prior art, a method and system for EEG identity authentication based on graph neural network are provided. The present invention can realize EEG identity authentication based on graph neural network, and can take into account the spatial characteristics of each channel in the EEG signal, effectively reduce the problem of overfitting of the classifier, improve the classification accuracy, and have better interpretability.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A method for EEG identity recognition based on graph neural network, comprising:

[0006] 1) Collecting EEG signals of users when performing identity recognition operations;

[0007] 2) extracting graph feature data from the EEG signal;

[0008] 3) Inputting the graph feature data into a pre-trained graph neural network to obtain an identity recognition result.

[0009] Optionally, the EEG signal when the user performs the identity recognition operation in step 1) is an EEG signal generated by resting state or steady-state visual evoked potential stimulation.

[0010] Optionally, step 2) includes:

[0011] 2.1) performing noise reduction preprocessing on the EEG signal;

[0012] 2.2) The channels in the EEG signal after noise reduction preprocessing are defined as nodes, and the node features and structural features of the EEG signal are extracted according to the channels, wherein the node features are the signal features of a single channel, and the structural features are indicators of functional connectivity.

[0013] Optionally, step 2.1) includes:

[0014] 2.1.1) downsampling the EEG signal;

[0015] 2.1.2) The downsampled EEG signal is subjected to high-pass and low-pass filtering using a Butterworth filter, and artifacts including electromyography and head movement signals are removed;

[0016] 2.1.3) Perform nose tip re-reference and baseline correction on the EEG signal after removing the artifacts;

[0017] 2.1.4) The EEG signals after nose tip re-reference and baseline correction are extracted using principal component analysis to obtain pre-processed EEG signals, and the EEG signals are divided into small segments of specified duration to expand the data set.

[0018] Optionally, the cutoff frequency of the high-pass filter in step 2.1.2) is 0.1 Hz, and the cutoff frequency of the low-pass filter is 30 Hz; the specified time length in step 2.1.4) is 2 s.

[0019] Optionally, the signal feature in step 2.2) is power spectral density PSD, differential entropy DE or wavelet feature, and the indicator of functional connection is coherence or phase synchronization.

[0020] Optionally, the graph neural network in step 3) includes a first graph convolution layer, a pooling layer, a second graph convolution layer, a graph readout operation layer and a classification network connected in sequence. After the graph feature data is input into a pre-trained graph neural network, the identity recognition result is finally output by the classification network.

[0021] Optionally, step 3) also includes the step of training the graph neural network:

[0022] S1) collecting EEG signal samples when the user performs identity recognition operations and attaching labels;

[0023] S2) extracting graph feature data samples from the EEG signal;

[0024] S3) constructing training data with the graph feature data samples and their additional labels, and training the graph neural network with the training data to obtain a trained graph neural network.

[0025] In addition, the present invention also provides an EEG identity recognition system based on a graph neural network, comprising an interconnected microprocessor and a memory, wherein the microprocessor is programmed or configured to execute the steps of the EEG identity recognition method based on a graph neural network.

[0026] In addition, the present invention also provides a computer-readable storage medium, which stores a computer program that is executed by a computer device to implement the graph neural network-based EEG identity recognition method.

[0027] Compared with the prior art, the present invention mainly has the following advantages: Graph neural network (also called graph convolutional neural network, abbreviated as GCN) is a classifier that can perform deep learning on graph data. GCN is a generalization of convolutional neural network in the graph domain, which can simultaneously perform end-to-end learning on the node information and structural information of the graph, and is currently the best choice for graph data learning tasks. EEG data is a typical graph data. Applying GCN to the classification learning of EEG data can take into account the temporal characteristics and spatial characteristics of each channel in the EEG signal. The EEG identity authentication method based on graph neural network of the present invention can combine the temporal characteristics and spatial characteristics of EEG signals, and by constructing the form of user graph data, while retaining the single-channel characteristics of EEG, introduces rich EEG data graph topological structure characteristics, which can better model the EEG signals of different individuals, reduce the overfitting problem of traditional classification, improve classification accuracy, and have better interpretability. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 Schematic diagram of the basic flow of the method of the embodiment of the present invention.

[0029] Figure 2 Schematic diagram of screen graphics for collecting EEG signals in an embodiment of the present invention.

[0030] Figure 3 Schematic diagram of the structure of the neural network in the embodiment of the present invention. DETAILED DESCRIPTION

[0031] like Figure 1 As shown, the EEG identity authentication method based on graph neural network in this embodiment includes:

[0032] 1) Collecting EEG signals of users when performing identity recognition operations;

[0033] 2) Extracting graph feature data from EEG signals;

[0034] 3) Input the graph feature data into the pre-trained graph neural network to obtain the identity recognition result. Graph neural network (GCN), also known as graph neural network, is a generalization of convolutional neural network in the graph domain. It can perform deep learning on graph data and can simultaneously perform end-to-end learning on the node information and structure information of the graph. It is currently the best choice for graph data learning tasks. In this embodiment, EEG data constitutes typical graph data. Applying graph neural network to the classification learning of EEG data can take into account the spatial characteristics of each channel, effectively reduce the problem of overfitting of the classifier, improve the classification accuracy, and have better interpretability.

[0035] Among them, the EEG signal when the user performs the identity recognition operation in step 1) is the EEG signal generated by the resting state or steady-state visual evoked potential stimulation. For example, as an optional implementation, in this embodiment, the EEG signal when the user performs the identity recognition operation in step 1) is the EEG signal generated by the steady-state visual evoked potential stimulation, and the screen graphics are as follows Figure 2 As shown, the stimulation frequency was 7 Hz and the stimulation lasted for 10 s.

[0036] In this embodiment, step 2) includes:

[0037] 2.1) performing noise reduction preprocessing on the EEG signal;

[0038] 2.2) The channels in the EEG signal after noise reduction preprocessing are defined as nodes, and the node features and structural features of the EEG signal are extracted by channel, where the node features are the signal features of a single channel and the structural features are the indicators of functional connectivity.

[0039] In this embodiment, step 2.1) includes:

[0040] 2.1.1) downsampling the EEG signal;

[0041] 2.1.2) The downsampled EEG signal is subjected to high-pass and low-pass filtering using a Butterworth filter, and artifacts including electromyography and head movement signals are removed;

[0042] 2.1.3) Perform nose tip re-reference and baseline correction on the EEG signal after removing the artifacts;

[0043] 2.1.4) The EEG signals after nose tip re-reference and baseline correction are extracted using principal component analysis to obtain pre-processed EEG signals, and the EEG signals are divided into small segments of specified duration to expand the data set.

[0044] In this embodiment, steps 2.1.1) to 2.1.4) are completed using EEGLAB software. The parameters of high-pass filtering and low-pass filtering can be selected according to actual needs. For example, in step 2.1.2) of this embodiment, the cutoff frequency of the high-pass filtering is 0.1 Hz, and the cutoff frequency of the low-pass filtering is 30 Hz; the specified time length in step 2.1.4) is 2 s.

[0045] Among them, the signal feature in step 2.2) may be power spectral density PSD, differential entropy DE or wavelet feature, and the indicator of functional connectivity may be coherence or phase synchronization. For example, as an optional implementation, the signal feature in step 2.2) of this embodiment is power spectral density PSD, and the indicator of functional connectivity is coherence (COH). It should be noted that power spectral density PSD, differential entropy DE or wavelet feature, coherence or phase synchronization are all existing features or indicators, so their specific extraction or calculation methods are no longer described in detail.

[0046] like Figure 3 As shown, the graph neural network in step 3) of this embodiment includes a first graph convolution layer, a pooling layer, a second graph convolution layer, a graph readout operation (Readout) layer and a classification network connected in sequence. After the graph feature data is input into the pre-trained graph neural network, the identity recognition result is finally output by the classification network. Figure 3 The graph neural network mainly includes two graph convolution layers. The first graph convolution layer is connected to the pooling layer, which is used to coarsen the graph into subgraphs. The second graph convolution layer is connected to the graph readout operation (Readout) layer, and the final graph representation is obtained using the sum / mean (sum / average) subgraph. Then, a classification network with classification capability is connected. The classification network can be a multi-layer perceptron MLP or a support vector machine SVM, etc. For example, as an optional implementation, the classification network in this embodiment specifically adopts a support vector machine SVM.

[0047] In this embodiment, before step 3), the step of training the graph neural network is also included:

[0048] S1) collecting EEG signal samples when the user performs identity recognition operations and attaching labels;

[0049] S2) extracting graph feature data samples from the EEG signal;

[0050] S3) constructing training data from the graph feature data samples and their additional labels, and training the graph neural network with the training data to obtain a trained graph neural network. Steps S1) to S2) are the same as the methods of steps 1) to 2) above, so they will not be repeated here.

[0051] In addition, this embodiment also provides an EEG identity recognition system based on a graph neural network, including a microprocessor and a memory connected to each other, and the microprocessor is programmed or configured to execute the steps of the EEG identity recognition method based on a graph neural network.

[0052] In addition, this embodiment also provides a computer-readable storage medium, which stores a computer program that is executed by a computer device to implement the graph neural network-based EEG identity recognition method.

[0053] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The present application is described 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 application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the functions in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0054] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for EEG identity recognition based on graph neural network, characterized in that: include: 1) Collect EEG signals when the user performs identity recognition operations; 2) extracting graph feature data from the EEG signal, including: 2.1) performing noise reduction preprocessing on the EEG signal; 2.2) defining channels in the EEG signal after noise reduction preprocessing as nodes, and extracting node features and structural features of the EEG signal according to the channels, wherein the node feature is the signal feature of a single channel, and the structural feature is an indicator of functional connection; the signal feature is power spectral density PSD, differential entropy DE or wavelet feature, and the indicator of functional connection is coherence or phase synchronization; 3) Inputting the graph feature data into a pre-trained graph neural network to obtain an identity recognition result; the graph neural network includes a first graph convolution layer, a pooling layer, a second graph convolution layer, a graph readout operation layer and a classification network connected in sequence. After inputting the graph feature data into the pre-trained graph neural network, the identity recognition result is finally output by the classification network.

2. The method for EEG identity recognition based on graph neural network according to claim 1 is characterized in that: In step 1), the EEG signal generated when the user performs the identity recognition operation is an EEG signal generated by resting state or steady-state visual evoked potential stimulation.

3. The method for EEG identity recognition based on graph neural network according to claim 2 is characterized in that: Step 2.1) includes: 2.1.1) Downsampling the EEG signal; 2.1.2) The downsampled EEG signal is subjected to high-pass and low-pass filtering using a Butterworth filter, and artifacts including electromyography and head movement signals are removed; 2.1.3) Perform nose tip re-reference and baseline correction on the EEG signals after removing artifacts; 2.1.4) The EEG signals after nose tip re-reference and baseline correction are extracted using principal component analysis to obtain the pre-processed EEG signals, and the EEG signals are divided into small segments of specified duration to expand the data set.

4. The method for EEG identity recognition based on graph neural network according to claim 3 is characterized in that: The cut-off frequency of the high-pass filter in step 2.1.2) is 0.1 Hz, and the cut-off frequency of the low-pass filter is 30 Hz; the specified time length in step 2.1.4) is 2 s.

5. The method for EEG identity recognition based on graph neural network according to claim 1, characterized in that: Step 3) also includes the steps of training the graph neural network: S1) Collect EEG signal samples when the user performs identity recognition operations and add labels; S2) extracting graph feature data samples from the EEG signal; S3) constructing training data with the graph feature data samples and their additional labels, and training the graph neural network with the training data to obtain a trained graph neural network.

6. An EEG identity recognition system based on graph neural network, comprising a microprocessor and a memory connected to each other, characterized in that: The microprocessor is programmed or configured to execute the steps of the graph neural network-based EEG identity recognition method described in any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program that is executed by a computer device to implement the EEG identity recognition method based on a graph neural network as described in any one of claims 1 to 5.

Images