KR20190030612A - System for providing subject-independent brain-computer interface and method thereof - Google Patents

Abstract

The present invention relates to a user-independent brain-computer interface (BCI) providing system and a method thereof. The method is executed by a BCI system for providing a user-independent BCI. The method includes the steps of: (a) obtaining brain signals through one or more paradigms from multiple users and using the obtained brain signals to extract the characteristic pattern information of brain signals of each user and storing reference characteristic information including the same in a BCI database; (b) obtaining idle period brain signals of a new user during a preset resting time when the approach of the new user is sensed and extracting the characteristic pattern information from the idle period brain signals to compute target characteristic information; (c) comparing the reference characteristic information stored in the BCI interface with the target characteristic information in order to select the characteristic pattern information of the brain signals having a characteristic pattern most similar to the characteristic pattern information of the target characteristic information in the reference characteristic information; and (d) applying pre-transition learning information formed in advance to correspond to the characteristic pattern information of the selected brain signals to the new user. The pre-transition learning information includes the brain wave characteristic information including the average, covariance, and deviation of brain waves as well as spectral power for each frequency band and classifier information for classifying the brain wave pattern.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a system and a method for providing a user-

The present invention relates to a system and method for providing a user-independent brain-computer interface that does not require a repetitive brain signal measurement process by extracting a brain signal feature pattern similar to a brain signal feature pattern of a new user through a large-capacity BCI database.

The Brain-Computer Interface (BCI) is an interface system that controls external devices without direct body movement through the connection between the human brain and the machine. BCI can be used for rehabilitation purposes as well as being a life support tool for people who can not move the body or patients who have difficulty moving the body. In addition, BCI can be used for a variety of purposes such as the entertainment industry, concentration-enhancing tools, drones control, and exercise applications based on the brain signal information of the general public, not only the patient.

There are two main methods of measuring brain signals in BCI. It is divided into an invasive method of opening the skull of the brain to acquire brain waves and a non-invasive method of acquiring brain waves outside the brain's skull. Electrocorticography (ECoG) is a method that is commonly used as an invasive method. Electrodes are placed on the skull to measure brain waves. The invasive method can acquire a more accurate and clean signal in a noninvasive method, but it has a limit in that an electrode for measuring brain waves must be inserted into the body. Electroencephalography (EEG), which measures electrical signals according to the activity of the brain, is widely known as a non-invasive method. In addition, changes in brain oxygen saturation (Magnetoencephalography, MEG) Functional magnetic resonance imaging (fMRI) or near-infrared spectroscopy (NIRS) are known.

 BCI technology measures external stimuli or spontaneous brain signals. At this time, the external stimulus can be divided into an involuntary method in which a specific signal given from the outside such as vision or hearing is reflected to the brain, or a voluntary method in which the brain reflects the motion.

The characteristics of involuntary EEGs induced by external stimuli include Steady-State Visual Evoked Potential (SSVEP) and Evnet-Related Potential (ERP). SSVEP is a method that allows a user to stimulate a specific frequency when the user strobes a visual stimulus that blinks at a specific frequency. The ERP is a response to a stimulus when the user repeatedly focuses on the stimulus It is a method to be induced in the occipital lobe of the brain. On the other hand, Sensory Motor Rhythm (SMR) is a feature of the brain wave that is spontaneously expressed when the user concentrates or imagines motion without external stimulus. In the SMR, when the user performs a motor imagery (MI) in which the user imagines the movement of a specific part in the somatosensory motor cortex, the signal changes in the left and right parts of the brain, Can be observed. It can be seen that the signal decreases in the frequency domain mainly at 8 ~ 12Hz (Mu rhythm) 18 ~ 25Hz (Beta rhythm), which is called Event-Related Desynchronization (ERD).

 The brain signal acquired from the brain consists of five signal processing processes. 1 is a flowchart illustrating signal processing operations of a conventional brain-computer interface system.

1, the signal processing process of the brain-computer interface system includes a brain signal extracting step S10 for collecting brain signals from at least one paradigm, a brain signal preprocessing step for preprocessing the extracted brain signals S20, A classifier generation step (S40) for classifying or predicting a class of a brain signal based on the extracted feature information, and a classifying step (S40) for classifying or predicting a result of the user And a performance evaluation step (s50) of recognizing the performance of the system.

The conventional BCI system performs a calibration process in the brain signal extraction step. This is because, in order to extract the feature of the BCI system, it is necessary to find the basic EEG feature pattern or classifier parameter value of the user.

This calibration process is performed each time considering the electrode resistance at the time of measurement, the change of the electrode position, the change of the state of the user's brain signal, etc. even if the user has previously measured the brain signal. Such a calibration process is essential, but needs to be minimized for the use of a more practical and economical BCI system.

In the case of the behavioral paradigm, the BCI user executes each class for about 20 minutes each time to build a stable BCI system, acquires brain waves related to the operation imagination, and adapts them to each class.

Recently, in order to reduce the calibration time, a method of extracting common features expressed by all users based on a large brain signal database has been proposed, and a user-independent BCI system which does not require calibration has been studied.

However, since the existing knowledge base is formed by integrating all users' information into one model (EEG feature or pattern) simply by using the existing database, it is applied to all persons having different characteristics It is difficult to form a common feature as much as possible and there is a limitation in that even if the common feature is applied to a new user, it is highly likely that the common feature is a feature that is not suitable for the user.

Korean Patent No. 10-1389015 (entitled " Electroencephalogram Analysis System Using Amplitude Modulated Steady State Visual Induced Potential Visual Stimulation "

In one embodiment of the present invention, a BCI database is constructed using feature pattern information of brain signals acquired from a plurality of users, and feature pattern information of a brain signal having information similar to a brain signal of a new user is selected And to provide a user-independent brain-computer interface providing system and method capable of learning a classifier using pre-transition learning information corresponding to feature pattern information of a selected brain signal.

It is to be understood, however, that the technical scope of the present invention is not limited to the above-described technical problems, and other technical problems may exist.

According to an aspect of the present invention, there is provided a method for providing a user-independent brain-computer interface, the method comprising: providing a user-independent brain-computer interface implemented by a brain-computer interface (BCI) A method for providing an interface, comprising the steps of: a) obtaining a brain signal from at least one or more paradigms from a plurality of users, extracting feature pattern information of a brain signal for each user using the obtained brain signal, Storing reference feature information including pattern information in a BCI database; b) acquiring a resting brain signal for a new user during a predetermined rest period when the access of a new user is detected, and extracting the characteristic pattern information from the resting brain signal to calculate subject characteristic information; c) comparing the reference feature information stored in the BCI database with the target feature information, and selecting feature pattern information of a brain signal having a feature pattern most similar to the feature pattern information of the target feature information among the reference feature information; And d) applying pre-transition learning information formed in advance corresponding to the feature pattern information of the selected brain signal to a new user, wherein the pre-transition learning information includes at least one of a mean, a covariance, a deviation, a spectral power , And classifier information for classifying the EEG pattern.

A system for providing a user-independent brain-computer interface for controlling an external device using a brain signal acquired from a user, the system comprising: a user-independent brain-computer interface Brain-Computer Interface, BCI); And a processor for executing the program, wherein the processor obtains a brain signal from at least one or more paradigms from a plurality of users by executing the program, And the reference feature information including the feature pattern information of the brain signal for each user is stored in the BCI database. If a new user's approach is detected, Extracts feature pattern information from the resting brain signal, calculates target feature information, compares reference feature information stored in the BCI database with the target feature information, and extracts features of the target feature information from the reference feature information The most similar feature to the pattern information And applying the pre-transition learning information formed in advance corresponding to the feature pattern information of the selected brain signal to the new user, wherein the pre-transition learning information includes at least one of a mean, a covariance, a deviation , EEG feature information including spectral power per frequency band, and classifier information for classifying EEG patterns.

According to the present invention, the user can immediately use the BCI system after attaching the brain signal electrodes, thereby eliminating the need for the user to measure the brain signal every time before using the BCI system. It is possible to greatly improve popularization and practicality.

Therefore, according to the present invention, the characteristic pattern information of the brain signal suitable for the user is selectively applied to the BCI database, thereby eliminating the existing calibration process and reducing the time required for preparation of the BCI system, It can easily access the BCI system without stimulation and express external brain signals to control external devices related to entertainment, education, and rehabilitation.

1 is a flowchart illustrating signal processing operations of a conventional brain-computer interface system.
2 is a diagram illustrating a configuration of a brain-computer interface system that provides a user-independent brain-computer interface according to an embodiment of the present invention.
3 is a block diagram illustrating the configuration of the processor of Fig.
4 is a flowchart illustrating a method of providing a user-independent brain-computer interface according to an embodiment of the present invention.
5 is a diagram for explaining the similarity determination process of FIG.
6 is a diagram illustrating a process of performing a frequency-space feature reconstruction algorithm for implementing a method for providing a user-independent brain-computer interface according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, which will be readily apparent to those skilled in the art. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and similar parts are denoted by like reference characters throughout the specification.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "electrically connected" with another part in between . Also, when an element is referred to as "including" an element, it is to be understood that the element may include other elements as well as other elements, And does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, parts, or combinations thereof.

In this specification, the term " part " means a unit realized by hardware or software, a unit realized by using both, and one unit may be realized by using two or more hardware, The above units may be realized by one hardware.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

2 is a diagram illustrating a configuration of a system for providing a user-independent brain-computer interface according to an embodiment of the present invention.

2, a system 100 for providing a user-independent brain-computer interface includes a memory 110, a processor 120, and a database 130.

The memory 110 records a program for performing a method of providing a user-independent brain-computer interface. In addition, the processor 120 performs a function of temporarily or permanently storing data to be processed. Herein, the memory 110 may include volatile storage media or non-volatile storage media, but the scope of the present invention is not limited thereto.

The processor 120 controls the entire process of providing a user-independent brain-computer interface providing method. Each step performed by the processor 120 will be described later with reference to FIG. 3 and FIG.

Here, the processor 120 may include any kind of device capable of processing data, such as a processor. Herein, the term " processor " may refer to a data processing apparatus embedded in hardware, for example, having a circuit physically structured to perform a function represented by a code or an instruction contained in the program. As an example of the data processing apparatus built in hardware, a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, an application-specific integrated circuit (ASIC) circuit, and a field programmable gate array (FPGA), but the scope of the present invention is not limited thereto.

The database 130 stores accumulated data while performing a method of providing a user-independent brain-computer interface. In particular, the database 130 may be a large-capacity BCI database that stores reference feature information including various feature patterns by using feature pattern information of brain signals acquired from a plurality of users.

In addition, the user-independent brain-computer interface providing system 100 further includes communication interface means for collecting brain signals or outputting control signals of external devices, a process for signal processing, and a display means for outputting results can do.

3 is a block diagram illustrating the configuration of the processor of Fig.

Referring to FIG. 3, the processor 120 includes an information management unit 121, an information extraction unit 122, a similarity determination unit 123, an information application unit 124, and a control unit 125.

The information management unit 121 generates and stores a large amount of reference feature information. That is, feature pattern information is extracted using brain signals acquired from a plurality of users, reference feature information including feature pattern information for each user is generated and stored in the BCI database.

The information extracting unit 122 acquires a resting brain signal for a new user during a predetermined rest period when a new user's approach is detected, and extracts the feature pattern information from the resting brain signal to calculate the feature information.

The similarity determination unit 123 compares the reference feature information of the BCI database and the target feature information of the new user to determine similarity.

The information application unit 124 applies the dictionary transition learning information formed in advance in correspondence with the feature pattern information of the brain signal determined to be similar to the feature information of the new user in the similarity determination unit 123 to the new user.

The control unit 125 applies the selected EEG feature information and classifier information to the new user through the information application unit 124, classifies the brain signals measured from the real time new user, and provides control information for controlling the external device .

FIG. 4 is a flowchart illustrating a method for providing a user-independent brain-computer interface according to an embodiment of the present invention, and FIG. 5 is a view for explaining the similarity determination process of FIG.

4 and 5, a method for providing a user-independent brain-computer interface includes acquiring brain signals from a plurality of users through at least one paradigm, and using the brain signals of a plurality of users, After extracting the pattern information, the reference feature information including the feature pattern information of the brain signal for each user is stored in the BCI database (S410).

In this case, the paradigm can be the experimental environment such as SSVEP, MI, ERP, etc., and the BCI system includes not only class (classifier) classifying the motion imagination (MI) of the right hand, left hand and foot but also paradigms such as SSVEP and ERP It is assumed that a classifier for classifying patterns is provided.

The BCI system acquires a resting brain signal for a new user for a predetermined rest period (for example, about one minute) when a new user's approach is detected, extracts feature pattern information from a resting brain signal, (S420). At this time, the target feature information is obtained by extracting the EEG feature information through the preprocessing process and the feature extraction process on the EEG acquired while the user rests, and the extracted EEG feature information includes at least one of the mean EEG and the covariance.

In the conventional BCI system, the user's brain signal is obtained by performing the experiment for about 20 minutes. However, the BCI system according to the embodiment of the present invention uses only the resting brain signal measured during the break time of about 1 minute. Therefore, the user using the BCI system can perform the brain signal extraction process more conveniently and quickly.

5, the BCI system judges the mutual similarity between the target feature information acquired from the new user and the reference feature information stored in the BCI database, and selects the most similar feature from the reference feature information to the feature pattern information of the target feature information Characteristic pattern information of the brain signal having the pattern is selected (S430).

The BCI system can use the distance information as a criterion for determining the similarity between the object feature information and the reference feature information. For example, the average distance between the EEG characteristics (e.g., at least one of the mean and the covariance) of the reference feature information in the BCI database and the EEG characteristics (e.g., at least one of the mean and covariance) of the feature information, The degree of change of the covariance value, and the like can be used.

More specifically, the distance information between the EEG characteristics of the object feature information and the reference feature information may be an Euclidean distance between the center points of each feature distribution, or may be based on a Kullback-Leibler divergence technique Lt; RTI ID = 0.0 > feature distribution. ≪ / RTI >

Accordingly, the BCI system according to an embodiment of the present invention extracts characteristics of a brain signal most similar to a feature of a brain signal of a new user in the BCI database, By applying the information (brain wave feature or pattern) directly to the new user (S440), the calibration process can be eliminated. At this time, the pre-transition learning information includes brain wave feature information including mean, covariance, deviation, spectral power per frequency band and classifier information for classifying EEG patterns.

The BCI system provides control information for controlling an external device by applying the EEG feature or classifier information selected in the BCI database to a new user (S450).

6 is a diagram illustrating a process of performing a frequency-space feature reconstruction algorithm for implementing a method for providing a user-independent brain-computer interface according to an embodiment of the present invention.

First, the pre-input information of the frequency-space feature reconstruction algorithm for implementing the user-independent brain-computer interface providing method is as follows.

이며, M채널과 T 샘플포인트들을 가지는 단일 시도 뇌파는

이고,

에서 클래스 라벨 매칭은

이다. 그리고, K는 주파수 필터 뱅크의 수, m은 공간 패턴 학습 알고리즘에서 공간 패턴의 반수(half number), P는 선택된 공간 필터 뱅크의 수, L은 CNN 레이어의 수이다. The set of training data in the source domain is {X, Y} and the set of single trial EEGs in total trial number (D)

, And a single attempted EEG with M channels and T sample points

ego,

Class label matching in

to be. K is the number of frequency filter banks, m is a half number of spatial patterns in the spatial pattern learning algorithm, P is the number of selected spatial filter banks, and L is the number of CNN layers.

6, a frequency-space feature reconstruction algorithm for implementing a method for providing a user-independent brain-computer interface provides a user-independent brain-computer interface based on MI paradigm among various paradigms, (S510), a feature reconstruction process (S520), and a feature fusion and classification process (S530).

The input information generating process (S510) collects the user's brain signals using 20 motion field EEG channels commonly used in MI according to the 10-20 International Nomenclature (10-20 International Nomenclature). Here, the 10-20 International Electrode Arrangement is an internationally well-known arrangement for the arrangement of electrodes placed on the head during EEG (Electroencephalogram) test or experiment.

Frequency-space features reproduction algorithm if the time of defining a plurality of frequency filter bank (b _k), including 8 ~ 30㎐, 11 ~ 20㎐ commonly used in the MI, MI and at the paradigm a start signal for performing the task EEG (For example, 1 second to 3.5 seconds) from a predetermined measurement time.

)를 연결하면,

이 된다. 여기서, N은 학습 피험자들의 수이다.Then, the frequency-space feature reconstruction algorithm connects the brain signals of all users (or subjects) measured. In other words,

),

. Where N is the number of study subjects.

The frequency-spatial feature reconstruction algorithm performs frequency filtering on the integrated EEG samples and extracts filtered EEG signals (E _k ) as shown in Equation (1).

[Equation 1]

이다. In Equation (1), E _k

to be.

Frequency-space features reproduction algorithm common spatial pattern (Common Spatial Pattern, CSP) applying an algorithm to obtain the spatial filter (W _k) as in equation (2), a spatial filter (W _k) for the filtered EEG signal (E _k) And a filtered EEG signal (E _k ), and then takes a log-variance value and extracts a feature vector (V _k ) as shown in Equations (3) and (4).

&Quot; (2) "

&Quot; (3) "

&Quot; (4) "

,

는 W_k에서 첫 번째 m과 마지막 m 컬럼 벡터들이다. In Equation (3)

,

Is the first m and last m column vectors in W _k .

In order to consider spatial characteristics, Common Spatial Pattern (CSP), which is a spatial filtering technique, is an algorithm for designing a spatial filter that maximizes dispersion between classes for two different classes. Maximize.

Frequency-spatial features reproduction algorithm is the extracted feature vector (V _k) frequency filter bank (b _p) and rearranged in the MI value is greater sequence existing frequency filter bank (b _k) to calculate the mutual information amount (MI) from .

The frequency-space feature reconstruction algorithm performs covariance on the signal values obtained through the matrix operation between the spatial filter (W _p ) extracted from the reordered frequency filter bank (b _p ) and the rearranged spatial filter (E _p ) To generate frequency-space input information (C _p ).

&Quot; (5) "

,

,

In the feature reproduction process S520, the frequency-space input information C _p generated from each frequency filter bank b _p is individually learned through the CNN, and the convolution layer (G _p ).

&Quot; (6) "

In the feature fusion and classification process S530, the convolutional features G _p in the L-th convolution layer are reduced to a one-dimensional vector form, and the reduced p 1-dimensional feature vectors are transformed into concatenated F- 7 as a single integrated vector (S ^concat ).

&Quot; (7) "

In the feature fusion and classification process (S530), a total of N class results are output through a fully-connected layer to the concatenated vector S ^concat and N class results are obtained through a soft max classifier as shown in Equation (8) Determine the class with a larger value.

&Quot; (8) "

,

,

,

,

As described above, the method of providing a user-independent brain-computer interface according to an embodiment of the present invention greatly improves the popularization and practicality of BCI technology by eliminating the hassle of measuring the brain signal every time before the user uses the BCI system .

The method of providing a user-independent brain-computer interface according to an embodiment of the present invention described above may be implemented in the form of a recording medium including instructions executable by a computer such as a program module executed by a computer. Such a recording medium includes a computer-readable medium, which may be any available medium that can be accessed by a computer, and includes both volatile and non-volatile media, both removable and non-removable media. The computer-readable medium may also include computer storage media, which may be volatile and non-volatile, implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data , Both removable and non-removable media.

It will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention as defined by the following claims. will be. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

Furthermore, while the methods and systems of the present invention have been described in terms of specific embodiments, some or all of those elements or operations may be implemented using a computer system having a general purpose hardware architecture.

It is intended that the present invention covers the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents. .

100: User-independent brain-computer interface providing system
110: Memory
120: Processor
120: Database

Claims (16)

A method of providing a user-independent brain-computer interface performed by a Brain-Computer Interface (BCI) system,
a) obtaining a brain signal from at least one or more paradigms from a plurality of users, extracting feature pattern information of a brain signal for each user by using the obtained brain signal, Storing feature information in a BCI database;
b) acquiring a resting brain signal for a new user during a predetermined rest period when the access of a new user is detected, and extracting the characteristic pattern information from the resting brain signal to calculate subject characteristic information;
c) comparing the reference feature information stored in the BCI database with the target feature information, and selecting the feature pattern information of the brain signal having the feature pattern most similar to the feature pattern information of the target feature information among the reference feature information; And
d) applying pre-transition learning information previously formed corresponding to the feature pattern information of the selected brain signal to a new user,
Wherein the pre-transition learning information includes EEG feature information including EEG covariance, covariance, variance, spectral power per frequency band, and classifier information for classifying EEG patterns. The method according to claim 1,
Wherein the resting time (t2) in the step b) is set to a short time (t2 <t1) compared to the time (t1) required for acquiring the brain signal in the step a) . The method according to claim 1,
The step b)
Extracting feature pattern information through a preprocessing process and a feature extraction process for the resting brain signal to calculate target feature information,
Wherein the object feature information is an EEG characteristic including an average and covariance of brain waves. The method according to claim 1,
Wherein the step c) determines similarity between the reference feature information and the brain signal of the feature information based on the distance information between the reference feature information and the object feature information. 5. The method of claim 4,
The distance information between the reference feature information and the object feature information may be any one of the Euclidean distance information or the change information of the joint value between the EEG feature in the reference feature information and the center point of each feature distribution of the EEG feature of the object feature information Wherein the computer-readable medium is a user-independent brain-computer interface. 5. The method of claim 4,
The distance information between the reference feature information and the object feature information is obtained by dividing the relative distance between the EEG feature in the reference feature information and each feature distribution of the EEG feature of the object feature information by a Kullback- Leibler divergence technique A method for providing a user-independent brain-computer interface, said method being probabilistic. The method according to claim 1,
The step d)
Learning of the classifier by applying the pre-transition learning information including the EEG feature information and the classifier information to a new user, classifying the real-time brain signal from the new user using the learned classifier, and generating control information for controlling the external device The method further comprising the steps of: The method according to claim 1,
The step a)
a-1) generating EEG samples by combining brain waves acquired from a plurality of users, extracting feature vectors from the EEG samples through frequency filtering and spatial filtering, and extracting mutual information (MI) from the feature vectors Generating frequency-space input information for each frequency information based on the frequency-space input information;
a-2) individually learning the frequency-space input information through a CNN (Convolution Neural Network) characteristic reconstruction algorithm to form a convolution feature through a convolution layer; And
a-3) reducing the dimension of the convolution feature to a low-dimensional feature vector, combining the reduced low-dimensional feature vectors into an integrator vector, Classifying the results of the class into a classifier. &Lt; Desc / Clms Page number 19 &gt; 9. The method of claim 8,
The step (a-1)
Collecting a brain signal of a user using a plurality of EEG channel data according to a predetermined electrode arrangement method;
Defining a plurality of frequency filter banks ( _bk ) using a frequency range used in each paradigm;
Selectively extracting only a brain signal for a predetermined measurement time from a time when a start signal for an EEG task is recognized in at least one paradigm;
Integrating and integrating EEG acquired from a plurality of users, extracting filtered EEG signals (E _k ) through frequency filtering of the integrated EEG samples;
The common spatial pattern on the filtered EEG signal _{(E k) (Common Spatial Pattern} , CSP) by applying the spatial filter algorithm (W _k) to obtain the spatial filter (W _k) and the filtered EEG signal (E _k) between Extracting a feature vector (V _k ) by taking a log-variance value after performing a matrix operation;
Calculating a mutual information amount MI from the extracted feature vector V _k and calculating the frequency filter bank b _p by rearranging the frequency filter banks b _{k in the} order of MI in ascending order; And
Area type information (C _p) - the reorder the frequency filter bank (b _p) a spatial filter (W _p) and the frequency to the covariance for the signal value obtained through matrix calculation between the rearranged spatial filter (E _p) extracted from The computer-readable medium having computer-executable instructions for performing the method. 10. The method of claim 9,
The step a-2)
Wherein the frequency-space input information (C _p ) generated from each frequency filter bank (b _p ) is learned separately via CNN to form a convolution feature (G _p ) in a convolution layer of L layers A method for providing a brain-computer interface. 11. The method of claim 10,
The a-3)
Reducing the convolutional features (G _p ) in the Lth convolution layer to a low dimensional vector form;
Integrating the reduced p low dimensional feature vectors into a single integrated vector (S ^concat ) in a concatenation fusion layer; And
Outputting a total of N class results through the fully-connected layer to the integrated vector S ^concat ; And
And classifying the class with a larger value of the N class results through the classifier. A system for providing a user-independent brain-computer interface for controlling an external device using a brain signal acquired from a user,
A memory in which a program for providing a user-independent brain-computer interface (BCI) is recorded; And
And a processor for executing the program,
Wherein the processor obtains a brain signal from at least one or more paradigms from a plurality of users by executing the program, extracts feature pattern information of a brain signal for each user by using the obtained brain signal, The reference feature information including the feature pattern information of the signal is stored in the BCI database,
Acquiring a resting brain signal for a new user during a predetermined rest period, extracting feature pattern information from the resting brain signal to calculate subject feature information,
Comparing the reference feature information stored in the BCI database with the object feature information to select feature pattern information of a brain signal having a feature pattern most similar to the feature pattern information of the object feature information among the reference feature information,
Applying the pre-transition learning information formed in advance corresponding to the feature pattern information of the selected brain signal to a new user,
Wherein the pre-transition learning information includes brain wave feature information including mean, covariance, deviation, spectral power per frequency band, and classifier information for classifying brain wave patterns. 13. The method of claim 12,
Wherein the processor determines the similarity between the reference feature information and the brain signals of the object feature information based on the distance information between the reference feature information and the object feature information. 14. The method of claim 13,
The distance information between the reference feature information and the object feature information may be any one of the Euclidean distance information or the change information of the joint value between the EEG feature in the reference feature information and the center point of each feature distribution of the EEG feature of the object feature information Wherein the computer-readable medium is a user-independent brain-computer interface providing system. 14. The method of claim 13,
The distance information between the reference feature information and the object feature information is obtained by dividing the relative distance between the EEG feature in the reference feature information and each feature distribution of the EEG feature of the object feature information by a Kullback- Leibler divergence technique A user-independent brain-computer interface providing system that is probabilistically calculated. 13. The method of claim 12,
Wherein the processor is configured to learn the classifier by applying the pre-transition learning information including the EEG feature information and the classifier information to a new user, classify the real-time brain signal from the new user using the classifier, Wherein the control information is generated by a user.

Images