CN114469139A

CN114469139A - Electroencephalogram signal recognition model training method, electroencephalogram signal recognition device and medium

Info

Publication number: CN114469139A
Application number: CN202210097544.9A
Authority: CN
Inventors: 胡晨; 陈屹
Original assignee: Agricultural Bank of China
Current assignee: Agricultural Bank of China
Priority date: 2022-01-27
Filing date: 2022-01-27
Publication date: 2022-05-13

Abstract

The embodiment of the invention discloses an electroencephalogram signal recognition model training method, an electroencephalogram signal recognition device and a medium. The electroencephalogram signal recognition model training method specifically comprises the following steps: acquiring electroencephalogram sample data of a target user and electrode position image sample data of the electroencephalogram sample data; generating electroencephalogram signal identification model input sample data according to the electroencephalogram signal sample data and the electrode position image sample data; the electroencephalogram signal identification model input sample data comprises an electroencephalogram signal time domain position diagram and an electroencephalogram signal frequency domain position diagram; and training an electroencephalogram signal identification model according to the electroencephalogram signal identification model input sample data to obtain a target electroencephalogram signal identification model matched with the target user. The technical scheme of the embodiment of the invention can improve the identification efficiency and the identification accuracy of the electroencephalogram signals.

Description

Electroencephalogram signal recognition model training method, electroencephalogram signal recognition device and medium

Technical Field

The embodiment of the invention relates to the technical field of computers, in particular to a method, a device and a medium for training an electroencephalogram signal identification model and identifying an electroencephalogram signal.

Background

The nervous system is one of three major physiological systems affecting human thinking and behavior, and is composed of a large number of neurons. Neurons are cells of the brain that are capable of generating, propagating, and processing electrical signals. Neurons are connected to other neurons to form functional networks, and the brain can be viewed as a collection of interacting neural networks. The brain electrical signals are spontaneous electrical potential activities that result from these nerve cell activities and are consistently present in the central nervous system. Electroencephalography records electroencephalogram signals by measuring changes in voltage on the scalp caused by cortical activity. Because brain electrical signals are highly dimensional and unstable, and are often mixed with various environmental noises, identifying brain electrical signals is a challenging task.

At present, with the continuous efforts of researchers, many methods for identifying electroencephalogram signals exist. The electroencephalogram signal identification method based on traditional machine learning generally divides feature extraction and feature classification into two steps for processing, but the electroencephalogram signal identification accuracy of the method is poor. With the continuous research of the deep learning method, the electroencephalogram signal identification method based on the deep learning is widely applied, the electroencephalogram signal is generally directly input into an electroencephalogram signal identification model for identification, but the electroencephalogram signal generally contains a lot of useless information, so that the electroencephalogram signal cannot be quickly and accurately identified by the method. Therefore, the existing electroencephalogram signal identification method is low in identification efficiency and poor in identification accuracy.

Disclosure of Invention

The embodiment of the invention provides an electroencephalogram signal recognition model training method, an electroencephalogram signal recognition device, an electroencephalogram signal recognition equipment and a medium, and can improve the electroencephalogram signal recognition efficiency and the electroencephalogram signal recognition accuracy.

According to an aspect of the present invention, there is provided an electroencephalogram signal recognition model training method, including:

acquiring electroencephalogram sample data of a target user and electrode position image sample data of the electroencephalogram sample data;

generating electroencephalogram signal identification model input sample data according to the electroencephalogram signal sample data and the electrode position image sample data; the electroencephalogram signal identification model input sample data comprises an electroencephalogram signal time domain position diagram and an electroencephalogram signal frequency domain position diagram;

and training an electroencephalogram signal identification model according to the electroencephalogram signal identification model input sample data to obtain a target electroencephalogram signal identification model matched with the target user.

According to another aspect of the present invention, there is provided an electroencephalogram signal identification method, including:

acquiring an electroencephalogram signal to be identified of a target user and electrode position image data of the electroencephalogram signal to be identified;

generating electroencephalogram signal identification model input data according to the electroencephalogram signal to be identified and the electrode position image data; the electroencephalogram signal identification model input data comprise an electroencephalogram signal time domain position diagram to be identified and an electroencephalogram signal frequency domain position diagram to be identified;

and inputting the EEG signal identification model input data into a target EEG signal identification model matched with the target user so as to identify the user identity of the target user through the target EEG signal identification model.

According to another aspect of the present invention, there is provided an electroencephalogram signal recognition model training apparatus, including:

the system comprises a sample data acquisition module, a data acquisition module and a data acquisition module, wherein the sample data acquisition module is used for acquiring electroencephalogram signal sample data of a target user and electrode position image sample data of the electroencephalogram signal sample data;

the model input sample data generating module is used for generating electroencephalogram signal identification model input sample data according to the electroencephalogram signal sample data and the electrode position image sample data; the electroencephalogram signal identification model input sample data comprises an electroencephalogram signal time domain position diagram and an electroencephalogram signal frequency domain position diagram;

and the target electroencephalogram signal identification model obtaining module is used for training an electroencephalogram signal identification model according to the electroencephalogram signal identification model input sample data to obtain a target electroencephalogram signal identification model matched with the target user.

According to another aspect of the present invention, there is provided an electroencephalogram signal identification apparatus, including:

the image acquisition module is used for acquiring the electroencephalogram signals to be identified of a target user and the electrode position image data of the electroencephalogram signals to be identified;

the electroencephalogram signal identification model input data generation module is used for generating electroencephalogram signal identification model input data according to the electroencephalogram signal to be identified and the electrode position image data; the electroencephalogram signal identification model input data comprise an electroencephalogram signal time domain position diagram to be identified and an electroencephalogram signal frequency domain position diagram to be identified;

and the user identity recognition module is used for inputting the electroencephalogram signal recognition model input data to the target electroencephalogram signal recognition model matched with the target user so as to recognize the user identity of the target user through the target electroencephalogram signal recognition model.

According to another aspect of the present invention, there is provided an electronic apparatus including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores a computer program executable by the at least one processor, the computer program being executable by the at least one processor to enable the at least one processor to perform the electroencephalogram recognition model training method according to any one of the embodiments of the present invention, or to perform the electroencephalogram recognition method according to any one of the embodiments of the present invention.

According to another aspect of the present invention, a computer-readable storage medium is provided, which stores computer instructions for enabling a processor to implement the electroencephalogram signal recognition model training method according to any one of the embodiments of the present invention, or implement the electroencephalogram signal recognition method according to any one of the embodiments of the present invention when the computer instructions are executed.

According to the technical scheme of the embodiment of the invention, the electroencephalogram signal sample data of the target user and the electrode position image sample data of the electroencephalogram signal sample data are obtained, the electroencephalogram signal identification model input sample data comprising an electroencephalogram signal time domain position map and an electroencephalogram signal frequency domain position map are generated according to the electroencephalogram signal sample data and the electrode position image sample data, the electroencephalogram signal identification model is trained according to the electroencephalogram signal identification model input sample data, so that the target electroencephalogram signal identification model matched with the target user is obtained, the user identity of the target user is identified by using the successfully trained target electroencephalogram signal identification model, the problems of low identification efficiency and poor identification accuracy of the existing electroencephalogram signal identification method are solved, and the identification efficiency and the identification accuracy of the electroencephalogram signal can be improved.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present invention, nor do they necessarily limit the scope of the invention. Other features of the present invention will become apparent from the following description.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a flowchart of a method for training an electroencephalogram signal recognition model according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of electrode position image sample data according to an embodiment of the present invention;

FIG. 3 is a flowchart illustrating an example of a method for training an electroencephalogram signal recognition model according to an embodiment of the present invention;

FIG. 4 is a flowchart of an electroencephalogram signal identification method according to a second embodiment of the present invention;

FIG. 5 is a schematic diagram of an electroencephalogram signal recognition model training device provided by a third embodiment of the present invention;

FIG. 6 is a schematic diagram of an electroencephalogram signal identification device according to a fourth embodiment of the present invention;

fig. 7 is a schematic structural diagram of an electronic device according to a fifth embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention.

It should be further noted that, for the convenience of description, only some but not all of the relevant aspects of the present invention are shown in the drawings. Before discussing exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although a flowchart may describe the operations (or steps) as a sequential process, many of the operations can be performed in parallel, concurrently or simultaneously. In addition, the order of the operations may be re-arranged. The process may be terminated when its operations are completed, but may have additional steps not included in the figure. The processes may correspond to methods, functions, procedures, subroutines, and the like.

The terms "first" and "second," and the like in the description and claims of embodiments of the invention and in the drawings, are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "comprising" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not set forth for a listed step or element but may include steps or elements not listed.

Example one

Fig. 1 is a flowchart of an electroencephalogram signal recognition model training method according to an embodiment of the present invention, which is applicable to a case where electroencephalogram signal recognition efficiency and accuracy are improved, and the method may be implemented by an electroencephalogram signal recognition model training apparatus, the apparatus may be implemented in a software and/or hardware manner, and may generally be directly integrated in an electronic device that executes the method, where the electronic device may be a terminal device or a server device, and the embodiment of the present invention does not limit the type of the electronic device that executes the electroencephalogram signal recognition model training method. Specifically, as shown in fig. 1, the electroencephalogram signal recognition model training method may specifically include the following steps:

s110, acquiring electroencephalogram sample data of a target user and electrode position image sample data of the electroencephalogram sample data.

The target user may be any user, for example, a user handling banking business, a user performing traffic safety inspection, or a user performing inspection in cooperation with related departments, and the like, which is not limited in this embodiment of the present invention. It can be understood that the user identity of the target user can be identified by identifying the electroencephalogram signal of the target user. The brain electrical signal sample data may be brain electrical signal data of the target user that can be used as a sample. The electrode position image sample data may be position image data of brain electrical electrodes that can be used as a sample.

In the embodiment of the invention, electroencephalogram signal sample data of a target user and electrode position image sample data of the electroencephalogram signal sample data are obtained, and electroencephalogram signal identification model input sample data are generated according to the electroencephalogram signal sample data and the electrode position image sample data. Specifically, the acquiring of the electroencephalogram signal sample data of the target user may be acquired by an electroencephalogram signal acquisition device. It can be understood that, the electroencephalogram signal sample data of the target user is acquired through the electroencephalogram signal acquisition equipment, the electroencephalogram signal sample data can be acquired through electrodes on the electroencephalogram signal acquisition equipment, different electrodes on the electroencephalogram signal acquisition equipment can correspond to different brain areas, namely, different electrodes can read signals of different brain areas, the position of the electroencephalogram electrode can be used for representing the position of the brain areas, and each electrode can be adjacent to a plurality of electrodes physically.

Optionally, acquiring electroencephalogram sample data of the target user may include: displaying a preset display image to a target user; acquiring electroencephalogram sample data generated when a target user watches a preset display image.

The preset display image can be any preset image displayed to the target user and can be used for displaying the preset image to the target user to acquire electroencephalogram sample data of the target user.

Specifically, the preset display image is displayed to the target user so as to obtain electroencephalogram signal sample data generated when the target user watches the preset display image. Optionally, the preset display image may include one target user face image and nine other arbitrary images. When the preset display image is displayed to the target user, each image in the preset display image may be randomly displayed, and the display time of each image may be 300 milliseconds. It will be appreciated that by analogy with the blind source classification method, brain electrical signal sample data can be thought of as multi-channel "speech" signals obtained from several "microphones" (associated with brain electrical electrodes) that record signals from a plurality of "speakers" (corresponding to the activity of cortical regions). That is, the electroencephalogram sample data may be a series of one-dimensional vectors, each of which may represent an electrode reading at a certain time.

Optionally, the electrode position image sample data for acquiring the electroencephalogram signal sample data may be a two-dimensional image. In a specific example, fig. 2 is a schematic diagram of electrode position image sample data provided by an embodiment of the present invention, electroencephalogram electrodes are distributed in a three-dimensional space of a scalp, and electrode positions can be projected from the three-dimensional space to a two-dimensional surface by isometric Projection (AEP) to obtain the electrode position image sample data of a two-dimensional image as shown in fig. 2. It will be appreciated that the shape of the electrode cap worn on a person's head may be approximated by a sphere, and that the projection of the electrode position on a 2D surface tangential to the apex of the head may be determined by the same method. By projecting the three-dimensional space to the two-dimensional image by using the method of isometric projection, the distance from the projection center to any other point can be kept unchanged, namely the relative distance between adjacent electrodes can be kept.

S120, generating electroencephalogram signal identification model input sample data according to the electroencephalogram signal sample data and the electrode position image sample data; the electroencephalogram signal identification model input sample data comprises an electroencephalogram signal time domain position diagram and an electroencephalogram signal frequency domain position diagram.

The input sample data of the electroencephalogram signal identification model may be sample data input to the electroencephalogram signal identification model. The electroencephalogram signal time domain position map can be an image obtained by combining a time domain value in electroencephalogram signal sample data and electrode position image sample data. The electroencephalogram signal frequency domain position map can be an image obtained by combining a frequency domain value in electroencephalogram signal sample data with electrode position image sample data.

In the embodiment of the invention, after the electroencephalogram signal sample data of the target user and the electrode position image sample data of the electroencephalogram signal sample data are obtained, the electroencephalogram signal identification model input sample data can be further generated according to the electroencephalogram signal sample data and the electrode position image sample data. Specifically, the input sample data of the electroencephalogram signal identification model may include an electroencephalogram signal time domain position map and an electroencephalogram signal frequency domain position map. The electroencephalogram identification method has the advantages that the electroencephalogram signals not only change along with time, but also the frequency and the phase of the electroencephalogram signals contain rich information, so that an electroencephalogram signal frequency domain position map can be generated according to electroencephalogram signal sample data and electrode position image sample data, the electroencephalogram signal frequency domain position map is input into an electroencephalogram signal identification model, and therefore the identification accuracy of the electroencephalogram signals is improved.

Optionally, before generating an electroencephalogram signal identification model input sample data according to the electroencephalogram signal sample data and the electrode position image sample data, noise reduction processing may be performed on the electroencephalogram signal sample data.

It can be understood that the electroencephalogram signal is easily polluted by various interferences in the acquisition process, such as power frequency interference generated by the electroencephalogram signal acquisition equipment and artifact signals generated by blinking, muscle activity and the like. The noise inevitably interferes with the characteristics of the electroencephalogram signal to influence the identification accuracy. Therefore, noise reduction processing is required after electroencephalogram signal sample data are acquired, so that the quality of electroencephalogram signals is improved, and the signal-to-noise ratio of the electroencephalogram signals is enhanced.

Optionally, generating an electroencephalogram time-domain position map according to the electroencephalogram signal sample data and the electrode position image sample data may include: determining a target electrode in the electrode position image sample data and a target electrode position of the target electrode in the electrode position image sample data; determining target electrode time signal data of a target electrode according to the electroencephalogram signal sample data; generating a plurality of continuous initial electroencephalogram signal time-domain position graphs according to the target electrode time signal data and the target electrode position; performing image supplement processing on the initial electroencephalogram signal time-domain position map by a linear interpolation method to obtain an electroencephalogram signal time-domain position map; and taking the time domain position map of each electroencephalogram signal moment in a preset time period as the time domain position map of the electroencephalogram signal.

The target electrode can be any electrode capable of acquiring electroencephalogram signal sample data. The target electrode position may be a position of the target electrode in the electrode position image sample data. The preset time period may be a preset time period. It can be understood that the electroencephalogram signal sample data may be an electroencephalogram signal within a preset time period. The target electrode time signal data may be electroencephalogram signal data of the target electrode at a specific time within a preset time period, that is, electrode reading of the target electrode at a specific time within a preset time period. The initial electroencephalogram time-domain position map can be an initial image of the electroencephalogram time-domain position map at a specific time within a preset time period. The electroencephalogram time-domain position map can be a time-domain position map of an electroencephalogram at a specific time within a preset time period.

Specifically, after acquiring electroencephalogram signal sample data of a target user and electrode position image sample data of the electroencephalogram signal sample data, the target electrode in the electrode position image sample data and the target electrode position of the target electrode in the electrode position image sample data can be further determined, target electrode time signal data of the target electrode is determined according to the electroencephalogram signal sample data, a plurality of continuous initial electroencephalogram signal time domain position maps are generated according to the target electrode time signal data and the target electrode position, image supplement processing is performed on the initial electroencephalogram signal time domain position maps through a linear interpolation method, an electroencephalogram signal time domain position map is obtained, and therefore each electroencephalogram signal time domain position map in preset time is used as an electroencephalogram signal time domain position map. Optionally, a plurality of continuous initial electroencephalogram time-domain position maps are generated according to the target electrode time signal data and the target electrode position, or a plurality of continuous initial electroencephalogram time-domain position maps are generated by filling the target electrode position into the target electrode time signal data.

According to the technical scheme, the initial electroencephalogram signal time-domain position map is subjected to image supplement processing through a linear interpolation method, the condition that the initial electroencephalogram signal time-domain position map is too sparse to influence the identification efficiency of the electroencephalogram signal identification model can be prevented, and therefore the identification efficiency of the electroencephalogram signal identification model is ensured.

Optionally, generating a electroencephalogram signal frequency domain position map according to the electroencephalogram signal sample data and the electrode position image sample data, which may include: determining a target electrode in the electrode position image sample data and a target electrode position of the target electrode in the electrode position image sample data; determining target electrode time signal data of a target electrode according to the electroencephalogram signal sample data; generating target electrode time frequency domain signal data of a target electrode according to the target electrode time signal data through discrete short-time Fourier transform; generating a plurality of continuous electroencephalogram signal time frequency domain position graphs according to the target electrode time frequency domain signal data and the target electrode position; and taking the time frequency domain position map of each electroencephalogram signal in a preset time period as the frequency domain position map of the electroencephalogram signal.

The target electrode time frequency domain signal data may be frequency domain data of an electroencephalogram signal of the target electrode at a specific time within a preset time period. The electroencephalogram signal time frequency domain position map can be a frequency domain position map of the electroencephalogram signal at a specific time within a preset time period.

Specifically, after acquiring electroencephalogram signal sample data of a target user and electrode position image sample data of the electroencephalogram signal sample data, the target electrode in the electrode position image sample data and the target electrode position of the target electrode in the electrode position image sample data can be further determined, target electrode time signal data of the target electrode is determined according to the electroencephalogram signal sample data, target electrode time frequency domain signal data of the target electrode is generated according to the target electrode time signal data through discrete short-time fourier transform, a plurality of continuous electroencephalogram signal time frequency domain position maps are generated according to the target electrode time frequency domain signal data and the target electrode position, and therefore each electroencephalogram signal time frequency domain position map in a preset time period is used as an electroencephalogram signal frequency domain position map.

Optionally, the target electrode time frequency domain signal data of the target electrode is generated according to the target electrode time signal data by discrete short-time fourier transform, and the target electrode time frequency domain signal data of the target electrode can be generated according to the target electrode time signal data by the following discrete short-time fourier transform formula:

wherein, T_iThe method is characterized in that the method represents the ith channel in the electroencephalogram signal, x represents a discrete value in a time domain, K represents a discrete value in a frequency domain, and L represents the number of electrode data points.

According to the technical scheme, the electroencephalogram signal time domain value and the electroencephalogram signal frequency domain value are converted into the electroencephalogram signal time domain position map and the electroencephalogram signal frequency domain position map of the two-dimensional image according to the position of the target electrode, so that the space structure of the electrode is reserved, and the time evolution of brain activity is represented through an image sequence in a preset time period.

S130, training an electroencephalogram signal recognition model according to the electroencephalogram signal recognition model input sample data to obtain a target electroencephalogram signal recognition model matched with the target user.

The electroencephalogram signal identification model can be a model for identifying electroencephalogram signals. The target brain electrical signal recognition model may be a target brain electrical signal recognition model that matches the target user. It can be understood that the electroencephalogram signal of each user is different, so that the model parameters of the electroencephalogram signal identification model obtained by inputting sample data according to the electroencephalogram signal identification model of different users for training are different, that is, the target electroencephalogram signal identification models matched with different target users are different.

In the embodiment of the invention, after the electroencephalogram signal identification model input sample data is generated according to the electroencephalogram signal sample data and the electrode position image sample data, the electroencephalogram signal identification model can be trained according to the electroencephalogram signal identification model input sample data to obtain the target electroencephalogram signal identification model matched with the target user.

Optionally, training the electroencephalogram signal recognition model according to the electroencephalogram signal recognition model input sample data may include: inputting the electroencephalogram signal time domain position map into a first convolution neural network of the electroencephalogram signal identification model, inputting the electroencephalogram signal frequency domain position map into a second convolution neural network of the electroencephalogram signal identification model, extracting electroencephalogram signal time domain characteristics of the electroencephalogram signal time domain position map through the first convolution neural network, and extracting electroencephalogram signal frequency domain characteristics of the electroencephalogram signal frequency domain position map through the second convolution neural network; performing feature splicing and fusion on the electroencephalogram signal time domain features and the electroencephalogram signal frequency domain features to obtain electroencephalogram signal fusion features; inputting the EEG signal fusion characteristics to a recurrent neural network of an EEG signal identification model to obtain EEG signal hidden layer information; classifying and identifying the EEG hidden layer information through a linear classifier of the EEG identification model to obtain a user identity identification result of the EEG identification model input sample data; and comparing the user identity recognition result with the user identity marking result of the sample data input by the electroencephalogram signal recognition model to determine the training effect of the electroencephalogram signal recognition model.

The first convolutional neural network can be a convolutional neural network in an electroencephalogram signal recognition model. The second convolutional neural network may be another convolutional neural network in the brain electrical signal recognition model. It will be appreciated that convolutional neural networks are a class of feed-forward neural networks that contain convolutional calculations and have a deep structure. It has three important features: local receptive field, weight sharing and down sampling. The local receptive field means that the connection of neurons between two layers of networks is local connection rather than full connection, and the structure can greatly reduce the number of weights and the complexity of a model, so that the networks are easier to converge. The weight sharing means that the connection weights among the neurons in the same layer are the same, and the number of the weights can be reduced because the weights are the same. The down-sampling is because the input is generally large, it is not necessary to analyze the original image in practice, the effective image feature is the key point, and the down-sampling process can be used to adjust the image size. The network structure of the convolutional neural network comprises a convolutional layer, a pooling layer and a full-link layer.

The electroencephalogram signal time domain features can be features extracted from an electroencephalogram signal time domain position diagram. The electroencephalogram signal frequency domain features may be features extracted in an electroencephalogram signal frequency domain location map. The electroencephalogram signal fusion characteristic can be a characteristic obtained by performing characteristic fusion on an electroencephalogram signal time domain characteristic and an electroencephalogram signal frequency domain characteristic. The EEG signal hidden layer information can be hidden state information output by a recurrent neural network of the EEG signal identification model. The user identification result may be a result obtained by identifying an electroencephalogram of the user. The user identity marking result may be a marking result of the user identity.

Specifically, after electroencephalogram signal identification model input sample data is generated according to electroencephalogram signal sample data and electrode position image sample data, an electroencephalogram signal time domain position map can be further input into a first convolution neural network of the electroencephalogram signal identification model, an electroencephalogram signal frequency domain position map is input into a second convolution neural network of the electroencephalogram signal identification model, electroencephalogram signal time domain characteristics of the electroencephalogram signal time domain position map are extracted through the first convolution neural network, electroencephalogram signal frequency domain characteristics of the electroencephalogram signal frequency domain position map are extracted through the second convolution neural network, characteristic splicing and fusion are carried out on the electroencephalogram signal time domain characteristics and the electroencephalogram signal frequency domain characteristics to obtain electroencephalogram signal fusion characteristics, the electroencephalogram signal fusion characteristics are input into the circulation neural network of the electroencephalogram signal identification model to obtain electroencephalogram signal hidden layer information, and the electroencephalogram signal hidden layer information is classified and identified through a linear classifier of the electroencephalogram signal identification model to obtain electroencephalogram signal hidden layer information The user identity recognition result of the sample data is input by the electric signal recognition model, so that the user identity recognition result is compared with the user identity marking result of the sample data input by the electroencephalogram signal recognition model, and the training effect of the electroencephalogram signal recognition model is further determined.

It can be understood that the loss value of the model can be calculated according to the user identity recognition result and the user identity marking result through the model loss function, so that the electroencephalogram signal recognition model can be trained according to the loss value condition. Alternatively, the cross entropy loss can be taken as a model loss function. If the loss value meets the training requirement, if the loss value tends to be stable or is always smaller than a set threshold value, the electroencephalogram signal recognition model successfully trained is indicated to be successfully trained, and the successfully trained electroencephalogram signal recognition model can be used as a target electroencephalogram signal recognition model so as to recognize the user identity of a target user through the target electroencephalogram signal recognition model.

Alternatively, the recurrent neural network may be a Long Short-term Memory network (LSTM). The EEG signal hidden layer information can be hidden state information output by the last layer of LSTM unit. The first convolutional layer of the first convolutional neural network and the first convolutional layer of the second convolutional neural network can both adopt a conditional parameter convolutional structure, and a compressive excitation module can be included between the first convolutional layer and the second convolutional layer.

It can be understood that the long-short term memory network is an improved recurrent neural network aiming at the problem that the gradient disappears or the gradient explodes when the traditional RNN trains, so that the farther sequence cannot be processed and the perception capability of the farther time is lost.

According to the technical scheme, the electroencephalogram signal identification model can have the capability of analyzing specific users by adopting a condition parameter convolution structure; by adding the extrusion excitation module between the first convolution layer and the second convolution layer, the electroencephalogram signal identification model can make full use of global information to comprehensively judge the information of a user, so that the identification accuracy of the electroencephalogram signal is improved.

In a specific example, fig. 3 is an example flowchart of a method for training an electroencephalogram signal recognition model according to an embodiment of the present invention, and as shown in fig. 3, the method specifically includes the following steps:

(1) determining an image sequence (namely a preset display image) corresponding to a target user; the image sequence may include one user face image and nine arbitrary images randomly selected from the image library. Acquiring electroencephalogram sample data when a target user watches an image sequence; wherein the images in the sequence of images are presented randomly in order and each image has a duration of 300 milliseconds.

(2) And projecting the electrode position from the three-dimensional space to the two-dimensional surface through isometric projection to acquire electrode position image sample data.

(3) And respectively converting the time domain value and the frequency domain value into two-dimensional images, namely an electroencephalogram signal time domain position map and an electroencephalogram signal frequency domain position map according to the electrode positions. Specifically, the electrode position is filled with electrode reading to obtain an electroencephalogram time domain position map, and a frequency domain signal is obtained through discrete short-time Fourier transform (DSTFT) to be filled with the electrode position to obtain an electroencephalogram frequency domain position map.

(4) Respectively inputting the EEG time domain position map and the EEG frequency domain position map into a convolutional neural network of an EEG recognition model for feature extraction to obtain EEG time domain features and EEG frequency domain features, fusing the EEG time domain features and the EEG frequency domain features and inputting the fused EEG time domain features and EEG frequency domain features into a long-short term memory network of the EEG recognition model, and classifying and recognizing the hidden layer information of the last layer of long-short term memory network unit through a linear layer (namely a linear classifier) of the EEG recognition model to obtain a classification result of the EEG.

(5) And training the electroencephalogram signal recognition model through a model loss function according to the classification result of the electroencephalogram signals. In particular, the cross entropy loss can be taken as a model loss function.

According to the technical scheme, the traditional one-dimensional vector electroencephalogram signal format is converted into the two-dimensional grid-shaped electroencephalogram signal hierarchical structure, the electrode reading and the electrode position information are mapped, and the correlation between the electroencephalogram signal and the corresponding brain area can be kept consistent.

According to the technical scheme, through acquiring the electroencephalogram signal sample data of the target user and the electrode position image sample data of the electroencephalogram signal sample data, generating the electroencephalogram signal identification model input sample data comprising an electroencephalogram signal time domain position map and an electroencephalogram signal frequency domain position map according to the electroencephalogram signal sample data and the electrode position image sample data, and training the electroencephalogram signal identification model according to the electroencephalogram signal identification model input sample data, the target electroencephalogram signal identification model matched with the target user is obtained, the problems of low identification efficiency and poor identification accuracy of an existing electroencephalogram signal identification method are solved, and the identification efficiency and the identification accuracy of the electroencephalogram signal can be improved.

Example two

Fig. 4 is a flowchart of an electroencephalogram signal identification method according to a second embodiment of the present invention, which is applicable to a case where electroencephalogram signal identification efficiency and accuracy are improved, and the method may be implemented by an electroencephalogram signal identification apparatus, and the apparatus may be implemented in a software and/or hardware manner, and may generally be directly integrated in an electronic device that executes the method, and the electronic device may be a terminal device or a server device. Specifically, as shown in fig. 4, the electroencephalogram signal identification method may specifically include the following steps:

s410, acquiring the electroencephalogram signal to be identified of the target user and the electrode position image data of the electroencephalogram signal to be identified.

The electroencephalogram signal to be identified can be an electroencephalogram signal to be identified. The electrode position image data may be position image data of brain electrical electrodes.

In the embodiment of the invention, the electroencephalogram signal to be identified of a target user and the electrode position image data of the electroencephalogram signal to be identified are obtained, so that electroencephalogram signal identification model input data are generated according to the electroencephalogram signal to be identified and the electrode position image data.

Optionally, acquiring the electroencephalogram signal to be identified of the target user may include: displaying a preset display image to a target user; acquiring the electroencephalogram signal to be recognized generated when a target user watches a preset display image.

Specifically, the electroencephalogram signal to be identified of the target user is obtained, and the preset display image can be displayed to the target user, so that the electroencephalogram signal to be identified, which is generated when the target user watches the preset display image, is obtained.

S420, generating electroencephalogram signal identification model input data according to the electroencephalogram signal to be identified and the electrode position image data; the electroencephalogram signal identification model input data comprise an electroencephalogram signal time domain position diagram to be identified and an electroencephalogram signal frequency domain position diagram to be identified.

The electroencephalogram signal identification model input data may be data input to the electroencephalogram signal identification model. The electroencephalogram signal time domain position map to be identified can be an image obtained by combining the time domain value in the electroencephalogram signal to be identified and the electrode position image data. The electroencephalogram signal frequency domain position map to be identified can be an image obtained by combining the frequency domain value in the electroencephalogram signal to be identified and the electrode position image data.

In the embodiment of the invention, after the electroencephalogram signal to be recognized of the target user and the electrode position image data of the electroencephalogram signal to be recognized are obtained, electroencephalogram signal recognition model input data can be further generated according to the electroencephalogram signal to be recognized and the electrode position image data. Specifically, the electroencephalogram signal identification model input data may include an electroencephalogram signal time domain position map to be identified and an electroencephalogram signal frequency domain position map to be identified.

Optionally, generating a time-domain position map of the electroencephalogram signal to be identified according to the electroencephalogram signal to be identified and the electrode position image data may include: determining a target electrode in the electrode position image data and a target electrode position of the target electrode in the electrode position image data; determining target electrode time to-be-identified signal data of a target electrode according to the electroencephalogram signal to be identified; generating a plurality of continuous time domain position graphs of initial electroencephalogram signals to be recognized according to the data of the signals to be recognized at the target electrode time and the position of the target electrode; performing image supplement processing on the time-domain position map of the electroencephalogram signal to be initially identified by a linear interpolation method to obtain the time-domain position map of the electroencephalogram signal to be identified; and taking the time domain position map of each electroencephalogram signal to be identified in a preset time period as the time domain position map of the electroencephalogram signal to be identified.

The signal data to be recognized at the target electrode moment can be electroencephalogram signal data to be recognized of the target electrode at a specific moment in a preset time period, namely, electrode reading of the target electrode at the specific moment in the preset time period to be recognized. The initial electroencephalogram signal time-domain position map to be recognized can be an initial image of the electroencephalogram signal time-domain position map to be recognized at a specific time within a preset time period. The time-domain position map of the electroencephalogram signal to be identified can be a time-domain position map of the electroencephalogram signal to be identified at a specific time within a preset time period.

Specifically, after acquiring the electroencephalogram signal to be recognized of the target user and the electrode position image data of the electroencephalogram signal to be recognized, the target electrode in the electrode position image data and the target electrode position of the target electrode in the electrode position image data can be further determined, the target electrode time to-be-recognized signal data of the target electrode is determined according to the electroencephalogram signal to be recognized, a plurality of continuous initial electroencephalogram signal time domain position maps to be recognized are generated according to the target electrode time to-be-recognized signal data and the target electrode position, the initial electroencephalogram signal time domain position map to be recognized is subjected to image supplement processing through a linear interpolation method to obtain the electroencephalogram signal time domain position map to be recognized, and therefore each electroencephalogram signal time domain position map to be recognized in a preset time period is used as the electroencephalogram signal time domain position map to be recognized.

Optionally, generating a frequency domain position map of the electroencephalogram signal to be identified according to the electroencephalogram signal to be identified and the electrode position image data may include: determining a target electrode in the electrode position image data and a target electrode position of the target electrode in the electrode position image data; determining target electrode time to-be-identified signal data of a target electrode according to the electroencephalogram signal to be identified; generating target electrode time to-be-identified frequency domain signal data of a target electrode according to the target electrode time to-be-identified signal data through discrete short-time Fourier transform; generating a plurality of continuous electroencephalogram signal time frequency domain position graphs to be identified according to the frequency domain signal data to be identified at the target electrode time and the position of the target electrode; and taking the time frequency domain position graph of each electroencephalogram signal to be identified in a preset time period as the frequency domain position graph of the electroencephalogram signal to be identified.

The frequency domain signal data to be identified at the target electrode time can be frequency domain data of the electroencephalogram signal to be identified at a specific time within a preset time period of the target electrode. The time frequency domain position map of the electroencephalogram signal to be identified can be a frequency domain position map of the electroencephalogram signal to be identified at a specific time within a preset time period.

Specifically, after acquiring the electroencephalogram signal to be recognized of the target user and the electrode position image data of the electroencephalogram signal to be recognized, the target electrode in the electrode position image data and the target electrode position of the target electrode in the electrode position image data can be further determined, the target electrode time to-be-recognized signal data of the target electrode is determined according to the electroencephalogram signal to be recognized, the target electrode time to-be-recognized frequency domain signal data of the target electrode is generated according to the target electrode time to-be-recognized signal data through discrete short-time fourier transform, a plurality of continuous electroencephalogram signal time frequency domain position maps to be recognized are generated according to the target electrode time to-be-recognized frequency domain signal data and the target electrode position, and therefore the electroencephalogram signal time frequency domain position maps to be recognized in a preset time period are used as the electroencephalogram signal position maps to be recognized.

And S430, inputting the electroencephalogram signal identification model input data to a target electroencephalogram signal identification model matched with the target user so as to identify the user identity of the target user through the target electroencephalogram signal identification model.

In the embodiment of the invention, after electroencephalogram signal identification model input data are generated according to the electroencephalogram signal to be identified and the electrode position image data, the electroencephalogram signal identification model input data can be further input into a target electroencephalogram signal identification model matched with a target user, so that the user identity of the target user can be identified through the target electroencephalogram signal identification model.

Optionally, inputting the electroencephalogram signal recognition model input data into a target electroencephalogram signal recognition model matched with the target user, so as to recognize the user identity of the target user through the target electroencephalogram signal recognition model, which may include: inputting the electroencephalogram signal time domain position map to be identified into a first target convolutional neural network of a target electroencephalogram signal identification model, inputting the electroencephalogram signal frequency domain position map to be identified into a second target convolutional neural network of the target electroencephalogram signal identification model, extracting electroencephalogram signal time domain characteristics to be identified of the electroencephalogram signal time domain position map to be identified through the first target convolutional neural network, and extracting electroencephalogram signal frequency domain characteristics to be identified of the electroencephalogram signal frequency domain position map to be identified through the second target convolutional neural network; performing feature splicing and fusion on the electroencephalogram signal time domain features to be identified and the electroencephalogram signal frequency domain features to be identified to obtain electroencephalogram signal fusion features to be identified; inputting the electroencephalogram signal fusion characteristics to be recognized into a target cyclic neural network of a target electroencephalogram signal recognition model to obtain electroencephalogram signal hidden layer information to be recognized; classifying and identifying the EEG hidden layer information to be identified through a linear classifier of the target EEG identification model to obtain a user identity identification result of data input by the EEG identification model so as to identify the user identity of the target user.

The first target convolutional neural network can be a convolutional neural network in the target brain electrical signal identification model. The second target convolutional neural network may be another convolutional neural network in the target brain electrical signal recognition model. The electroencephalogram signal time domain features to be identified can be features extracted from an electroencephalogram signal time domain position map to be identified. The electroencephalogram signal frequency domain features to be identified can be features extracted from the electroencephalogram signal frequency domain position map to be identified. The electroencephalogram signal fusion characteristics to be identified can be characteristics obtained by performing characteristic fusion on the electroencephalogram signal time domain characteristics to be identified and the electroencephalogram signal frequency domain characteristics to be identified. The hidden layer information of the electroencephalogram signal to be recognized can be hidden state information output by a recurrent neural network of a target electroencephalogram signal recognition model.

Specifically, after electroencephalogram signal identification model input data are generated according to an electroencephalogram signal to be identified and electrode position image data, an electroencephalogram signal time domain position map to be identified can be further input into a first target convolutional neural network of a target electroencephalogram signal identification model, an electroencephalogram signal frequency domain position map to be identified is input into a second target convolutional neural network of the target electroencephalogram signal identification model, electroencephalogram signal time domain features to be identified of the electroencephalogram signal time domain position map to be identified are extracted through the first target convolutional neural network, and electroencephalogram signal frequency domain features to be identified of the electroencephalogram signal frequency domain position map to be identified are extracted through the second target convolutional neural network. And performing feature splicing and fusion on the time domain features of the electroencephalogram signal to be recognized and the frequency domain features of the electroencephalogram signal to be recognized to obtain electroencephalogram signal fusion features to be recognized, inputting the electroencephalogram signal fusion features to be recognized into a target recurrent neural network of a target electroencephalogram signal recognition model to obtain electroencephalogram signal hidden layer information to be recognized, and performing classification and recognition on the electroencephalogram signal hidden layer information to be recognized through a linear classifier of the target electroencephalogram signal recognition model to obtain a user identity recognition result of data input by the electroencephalogram signal recognition model, so that the user identity of a target user is recognized.

Alternatively, the target recurrent neural network may be a long-short term memory network. The EEG hidden layer information to be identified can be hidden state information output by the last layer of LSTM unit. The first convolutional layer of the first target convolutional neural network and the first convolutional layer of the second target convolutional neural network can both adopt a conditional parameter convolutional structure, and a squeeze excitation module can be included between the first convolutional layer and the second convolutional layer.

In a specific example, an application scenario of user identification when banking is handled is taken as an example for explanation. Identity authentication is a basic security precaution for banking. Conventional authentication systems are based on passwords entered using a keyboard, touch screen or mouse, and these identification methods are easily lost, forgotten and stolen. With the ever increasing security risks and limitations of traditional models, biometric mechanisms, particularly neural activity-based approaches, continue to emerge. While these authentication methods may utilize unique biometric identification techniques to improve security, they still face the following challenges: the biometric features can be replicated; after the body part with the characteristics is damaged, the body part cannot be continuously identified; there are certain limitations to the use of different types of people, for example, disabled people without arms cannot use fingerprint recognition, and blind people cannot use iris recognition. For various banking services, such as money transfer and remittance, if the bank card password is stolen or cracked, the fingerprint is copied, which causes great loss to users. Compared with the traditional biological identification technology (such as human face, fingerprint or DNA), the electroencephalogram signals cannot be copied and stolen by the current technology, cannot be obtained from the dead, and are difficult to coerce or violently obtain, so that the great stress generated by people in danger can cause severe influence on brain activity. Therefore, the bank business transaction uses the identification technology based on the electroencephalogram signals to provide greater guarantee for the property safety of the client.

The electroencephalogram signal identification method widely used at present depends on characteristics constructed manually, and a large amount of preprocessing needs to be carried out on the electroencephalogram signals before identification is carried out. The conventional method generally comprises two stages: a feature processing stage and a feature classification stage. The feature processing stage is an extremely important part, and the excellent feature extraction can greatly improve the feature classification precision of the model. However, the conventional method usually uses manually designed data features, which leads to a complicated electroencephalogram signal identification process, is prone to error, and reduces identification efficiency and accuracy.

The identification of the user handling the banking business is carried out by identifying the electroencephalogram signal, and the identification method can specifically comprise the following steps: acquiring the electroencephalogram signal to be identified when a target user watches the image sequence, and acquiring the image data of the electrode position. And determining a target electrode in the electrode position image data and a target electrode position of the target electrode in the electrode position image data, and determining target electrode time to-be-identified signal data of the target electrode according to the electroencephalogram signals to be identified. Inputting target electrode time to-be-identified signal data to a target electrode position to obtain a plurality of continuous initial electroencephalogram time and time domain position maps to-be-identified, performing image supplement processing on the initial electroencephalogram time and time domain position maps to-be-identified through a linear interpolation method to obtain electroencephalogram time and time domain position maps to-be-identified, and taking each electroencephalogram time and time domain position map to-be-identified in a preset time period as an electroencephalogram time and time domain position map to-be-identified. And generating target electrode time to-be-identified frequency domain signal data of a target electrode according to the target electrode time to-be-identified signal data through discrete short-time Fourier transform, and generating a plurality of continuous electroencephalogram time frequency domain position maps to be identified according to the target electrode time to-be-identified frequency domain signal data and the position of the target electrode, so that each electroencephalogram time frequency domain position map to be identified in a preset time period is used as an electroencephalogram frequency domain position map to be identified. The electroencephalogram signal time domain position map to be recognized is input into a first target convolutional neural network of a target electroencephalogram signal recognition model, the electroencephalogram signal frequency domain position map to be recognized is input into a second target convolutional neural network of the target electroencephalogram signal recognition model, electroencephalogram signal time domain characteristics to be recognized of the electroencephalogram signal time domain position map to be recognized are extracted through the first target convolutional neural network, and electroencephalogram signal frequency domain characteristics to be recognized of the electroencephalogram signal frequency domain position map to be recognized are extracted through the second target convolutional neural network. The method comprises the steps of performing feature splicing and fusion on electroencephalogram time domain features to be recognized and electroencephalogram frequency domain features to be recognized to obtain electroencephalogram fusion features to be recognized, inputting the electroencephalogram fusion features to be recognized into a target recurrent neural network of a target electroencephalogram recognition model to obtain electroencephalogram hidden layer information to be recognized, and performing classification and recognition on the electroencephalogram hidden layer information to be recognized through a linear classifier of the target electroencephalogram recognition model to obtain a user identity recognition result of electroencephalogram recognition model input data, so that the user identity of a target user is recognized.

According to the technical scheme, the recognition of the electroencephalogram signals is combined with the financial business handling, so that the safety of the financial business can be guaranteed, and convenience is provided for the financial business handling.

According to the technical scheme, electroencephalogram signals to be recognized of a target user and electrode position image data of the electroencephalogram signals to be recognized are obtained, electroencephalogram signal recognition model input data including an electroencephalogram signal time domain position diagram to be recognized and an electroencephalogram signal frequency domain position diagram to be recognized are generated according to the electroencephalogram signals to be recognized and the electrode position image data, the electroencephalogram signal recognition model input data are input to a target electroencephalogram signal recognition model matched with the target user, the user identity of the target user is recognized through the target electroencephalogram signal recognition model, the problems that an existing electroencephalogram signal recognition method is low in recognition efficiency and poor in recognition accuracy are solved, and the electroencephalogram signal recognition efficiency and the recognition accuracy can be improved.

EXAMPLE III

Fig. 5 is a schematic diagram of an electroencephalogram signal recognition model training device provided by a third embodiment of the present invention, and as shown in fig. 5, the device includes: a sample data obtaining module 510, a model input sample data generating module 520, and a target electroencephalogram signal recognition model obtaining module 530, wherein:

a sample data obtaining module 510, configured to obtain electroencephalogram signal sample data of a target user and electrode position image sample data of the electroencephalogram signal sample data;

a model input sample data generating module 520, configured to generate electroencephalogram signal identification model input sample data according to the electroencephalogram signal sample data and the electrode position image sample data; the electroencephalogram signal identification model input sample data comprises an electroencephalogram signal time domain position diagram and an electroencephalogram signal frequency domain position diagram;

and the target electroencephalogram signal identification model obtaining module 530 is used for training an electroencephalogram signal identification model according to the electroencephalogram signal identification model input sample data to obtain a target electroencephalogram signal identification model matched with the target user.

Optionally, the sample data obtaining module 510 may be specifically configured to: displaying a preset display image to a target user; acquiring electroencephalogram sample data generated when a target user watches a preset display image.

Optionally, the model input sample data generating module 520 may be specifically configured to: determining a target electrode in the electrode position image sample data and a target electrode position of the target electrode in the electrode position image sample data; determining target electrode time signal data of a target electrode according to the electroencephalogram signal sample data; generating a plurality of continuous initial electroencephalogram signal time-domain position graphs according to the target electrode time signal data and the target electrode position; performing image supplement processing on the initial electroencephalogram signal time-domain position map by a linear interpolation method to obtain an electroencephalogram signal time-domain position map; and taking the time domain position map of each electroencephalogram signal moment in a preset time period as the time domain position map of the electroencephalogram signal.

Optionally, the model input sample data generating module 520 may be specifically configured to: determining a target electrode in the electrode position image sample data and a target electrode position of the target electrode in the electrode position image sample data; determining target electrode time signal data of a target electrode according to the electroencephalogram signal sample data; generating target electrode time frequency domain signal data of a target electrode according to the target electrode time signal data through discrete short-time Fourier transform; generating a plurality of continuous electroencephalogram signal time frequency domain position graphs according to the target electrode time frequency domain signal data and the target electrode position; and taking the time frequency domain position map of each electroencephalogram signal in a preset time period as the frequency domain position map of the electroencephalogram signal.

Optionally, the target electroencephalogram signal identification model obtaining module 530 may be specifically configured to: inputting the electroencephalogram signal time domain position map into a first convolution neural network of the electroencephalogram signal identification model, inputting the electroencephalogram signal frequency domain position map into a second convolution neural network of the electroencephalogram signal identification model, extracting electroencephalogram signal time domain characteristics of the electroencephalogram signal time domain position map through the first convolution neural network, and extracting electroencephalogram signal frequency domain characteristics of the electroencephalogram signal frequency domain position map through the second convolution neural network; performing feature splicing and fusion on the electroencephalogram signal time domain features and the electroencephalogram signal frequency domain features to obtain electroencephalogram signal fusion features; inputting the EEG signal fusion characteristics to a recurrent neural network of an EEG signal identification model to obtain EEG signal hidden layer information; classifying and identifying the EEG hidden layer information through a linear classifier of the EEG identification model to obtain a user identity identification result of the EEG identification model input sample data; and comparing the user identity recognition result with the user identity marking result of the sample data input by the electroencephalogram signal recognition model to determine the training effect of the electroencephalogram signal recognition model.

The electroencephalogram signal recognition model training device can execute the electroencephalogram signal recognition model training method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method. For details of the technology that are not described in detail in this embodiment, reference may be made to the electroencephalogram signal recognition model training method provided in any embodiment of the present invention.

Since the above-described electroencephalogram signal recognition model training device is a device capable of executing the electroencephalogram signal recognition model training method in the embodiment of the present invention, based on the electroencephalogram signal recognition model training method described in the embodiment of the present invention, a person skilled in the art can understand the specific implementation manner and various variations of the electroencephalogram signal recognition model training device of the embodiment, and therefore, how the electroencephalogram signal recognition model training device implements the electroencephalogram signal recognition model training method in the embodiment of the present invention is not described in detail here. As long as the person skilled in the art implements the device adopted by the electroencephalogram signal recognition model training method in the embodiment of the invention, the device belongs to the scope to be protected by the present application.

Example four

Fig. 6 is a schematic diagram of an electroencephalogram signal identification device according to a fourth embodiment of the present invention, and as shown in fig. 6, the device includes: an image acquisition module 610, a model input data generation module 620, and a user identification module 630, wherein:

the image acquisition module 610 is used for acquiring an electroencephalogram signal to be identified of a target user and electrode position image data of the electroencephalogram signal to be identified;

the model input data generation module 620 is used for generating electroencephalogram signal identification model input data according to the electroencephalogram signal to be identified and the electrode position image data; the electroencephalogram signal identification model input data comprise an electroencephalogram signal time domain position diagram to be identified and an electroencephalogram signal frequency domain position diagram to be identified;

the user identity recognition module 630 is configured to input the electroencephalogram signal recognition model input data to the target electroencephalogram signal recognition model matched with the target user, so as to recognize the user identity of the target user through the target electroencephalogram signal recognition model.

The electroencephalogram signal identification device can execute the electroencephalogram signal identification method provided by any embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method. For details of the technology that are not described in detail in this embodiment, reference may be made to the electroencephalogram signal identification method provided in any embodiment of the present invention.

Since the electroencephalogram signal identification device described above is a device capable of executing the electroencephalogram signal identification method in the embodiment of the present invention, based on the electroencephalogram signal identification method described in the embodiment of the present invention, a person skilled in the art can understand the specific implementation of the electroencephalogram signal identification device of the embodiment and various variations thereof, and therefore, how the electroencephalogram signal identification device implements the electroencephalogram signal identification method in the embodiment of the present invention is not described in detail here. As long as the device adopted by the electroencephalogram signal identification method in the embodiment of the invention is implemented by persons skilled in the art, the device belongs to the scope of the protection to be claimed in the application.

EXAMPLE five

Fig. 7 is a schematic structural diagram of an electronic device according to a fifth embodiment of the present invention. FIG. 7 illustrates a block diagram of an exemplary electronic device 12 suitable for use in implementing embodiments of the present invention. The electronic device 12 shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiment of the present invention.

As shown in FIG. 7, electronic device 12 is embodied in the form of a general purpose computing device. The components of electronic device 12 may include, but are not limited to: one or more processors 16, a memory 28, and a bus 18 that connects the various system components (including the memory 28 and the processors 16).

Bus 18 represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, and a processor or local bus using any of a variety of bus architectures. By way of example, such architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MCA) bus, an enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

Electronic device 12 typically includes a variety of computer system readable media. Such media may be any available media that is accessible by electronic device 12 and includes both volatile and nonvolatile media, removable and non-removable media.

Memory 28 may include computer system readable media in the form of volatile Memory, such as Random Access Memory (RAM) 30 and/or cache Memory 32. The electronic device 12 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 7, and commonly referred to as a "hard drive"). Although not shown in FIG. 7, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a Compact disk-Read Only Memory (CD-ROM), a Digital Video disk (DVD-ROM), or other optical media) may be provided. In these cases, each drive may be connected to bus 18 by one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the invention.

A program/utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. Program modules 42 generally carry out the functions and/or methodologies of the described embodiments of the invention.

Electronic device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), with one or more devices that enable a user to interact with electronic device 12, and/or with any devices (e.g., network card, modem, etc.) that enable electronic device 12 to communicate with one or more other computing devices. Such communication may be through an Input/Output (I/O) interface 22. Also, the electronic device 12 may communicate with one or more networks (e.g., a Local Area Network (LAN), Wide Area Network (WAN), and/or a public Network such as the internet) via the Network adapter 20. As shown, the network adapter 20 communicates with other modules of the electronic device 12 via the bus 18. It should be appreciated that although not shown in FIG. 7, other hardware and/or software modules may be used in conjunction with electronic device 12, including but not limited to: microcode, device drivers, Redundant processing units, external disk drive Arrays, (Redundant Arrays of Independent Disks, RAID) systems, tape drives, and data backup storage systems, to name a few.

The processor 16 executes various functional applications and data processing by running the program stored in the memory 28, so as to implement the electroencephalogram signal recognition model training method provided by the first embodiment of the present invention: acquiring electroencephalogram sample data of a target user and electrode position image sample data of the electroencephalogram sample data; generating electroencephalogram signal identification model input sample data according to the electroencephalogram signal sample data and the electrode position image sample data; the electroencephalogram signal identification model input sample data comprises an electroencephalogram signal time domain position diagram and an electroencephalogram signal frequency domain position diagram; and training an electroencephalogram signal identification model according to the electroencephalogram signal identification model input sample data to obtain a target electroencephalogram signal identification model matched with the target user.

Or the electroencephalogram signal identification method provided by the second embodiment of the invention is realized as follows: acquiring an electroencephalogram signal to be identified of a target user and electrode position image data of the electroencephalogram signal to be identified; generating electroencephalogram signal identification model input data according to the electroencephalogram signal to be identified and the electrode position image data; the electroencephalogram signal identification model input data comprise an electroencephalogram signal time domain position diagram to be identified and an electroencephalogram signal frequency domain position diagram to be identified; and inputting the EEG signal identification model input data into a target EEG signal identification model matched with the target user so as to identify the user identity of the target user through the target EEG signal identification model.

EXAMPLE six

An embodiment of the present invention further provides a computer storage medium storing a computer program, where the computer program is executed by a computer processor to perform the electroencephalogram signal recognition model training method according to the first embodiment of the present invention: acquiring electroencephalogram sample data of a target user and electrode position image sample data of the electroencephalogram sample data; generating electroencephalogram signal identification model input sample data according to the electroencephalogram signal sample data and the electrode position image sample data; the electroencephalogram signal identification model input sample data comprises an electroencephalogram signal time domain position diagram and an electroencephalogram signal frequency domain position diagram; and training an electroencephalogram signal identification model according to the electroencephalogram signal identification model input sample data to obtain a target electroencephalogram signal identification model matched with the target user.

Or the computer program is used for executing the electroencephalogram signal identification method according to the second embodiment of the present invention when executed by a computer processor: acquiring an electroencephalogram signal to be identified of a target user and electrode position image data of the electroencephalogram signal to be identified; generating electroencephalogram signal identification model input data according to the electroencephalogram signal to be identified and the electrode position image data; the electroencephalogram signal identification model input data comprise an electroencephalogram signal time domain position diagram to be identified and an electroencephalogram signal frequency domain position diagram to be identified; and inputting the EEG signal identification model input data into a target EEG signal identification model matched with the target user so as to identify the user identity of the target user through the target EEG signal identification model.

Computer storage media for embodiments of the invention may employ any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a Read-Only Memory (ROM), an Erasable Programmable Read-Only Memory (EPROM) or flash Memory), an optical fiber, a portable compact disc Read-Only Memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, Radio Frequency (RF), etc., or any suitable combination of the foregoing.

Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).

It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. An electroencephalogram signal recognition model training method is characterized by comprising the following steps:

2. The method of claim 1, wherein obtaining brain electrical signal sample data of a target user comprises:

displaying a preset display image to the target user;

acquiring electroencephalogram sample data generated when the target user watches the preset display image.

3. The method of claim 1, wherein generating a brain electrical signal time domain position map from the brain electrical signal sample data and the electrode position image sample data comprises:

determining a target electrode in the electrode position image sample data and a target electrode position of the target electrode in the electrode position image sample data;

determining target electrode time signal data of the target electrode according to the electroencephalogram signal sample data;

generating a plurality of continuous initial electroencephalogram signal time-domain position graphs according to the target electrode time signal data and the target electrode position;

performing image supplement processing on the initial electroencephalogram signal time-domain position map by a linear interpolation method to obtain an electroencephalogram signal time-domain position map;

and taking each electroencephalogram time domain position map within a preset time period as the electroencephalogram time domain position map.

4. The method of claim 1, wherein generating a brain electrical signal frequency domain location map from the brain electrical signal sample data and the electrode location image sample data comprises:

generating target electrode time frequency domain signal data of the target electrode according to the target electrode time signal data through discrete short-time Fourier transform;

generating a plurality of continuous electroencephalogram signal time frequency domain position graphs according to the target electrode time frequency domain signal data and the target electrode position;

and taking each electroencephalogram time frequency domain position map in a preset time period as the electroencephalogram frequency domain position map.

5. The method of claim 1, wherein training a brain electrical signal recognition model from the brain electrical signal recognition model input sample data comprises:

inputting the electroencephalogram signal time domain position map into a first convolution neural network of the electroencephalogram signal identification model, inputting the electroencephalogram signal frequency domain position map into a second convolution neural network of the electroencephalogram signal identification model, so as to extract electroencephalogram signal time domain characteristics of the electroencephalogram signal time domain position map through the first convolution neural network, and extract electroencephalogram signal frequency domain characteristics of the electroencephalogram signal frequency domain position map through the second convolution neural network;

performing feature splicing and fusion on the electroencephalogram signal time domain features and the electroencephalogram signal frequency domain features to obtain electroencephalogram signal fusion features;

inputting the EEG signal fusion characteristics to a recurrent neural network of the EEG signal identification model to obtain EEG signal hidden layer information;

classifying and identifying the EEG hidden layer information through a linear classifier of the EEG identification model to obtain a user identity identification result of the EEG identification model input sample data;

and comparing the user identity recognition result with the user identity marking result of the input sample data of the electroencephalogram signal recognition model to determine the training effect of the electroencephalogram signal recognition model.

6. An electroencephalogram signal identification method, characterized by comprising:

7. The utility model provides an EEG signal recognition model trainer which characterized in that includes:

8. An electroencephalogram signal identification device, characterized by comprising:

the model input data generation module is used for generating electroencephalogram signal identification model input data according to the electroencephalogram signal to be identified and the electrode position image data; the electroencephalogram signal identification model input data comprise an electroencephalogram signal time domain position diagram to be identified and an electroencephalogram signal frequency domain position diagram to be identified;

and the user identity identification module is used for inputting the electroencephalogram signal identification model input data to the target electroencephalogram signal identification model matched with the target user so as to identify the user identity of the target user through the target electroencephalogram signal identification model.

9. An electronic device, characterized in that the electronic device comprises:

at least one processor; and

the memory stores a computer program executable by the at least one processor, the computer program being executable by the at least one processor to enable the at least one processor to perform the brain electrical signal recognition model training method of any one of claims 1-5 or to perform the brain electrical signal recognition method of claim 6.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores computer instructions for causing a processor to implement, when executed, the electroencephalogram signal recognition model training method of any one of claims 1 to 5, or the electroencephalogram signal recognition method of claim 6.