CN108108763B - EEG classification model generation method, device and electronic device - Google Patents

Abstract

The application provides an electroencephalogram classification model generation method, an electroencephalogram classification model generation device, an electronic device and a computer-readable storage medium, wherein the electroencephalogram classification model generation method comprises the following steps: obtaining sample data of K subjects, wherein the sample data comprises: the classified electroencephalogram information and the classification result of the corresponding electroencephalogram information, wherein K is greater than or equal to 2; calculating an orthogonal transformation matrix which enables the first target function to take the minimum value based on sample data of K tested objects and a preset first target function, wherein the first target function is a function related to the orthogonal transformation matrix and electroencephalogram information of the K tested objects, and the orthogonal transformation matrix is used for transforming the electroencephalogram information of the K tested objects into correlation information among the K tested objects; and generating an electroencephalogram classification model based on the orthogonal transformation matrix. The technical scheme is used for generating the electroencephalogram classification models applicable to a plurality of subjects, and the maintenance cost of the electroencephalogram classification models is saved.

Description

EEG classification model generation method, device and electronic device

technical field

The present application belongs to the technical field of brain-computer interaction, and in particular, relates to a method for generating an EEG classification model, a device for generating an EEG classification model, an electronic device, and a computer-readable storage medium.

Background technique

Electroencephalogram (EEG) is the overall reflection of the electrophysiological activity of brain nerve cells on the surface of the cerebral cortex or scalp.

At present, brain-computer interaction technology based on EEG signals has become a research hotspot in the industry. The key technology of brain-computer interaction technology is how to extract EEG information quickly and effectively and improve the recognition accuracy. Considering the high degree of non-stationarity and individual differences of EEG signals, the EEG classification models trained based on the EEG information of different subjects have significant differences. Therefore, in the existing brain-computer interaction system, targeting at Each subject trains an independent EEG classification model (that is, each EEG classification model is adapted to a subject), in order to classify and process the EEG information of the corresponding subject through the trained EEG classification model. , so as to improve the recognition accuracy. Since the EEG classification model can only be applied to one subject, the generalization performance of the existing brain-computer interaction system is poor, and as the number of subjects increases, the brain-computer interaction system needs to maintain There are more and more EEG classification models, and correspondingly, the maintenance cost of EEG classification models also increases.

SUMMARY OF THE INVENTION

In view of this, the present application provides an EEG classification model generation method, an EEG classification model generation device, an electronic device and a computer-readable storage medium for generating an EEG classification model applicable to multiple subjects, Save maintenance cost of EEG classification model.

A first aspect of the embodiments of the present application provides a method for generating an EEG classification model, including:

Obtaining sample data of K subjects, wherein the sample data includes: classified EEG information and classification results of the corresponding EEG information, and the K is greater than or equal to 2;

Based on the sample data of the K subjects and a preset first objective function, an orthogonal transformation matrix that makes the first objective function take a minimum value is calculated, wherein the first objective function is an orthogonal transformation with The matrix is a function related to the EEG information of the K subjects, and the orthogonal transformation matrix is used to transform the respective EEG information of the K subjects into the correlation between the K subjects information;

An EEG classification model is generated based on the orthogonal transformation matrix, so as to classify the EEG information of any one of the K subjects by using the EEG classification model.

Based on the first aspect of the present application, in a first possible implementation manner, the first objective function is:

所述第一目标函数中的N_k表示第k个受试对象的样本数据个数，N_l表示第l个受试对象的样本数据个数，P表示正交变换矩阵，P^T为P的转置，

表示x_i,k的转置，x_i,k表示第k个受试对象的第i个样本数据中的脑电信息，

表示x_j,l的转置,x_j,l表示第l个受试对象的第j个样本数据中的脑电信息；

In the first objective function, N _k represents the number of sample data of the k-th subject, N _l represents the number of sample data of the l-th subject, P represents an orthogonal transformation matrix, and P ^T is the Transpose,

represents the transpose of x i, _k , x _i,k represents the EEG information in the i-th sample data of the k-th subject,

represents the transpose of x j, _l , x _j,l represents the EEG information in the jth sample data of the lth subject;

Described based on the sample data of the K subjects and the preset first objective function, the orthogonal transformation matrix that calculates the minimum value of the first objective function is:

Under the condition that PP ^T =I is satisfied, based on the sample data of the K subjects, calculate P that minimizes the first objective function, where I is an identity matrix.

Based on the first possible implementation manner of the first aspect of the present application, in a second possible implementation manner, the generating an EEG classification model based on the orthogonal transformation matrix includes:

According to the obtained orthogonal transformation matrix, the first objective function and the preset second objective function, determine the first weight vector, the second weight vector and the second objective function based on the Lagrangian expression offset value;

generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value;

Wherein, the second objective function is:

且所述第二目标函数满足：

And the second objective function satisfies:

为W_k的转置，W_k表示与第k个受试对象相关的第一权向量，V^T为V的转置,V表示第二权向量，b_k表示与第k个受试对象相关的偏置值。ρ _i,k represents the contribution of the i-th sample data of the k-th subject, ε _i,k is the slack variable of x _i,k , C _k and λ are values greater than 0, y _i,k represents the ith The classification result of the ith sample data of k subjects,

is the transpose of W _k , W _k represents the first weight vector related to the k-th subject, V ^T is the transpose of V, V represents the second weight vector, and b _k represents the k-th subject related to the subject offset value.

Based on the second possible implementation manner of the first aspect of the present application, in a third possible implementation manner, the first weight vector and the second weight vector determined based on the orthogonal transformation matrix and the determined The vector and the bias value generate the EEG classification model as:

A classification algorithm for generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value;

The classification algorithm is:

且，α_i,k'和b_k满足等式[K+Z]·α＝Y，在所述等式中，

Among them, when k'=k, Z _i,k =Px _i,k , when k'≠k, Z _i,k =Qx _i,k , and,

And, α _i,k' and b _k satisfy the equation [K+Z]·α=Y, in which,

A second aspect of the present application provides a device for generating an EEG classification model, including:

an obtaining unit, configured to obtain sample data of K subjects, wherein the sample data includes: classified EEG information and a classification result of the corresponding EEG information, and the K is greater than or equal to 2;

A computing unit for calculating an orthogonal transformation matrix that makes the first objective function take a minimum value based on the sample data of the K subjects and a preset first objective function, wherein the first objective function is a function related to an orthogonal transformation matrix and the EEG information of the K subjects, and the orthogonal transformation matrix is used to transform the respective EEG information of the K subjects into the K subjects Correlation information between;

A generating unit, configured to generate an EEG classification model based on the orthogonal transformation matrix, so as to classify the EEG information of any one of the K subjects by using the EEG classification model.

Based on the second aspect of the present application, in a first possible implementation manner, the first objective function is:

所述第一目标函数中的N_k表示第k个受试对象的样本数据个数，N_l表示第l个受试对象的样本数据个数，P表示正交变换矩阵，P^T为P的转置，

表示x_i,k的转置，x_i,k表示第k个受试对象的第i个样本数据中的脑电信息，

表示x_j,l的转置,x_j,l表示第l个受试对象的第j个样本数据中的脑电信息；

In the first objective function, N _k represents the number of sample data of the k-th subject, N _l represents the number of sample data of the l-th subject, P represents an orthogonal transformation matrix, and P ^T is the Transpose,

represents the transpose of x i, _k , x _i,k represents the EEG information in the i-th sample data of the k-th subject,

represents the transpose of x j, _l , x _j,l represents the EEG information in the jth sample data of the lth subject;

The computing unit is specifically configured to: under the condition that ^PPT =I is satisfied, based on the sample data of the K subjects, calculate P that makes the first objective function take the minimum value, wherein the I is identity matrix.

Based on the first possible implementation manner of the second aspect of the present application, in the second possible implementation manner, the generating unit includes:

The determining unit is used to determine the first objective function in the second objective function based on the Lagrangian expression according to the orthogonal transformation matrix, the first objective function and the preset second objective function calculated by the computing unit. a weight vector, a second weight vector and a bias value;

a sub-generating unit, configured to generate an EEG classification model based on the orthogonal transformation matrix and the first weight vector, the second weight vector and the bias value determined by the determining unit;

Wherein, the second objective function is:

且所述第二目标函数满足：

And the second objective function satisfies:

为W_k的转置，W_k表示与第k个受试对象相关的第一权向量，V^T为V的转置,V表示第二权向量，b_k表示与第k个受试对象相关的偏置值。ρ _i,k represents the contribution of the i-th sample data of the k-th subject, ε _i,k is the slack variable of x _i,k , C _k and λ are values greater than 0, y _i,k represents the ith The classification result of the ith sample data of k subjects,

is the transpose of W _k , W _k represents the first weight vector related to the k-th subject, V ^T is the transpose of V, V represents the second weight vector, and b _k represents the k-th subject related to the subject offset value.

Based on the second possible implementation manner of the second aspect of the present application, in the third possible implementation manner, the sub-generating model is specifically used for:

A classification algorithm for generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value;

The classification algorithm is:

且，α_i,k'和b_k满足等式[K+Z]·α＝Y，在所述等式中，

Among them, when k'=k, Z _i,k =Px _i,k , when k'≠k, Z _i,k =Qx _i,k , and,

And, α _i,k' and b _k satisfy the equation [K+Z]·α=Y, in which,

A third aspect of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the first aspect or the first aspect when the processor executes the computer program. The EEG classification model generation method mentioned in any possible implementation manner of .

A fourth aspect of the present application provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the foregoing first aspect or any possible implementation manner of the foregoing first aspect is implemented The EEG classification model generation method mentioned in .

A fifth aspect of the present application provides a computer program product, the computer program product includes a computer program, and the computer program implements the first aspect or any possible implementation of the first aspect when the computer program is executed by one or more processors The EEG classification model generation method mentioned in the method.

As can be seen from the above, the present application obtains the sample data of K subjects, and calculates an orthogonal transformation matrix based on the sample data of K subjects and the first objective function, and then generates an EEG classification model based on the orthogonal transformation matrix. . Since the orthogonal transformation matrix can be used to transform the respective EEG information of the K subjects into the correlation information between the K subjects, the EEG classification model generated based on the orthogonal transformation matrix can adapt to K The difference of EEG information between subjects makes the generated EEG classification model applicable to K subjects. Since multiple subjects can share the same EEG classification model, compared with the traditional scheme, The present application can maintain an EEG classification mode for multiple subjects, which effectively saves the maintenance cost of the EEG classification model.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present application more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only for the present application. In some embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without any creative effort.

1-a is a schematic flowchart of an embodiment of a method for generating an EEG classification model provided by the application;

1-b is a schematic diagram of the algorithm framework of the EEG classification model provided by the application;

2 is a schematic structural diagram of an embodiment of an EEG classification model generation device provided by the present application;

FIG. 3 is a schematic structural diagram of an embodiment of the electronic device provided by the present application.

Detailed ways

In the following description, for the purpose of illustration rather than limitation, specific details such as a specific system structure and technology are set forth in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be practiced in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

It should be understood that the size of the sequence numbers of the steps in the following method embodiments does not mean the sequence of execution, and the execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of each embodiment. .

In order to illustrate the technical solutions described in the present application, the following specific embodiments are used for description.

Example 1

The embodiment of the present application provides a method for generating an EEG classification model. The method for generating an EEG classification model can be applied to a device for generating an EEG classification model. The device for generating an EEG classification model can be an independent device, or an EEG classification model. The model generation apparatus can also be integrated in electronic devices (such as smartphones, tablets, computers, and wearable devices, etc.). Optionally, the operating system carried by the device or electronic device that integrates the EEG classification model generating apparatus may be an ios system, an android system, a windows system or other operating systems, which is not limited here.

Referring to FIG. 1-a, the method for generating an EEG classification model in the embodiment of the present application may include:

Step 101, obtain the sample data of K subjects;

In the embodiment of the present application, the sample data of the subject includes: the classified EEG information of the subject and the classification result of the corresponding EEG information. In practical applications, for any subject among the K subjects, technicians can manually classify part of the subject's EEG information to obtain the classification result of the corresponding EEG information, and then obtain the subject's EEG information. For the sample data of the subject, after the generation (ie, training) of the EEG classification model is completed based on the sample data, the generated EEG classification model can be used to automatically classify the EEG information of the subject.

In the embodiments of this application, EEG information refers to feature information extracted from EEG signals. Specifically, the process of collecting EEG signals and the process of extracting EEG information can be implemented with reference to the prior art, which is not repeated here. Repeat.

In the embodiment of the present application, the above K is greater than or equal to 2, that is, the sample data of a plurality of subjects is acquired in step 101 .

Step 102, based on the sample data of the above-mentioned K subjects and the preset first objective function, calculate the orthogonal transformation matrix that makes the above-mentioned first objective function take the minimum value;

In this embodiment of the present application, each subject can be regarded as each task in multi-task learning. Combined with the theory of latent structure learning, it is assumed that there is a shared latent space, and the information contained in the shared latent space is obtained by transforming the EEG information of multiple subjects through an orthogonal transformation matrix. relevance information. By using the maximum joint probability distribution criterion, the correlation information of multiple tasks mapped to the shared latent space is explored, and the correlation information is used to help improve the learning effect of each task, so as to realize multi-task learning.

In step 102, the orthogonal transformation matrix is used to transform the respective EEG information of the K subjects into the correlation information between the K subjects. The first objective function is a function related to the orthogonal transformation matrix and the EEG information of the K subjects. Specifically, the first objective function can be derived in the following way:

对于隐含空间中任意第k个任务和第l个任务，其Parzen窗密度估计可以分别表示为：According to the Gaussian distribution function

For any k-th task and l-th task in the latent space, the Parzen window density estimation can be expressed as:

最小，也即，最小化如下式

经过一系列数学推导，可得到如下求解正交变换矩阵P的第一目标函数：After derivation, minimizing the distribution difference between the k-th task and the l-th task ∫ _k,l (P _k (X)-P _l (X)) ² dX should be such that

Minimum, that is, minimize the following formula

After a series of mathematical derivations, the first objective function for solving the orthogonal transformation matrix P can be obtained as follows:

在上述第一目标函数中，N_k表示第k个受试对象的样本数据个数，N_l表示第l个受试对象的样本数据个数，P表示正交变换矩阵，P^T为P的转置，

表示x_i，k的转置，x_i,k表示第k个受试对象的第i个样本数据中的脑电信息，

表示x_j,l的转置,x_j,l表示第l个受试对象的第j个样本数据中的脑电信息。

In the above first objective function, N _k represents the number of sample data of the k-th subject, N _l represents the number of sample data of the l-th subject, P represents the orthogonal transformation matrix, and P ^T is the Transpose,

represents the transpose of x i, _k , x _i,k represents the EEG information in the i-th sample data of the k-th subject,

represents the transpose of x j, _l , and x _j,l represents the EEG information in the jth sample data of the lth subject.

Based on the above-mentioned first objective function, step 102 can specifically be expressed as: under the condition that ^PPT = 1 is satisfied, based on the sample data of the above-mentioned K subjects, calculate P (that is, the minimum value of the above-mentioned first objective function) Orthogonal transformation matrix), wherein, the above-mentioned E is the identity matrix. Specifically, under the condition that ^PPT =I is satisfied, based on the sample data of the K subjects, the gradient descent method can be used to calculate and calculate P that makes the first objective function take the minimum value.

Optionally, in the process of calculating the above-mentioned orthogonal transformation matrix, in addition to the classified EEG information, the collected but unclassified EEG information may also be used. The sample data obtained in step 101 may further include: collected but unclassified EEG information.

Step 103, generating an EEG classification model based on the above-mentioned orthogonal transformation matrix;

In step 103, an EEG classification model is generated based on the orthogonal transformation matrix calculated in step 102, so as to use the EEG classification model to classify the EEG information of any one of the above K subjects.

Assume that the algorithm framework of the EEG classification model in the embodiment of the present application is shown in Fig. 1-b. The EEG classification model includes two learning parts, namely learning based on the original data space and learning based on the shared latent space. Among them, the data in the original data space is the respective EEG information of multiple subjects, and the data in the shared latent space is the correlation information between different subjects (the correlation information is the transformation matrix transformation). Finally, the learning results based on the original data space and the learning results based on the shared latent space are combined. Denote the dimension of the original data space as d and the dimension of the shared latent space as r, then the learning process of the above two parts can be described by the following optimization problem:

stPP ^T =I _d×d

和

P为将各受试对象的脑电信息从原始数据空间投影到共享隐空间的正交变换矩阵

共享隐空间的维度记为r。φ(·)则表示共享隐空间中各受试对象之间的相关性信息，可表示为：where W _k represents the first weight vector associated with the k-th subject, V represents the second weight vector, and

and

P is the orthogonal transformation matrix that projects the EEG information of each subject from the original data space to the shared latent space

The dimension of the shared latent space is denoted as r. φ( ) represents the correlation information between subjects in the shared latent space, which can be expressed as:

的独立性信息最优化，可表示为：For K tasks (as mentioned above, each subject is regarded as a task), g _k ( ) is the k-th task

The independence information optimization of , can be expressed as:

where b _k is the bias value associated with the k-th subject, and C _k and λ are values greater than 0.

On this basis, step 103 may include:

According to the obtained orthogonal transformation matrix, the first objective function and the preset second objective function, determine the first weight vector, the second weight vector and the second objective function based on the Lagrangian expression offset value;

generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value;

Wherein, the second objective function is:

且所述第二目标函数满足：

And the second objective function satisfies:

为W_k的转置，W_k表示与第k个受试对象相关的第一权向量，V^T为V的转置,V表示第二权向量。ρ _i,k represents the contribution of the i-th sample data of the k-th subject, ε _i,k is the slack variable of x _i,k , y _i,k represents the i-th sample of the k-th subject The classification result of the data,

is the transpose of W _k , W _k represents the first weight vector related to the k-th subject, V ^T is the transpose of V, and V represents the second weight vector.

Further, the above-mentioned generation of the EEG classification model based on the above-mentioned orthogonal transformation matrix and the determined above-mentioned first weight vector, above-mentioned second weight vector and above-mentioned bias value is:

A classification algorithm for generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value;

The classification algorithm is:

Among them, when k'=k, Z _i,k =Px _i,k , when k'≠k, Z _i,k =Qx _i,k , and,

且，α_i,k'和b_k满足等式[K+Z]·α＝Y，在上述等式中，

And, α _i,k' and b _k satisfy the equation [K+Z]·α=Y, in the above equation,

After the embodiment of the present application, the EEG information of the subject can be classified by the above classification algorithm.

As can be seen from the above, the embodiment of the present application obtains the sample data of the K subjects, and calculates an orthogonal transformation matrix based on the sample data of the K subjects and the first objective function, and then generates an EEG based on the orthogonal transformation matrix. classification model. Since the orthogonal transformation matrix can be used to transform the respective EEG information of the K subjects into the correlation information between the K subjects, the EEG classification model generated based on the orthogonal transformation matrix can adapt to K The difference of EEG information between subjects makes the generated EEG classification model applicable to K subjects. Since multiple subjects can share the same EEG classification model, compared with the traditional scheme, The present application can maintain an EEG classification mode for multiple subjects, which effectively saves the maintenance cost of the EEG classification model. Further, since the process of generating the EEG classification model is based on the sample data of multiple subjects for training, and the correlation information between multiple subjects is used to revise the training process, therefore, compared with the traditional Based on the EEG classification model obtained by training the sample data of a single subject, the EEG classification model generated in the embodiment of the present application can improve the EEG classification accuracy of each subject to a certain extent.

Embodiment 2

An embodiment of the present application provides an apparatus for generating an EEG classification model. As shown in FIG. 2 , the apparatus for generating an EEG classification model 200 in the embodiment of the present application includes:

The obtaining unit 201 is configured to obtain sample data of K subjects, wherein the sample data includes: classified EEG information and a classification result of the corresponding EEG information, and the K is greater than or equal to 2;

A computing unit 202, configured to calculate an orthogonal transformation matrix that minimizes the first objective function based on the sample data of the K subjects and a preset first objective function, wherein the first objective The function is a function related to an orthogonal transformation matrix and the EEG information of the K subjects, and the orthogonal transformation matrix is used to transform the respective EEG information of the K subjects into the K subjects. Correlation information between objects;

The generating unit 203 is configured to generate an EEG classification model based on the orthogonal transformation matrix, so as to use the EEG classification model to classify the EEG information of any one of the K subjects.

Optionally, the first objective function is:

所述第一目标函数中的N_k表示第k个受试对象的样本数据个数，N_l表示第l个受试对象的样本数据个数，P表示正交变换矩阵，P^T为P的转置，

表示x_i,k的转置，x_i,k表示第k个受试对象的第i个样本数据中的脑电信息，

表示x_j,l的转置,x_j,l表示第l个受试对象的第j个样本数据中的脑电信息；

In the first objective function, N _k represents the number of sample data of the k-th subject, N _l represents the number of sample data of the l-th subject, P represents an orthogonal transformation matrix, and P ^T is the Transpose,

represents the transpose of x i, _k , x _i,k represents the EEG information in the i-th sample data of the k-th subject,

represents the transpose of x j, _l , x _j,l represents the EEG information in the jth sample data of the lth subject;

The calculation unit 202 is specifically configured to: under the condition that ^PPT =I is satisfied, based on the sample data of the K subjects, calculate P that makes the first objective function take the minimum value, wherein the I is a unit matrix.

Optionally, the generating unit 203 includes:

The determining unit is used to determine the first objective function in the second objective function based on the Lagrangian expression according to the orthogonal transformation matrix, the first objective function and the preset second objective function calculated by the computing unit 202 weight vector, second weight vector and bias value;

a sub-generating unit, configured to generate an EEG classification model based on the orthogonal transformation matrix and the first weight vector, the second weight vector and the bias value determined by the determining unit;

Wherein, the second objective function is:

且所述第二目标函数满足：

And the second objective function satisfies:

为W_k的转置，W_k表示与第k个受试对象相关的第一权向量，V^T为V的转置,V表示第二权向量，b_k表示与第k个受试对象相关的偏置值。ρ _i,k represents the contribution of the i-th sample data of the k-th subject, ε _i,k is the slack variable of x _i,k , C _k and λ are values greater than 0, y _i, represents the k-th The classification result of the i-th sample data of the subjects,

is the transpose of W _k , W _k represents the first weight vector related to the k-th subject, V ^T is the transpose of V, V represents the second weight vector, and b _k represents the k-th subject related to the subject offset value.

Optionally, the sub-generating model is specifically used for:

A classification algorithm for generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value;

The classification algorithm is:

且，α_i,k'和b_k满足等式[K+Z]·α＝Y，在所述等式中，

Among them, when k'=k, Z _i,k =Px _i,k , when k'≠k, Z _i,k =Qx _i,k , and,

And, α _i,k' and b _k satisfy the equation [K+Z]·α=Y, in which,

It should be noted that the apparatus for generating an EEG classification model in this embodiment of the present application may be an independent device, or the apparatus for generating an EEG classification model may also be integrated into electronic devices (such as smartphones, tablet computers, computers, and wearable devices). etc.) in. Optionally, the operating system carried by the device or electronic device that integrates the EEG classification model generating apparatus may be an ios system, an android system, a windows system or other operating systems, which is not limited here.

As can be seen from the above, the embodiment of the present application obtains the sample data of the K subjects, and calculates an orthogonal transformation matrix based on the sample data of the K subjects and the first objective function, and then generates an EEG based on the orthogonal transformation matrix. classification model. Since the orthogonal transformation matrix can be used to transform the respective EEG information of the K subjects into the correlation information between the K subjects, the EEG classification model generated based on the orthogonal transformation matrix can adapt to K The difference of EEG information between subjects makes the generated EEG classification model applicable to K subjects. Since multiple subjects can share the same EEG classification model, compared with the traditional scheme, The present application can maintain an EEG classification mode for multiple subjects, which effectively saves the maintenance cost of the EEG classification model. Further, since the process of generating the EEG classification model is based on the sample data of multiple subjects for training, and the correlation information between multiple subjects is used to revise the training process, therefore, compared with the traditional Based on the EEG classification model obtained by training the sample data of a single subject, the EEG classification model generated in the embodiment of the present application can improve the EEG classification accuracy of each subject to a certain extent.

Embodiment 3

An embodiment of the present application provides an electronic device, please refer to FIG. 3 , the electronic device in the embodiment of the present application includes: a memory 301 , one or more processors 302 (only one is shown in FIG. A computer program that can be run on a processor. The memory 301 is used for storing software programs and modules, and the processor 302 executes various functional applications and data processing by running the software programs and units stored in the memory 301 . Specifically, the processor 302 implements the following steps by running the above-mentioned computer program stored in the memory 301:

Obtaining sample data of K subjects, wherein the sample data includes: classified EEG information and classification results of the corresponding EEG information, and the K is greater than or equal to 2;

Based on the sample data of the K subjects and a preset first objective function, an orthogonal transformation matrix that makes the first objective function take a minimum value is calculated, wherein the first objective function is an orthogonal transformation with The matrix is a function related to the EEG information of the K subjects, and the orthogonal transformation matrix is used to transform the respective EEG information of the K subjects into the correlation between the K subjects information;

An EEG classification model is generated based on the orthogonal transformation matrix, so as to classify the EEG information of any one of the K subjects by using the EEG classification model.

Assuming that the above is the first possible implementation manner, in the second possible implementation manner provided on the basis of the first possible implementation manner, the first objective function is:

所述第一目标函数中的N_k表示第k个受试对象的样本数据个数，N_l表示第l个受试对象的样本数据个数，P表示正交变换矩阵，P^T为P的转置，

表示x_i,k的转置，x_i,k表示第k个受试对象的第i个样本数据中的脑电信息，

表示x_j,l的转置,x_j,l表示第l个受试对象的第j个样本数据中的脑电信息；

In the first objective function, N _k represents the number of sample data of the k-th subject, N _l represents the number of sample data of the l-th subject, P represents an orthogonal transformation matrix, and P ^T is the Transpose,

represents the transpose of x i, _k , x _i,k represents the EEG information in the i-th sample data of the k-th subject,

represents the transpose of x j, _l , x _j,l represents the EEG information in the jth sample data of the lth subject;

Described based on the sample data of the K subjects and the preset first objective function, the orthogonal transformation matrix that calculates the minimum value of the first objective function is:

Under the condition that PP ^T =I is satisfied, based on the sample data of the K subjects, calculate P that minimizes the first objective function, where I is an identity matrix.

In a third possible implementation manner provided on the basis of the above-mentioned second possible implementation manner, the generating an EEG classification model based on the orthogonal transformation matrix includes:

According to the obtained orthogonal transformation matrix, the first objective function and the preset second objective function, determine the first weight vector, the second weight vector and the second objective function based on the Lagrangian expression offset value;

generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value;

Wherein, the second objective function is:

且所述第二目标函数满足：

And the second objective function satisfies:

为W_k的转置，W_k表示与第k个受试对象相关的第一权向量，V^T为V的转置,V表示第二权向量，b_k表示与第k个受试对象相关的偏置值。ρ _i,k represents the contribution of the i-th sample data of the k-th subject, ε _i,k is the slack variable of x _i,k , C _k and λ are values greater than 0, y _i,k represents the ith The classification result of the ith sample data of k subjects,

is the transpose of W _k , W _k represents the first weight vector related to the k-th subject, V ^T is the transpose of V, V represents the second weight vector, and b _k represents the k-th subject related to the subject offset value.

In the fourth possible implementation manner provided on the basis of the above-mentioned third possible implementation manner, the first weight vector, the second weight vector and the determined based on the orthogonal transformation matrix and the The bias value generates an EEG classification model as follows:

A classification algorithm for generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value;

The classification algorithm is:

Among them, when k'=k, Z _i,k =Px _i,k , when k'≠k, Z _i,k =Qx _i,k , and,

且，α_i,k'和b_k满足等式[K+Z]·α＝Y，在所述等式中，

And, α _i,k' and b _k satisfy the equation [K+Z]·α=Y, in which,

Optionally, as shown in FIG. 3 , the above electronic device may further include: one or more input devices 303 (only one is shown in FIG. 3 ) and one or more output devices 304 (only one is shown in FIG. 3 ) . The memory 301 , the processor 302 , the input device 303 and the output device 304 are connected by a bus 305 .

It should be understood that, in this embodiment of the present application, the processor 302 may be a central processing unit (Central Processing Unit, CPU), and the processor may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), special-purpose processors An integrated circuit (Application Specific Integrated Circuit, ASIC), an off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The input device 303 may include a keyboard, a touchpad, a fingerprint sensor (for collecting user's fingerprint information and fingerprint direction information), a microphone, and the like, and the output device 304 may include a display, a speaker, and the like.

Memory 304 may include read-only memory and random access memory, and provides instructions and data to processor 301 . A portion or all of memory 304 may also include non-volatile random access memory. For example, memory 304 may also store device type information.

As can be seen from the above, the present application obtains the sample data of K subjects, and calculates an orthogonal transformation matrix based on the sample data of K subjects and the first objective function, and then generates an EEG classification model based on the orthogonal transformation matrix. . Since the orthogonal transformation matrix can be used to transform the respective EEG information of the K subjects into the correlation information between the K subjects, the EEG classification model generated based on the orthogonal transformation matrix can adapt to K The difference of EEG information between subjects makes the generated EEG classification model applicable to K subjects. Since multiple subjects can share the same EEG classification model, compared with the traditional scheme, The present application can maintain an EEG classification mode for multiple subjects, which effectively saves the maintenance cost of the EEG classification model. Further, since the process of generating the EEG classification model is based on the sample data of multiple subjects for training, and the correlation information between multiple subjects is used to revise the training process, therefore, compared with the traditional Based on the EEG classification model obtained by training the sample data of a single subject, the EEG classification model generated in the embodiment of the present application can improve the EEG classification accuracy of each subject to a certain extent.

Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. Module completion, that is, dividing the internal structure of the above device into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment may be integrated in one processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit, and the above-mentioned integrated units may adopt hardware. It can also be realized in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing from each other, and are not used to limit the protection scope of the present application. For the specific working processes of the units and modules in the above-mentioned system, reference may be made to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

In the foregoing embodiments, the description of each embodiment has its own emphasis. For parts that are not described or described in detail in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application.

In the embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the system embodiments described above are only illustrative. For example, the division of the above-mentioned modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined. Either it can be integrated into another system, or some features can be omitted, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.

The units described above as separate components may or may not be physically separated, and components shown as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

If the above-mentioned integrated units are implemented in the form of software functional units and sold or used as independent products, they may be stored in a computer-readable storage medium. Based on this understanding, the present application realizes all or part of the processes in the methods of the above-mentioned embodiments, and can also be completed by instructing the relevant hardware through a computer program. The above-mentioned computer program can be stored in a computer-readable storage medium. The computer program When executed by a processor, the steps of each of the above method embodiments can be implemented. Wherein, the above-mentioned computer program includes computer program code, and the above-mentioned computer program code may be in the form of source code, object code form, executable file or some intermediate form. The above-mentioned computer-readable medium may include: any entity or device capable of carrying the above-mentioned computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random Access memory (RAM, RandomAccess Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the above-mentioned computer-readable media may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer-readable media does not Included are electrical carrier signals and telecommunication signals.

The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the above-mentioned embodiments, those of ordinary skill in the art should understand that the above-mentioned embodiments can still be used for The recorded technical solutions are modified, or some technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the application, and should be included in the present application. within the scope of protection of the application.

Claims (8)

1. A method for generating an EEG classification model, comprising: Obtaining sample data of K subjects, wherein the sample data includes: classified EEG information and classification results of the corresponding EEG information, and the K is greater than or equal to 2; Based on the sample data of the K subjects and a preset first objective function, an orthogonal transformation matrix that makes the first objective function take a minimum value is calculated, wherein the first objective function is an orthogonal transformation with The matrix is a function related to the EEG information of the K subjects, and the orthogonal transformation matrix is used to transform the respective EEG information of the K subjects into the correlation between the K subjects information; generating an EEG classification model based on the orthogonal transformation matrix, so as to use the EEG classification model to classify the EEG information of any one of the K subjects; The first objective function is:

In the first objective function, N _k represents the number of sample data of the k-th subject, N ₁ represents the number of sample data of the first subject, P represents an orthogonal transformation matrix, and P ^T is the Transpose,

represents the transpose of x i, _k , x _{i, k} represents the EEG information in the i-th sample data of the k-th subject,

represents the transpose of x j, _l , x _{j, l} represents the EEG information in the jth sample data of the lth subject; Described based on the sample data of the K subjects and the preset first objective function, the orthogonal transformation matrix that calculates the minimum value of the first objective function is: Under the condition that PP ^T =I is satisfied, based on the sample data of the K subjects, calculate P that minimizes the first objective function, where I is an identity matrix. 2. The method for generating an EEG classification model according to claim 1, wherein the generating an EEG classification model based on the orthogonal transformation matrix comprises: According to the obtained orthogonal transformation matrix, the first objective function and the preset second objective function, determine the first weight vector, the second weight vector and the second objective function based on the Lagrangian expression offset value; generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value; Wherein, the second objective function is:

And the second objective function satisfies:

ρ _i,k represents the contribution of the i-th sample data of the k-th subject, ε _i,k is the slack variable of x _i,k , C _k and λ are values greater than 0, y _i,k represents the ith The classification result of the ith sample data of k subjects,

is the transpose of W _k , W _k represents the first weight vector related to the k-th subject, V ^T is the transpose of V, V represents the second weight vector, and b _k represents the k-th subject related to the subject offset value. 3 . The method for generating an EEG classification model according to claim 2 , wherein the first weight vector, the second weight vector and the partial weight are determined based on the orthogonal transformation matrix and the The value-generating EEG classification model is: A classification algorithm for generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value; The classification algorithm is:

Among them, when k'=k, Z _i,k =Px _i,k , when k'≠k, Z _i,k =Qx _i,k , and,

And, α _{i, k'} and b _k satisfy the equation [K+z]·α=Y, in which,

4. A device for generating an EEG classification model, comprising: an obtaining unit, configured to obtain sample data of K subjects, wherein the sample data includes: classified EEG information and a classification result of the corresponding EEG information, and the K is greater than or equal to 2; A computing unit for calculating an orthogonal transformation matrix that makes the first objective function take a minimum value based on the sample data of the K subjects and a preset first objective function, wherein the first objective function is a function related to an orthogonal transformation matrix and the EEG information of the K subjects, and the orthogonal transformation matrix is used to transform the respective EEG information of the K subjects into the K subjects Correlation information between; a generating unit, configured to generate an EEG classification model based on the orthogonal transformation matrix, so that the EEG classification model is used to classify the EEG information of any one of the K subjects; The first objective function is:

In the first objective function, N _k represents the number of sample data of the k-th subject, N ₁ represents the number of sample data of the first subject, P represents an orthogonal transformation matrix, and P ^T is the Transpose,

represents the transpose of x i, _k , x _{i, k} represents the EEG information in the i-th sample data of the k-th subject,

represents the transpose of x j, _l , x _{j, l} represents the EEG information in the jth sample data of the first subject; The computing unit is specifically configured to: under the condition that ^PPT =I is satisfied, based on the sample data of the K subjects, calculate P that makes the first objective function take the minimum value, wherein the I is identity matrix. 5. The device for generating an EEG classification model according to claim 4, wherein the generating unit comprises: The determining unit is used to determine the first objective function in the second objective function based on the Lagrangian expression according to the orthogonal transformation matrix, the first objective function and the preset second objective function calculated by the computing unit. a weight vector, a second weight vector and a bias value; a sub-generating unit, configured to generate an EEG classification model based on the orthogonal transformation matrix and the first weight vector, the second weight vector and the bias value determined by the determining unit; Wherein, the second objective function is:

And the second objective function satisfies:

ρ _i,k represents the contribution of the i-th sample data of the k-th subject, ε _i,k is the slack variable of x _i,k , C _k and λ are values greater than 0, y _i,k represents the ith The classification result of the ith sample data of k subjects,

is the transpose of W _k , W _k represents the first weight vector related to the k-th subject, V ^T is the transpose of V, V represents the second weight vector, and b _k represents the k-th subject related to the subject offset value. 6. The device for generating an EEG classification model according to claim 5, wherein the sub-generating model is specifically used for: A classification algorithm for generating an EEG classification model based on the orthogonal transformation matrix and the determined first weight vector, the second weight vector and the bias value; The classification algorithm is:

Among them, when k'=k, Z _i,k =Px _i,k , when k'≠k, Z _i,k =Qx _i,k , and,

And, α _{i, k'} and b _k satisfy the equation [K+z]·α=Y, in which,

7. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the computer program as claimed in the claims Steps of any one of 1 to 3 of the method. 8. A computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented .

Images