KR101740894B1

KR101740894B1 - Brain-computer interface system and method using spatial filter

Info

Publication number: KR101740894B1
Application number: KR1020150159799A
Authority: KR
Inventors: 정천기; 김준식; 염홍기; 방문석; 김성완; 최현
Original assignee: 대한민국
Priority date: 2015-11-13
Filing date: 2015-11-13
Publication date: 2017-05-29
Also published as: KR20170056295A

Abstract

A brain-machine interface system and method using a spatial filter is disclosed. The brain-machine interface system using the disclosed spatial filter applies a spatial filter of EEG data generated in a spatial filter generation system of brain wave data to new EEG data measured by the new EEG data measuring unit, A spatial filter applying unit for amplifying EEG data, and an operation predicting unit for predicting an intended operation of the subject based on the amplified new EEG data.

Description

Technical Field [0001] The present invention relates to a brain-machine interface system and method using a spatial filter,

A brain-machine interface system and method using a spatial filter is disclosed. More particularly, a spatial filter generation system, a brain-machine interface system, and a brain-machine interfacing method of human brain wave data applicable to a brain-machine interface are disclosed.

The brain-computer interface technique (BCI) is a technique for measuring the brain wave data and recognizing the subject's intention from the brain wave computer interface, so that the computer or the machine can be controlled by thinking. Therefore, by using the above-described technique, a person with general paralysis who can not move his / her body at all can express his / her intention, pick up objects, or control the moving means. Therefore, this technique is a necessary and highly valuable technology for the handicapped.

In addition, brain-machine interface technology is a very useful and ideal user interface (UI) technology for the general public. For example, the technology can be used to control various electronic devices such as a general person changing a channel of a television, adjusting a temperature of an air conditioner, or adjusting a volume of music. Also, since the above technology can be applied to the entertainment field such as a game machine, the military field, or the elderly who are hard to move much, the social and economic ripple effect of the technology is very large.

There are two main ways to implement brain-machine interface technology. The first method is to choose one of several predefined choices. For example, the above method is a method of selecting the direction (front, back, left, right) of the electric wheelchair or selecting letters (A, B, C, etc.). In this case, a classification algorithm that selects one of several possible choices is used. The second method is to predict continuous values. For example, the method is a method of continuously predicting consecutive arm movements rather than discrete values like left and right segmentation. In this case, a regression algorithm is used.

The BCI using the classification algorithm has the advantage of being able to select the desired one, but has a disadvantage in that it can be selected only within a few predetermined choices. However, BCI using the regression algorithm is ideal because it can implement the desired behavior close to infinity through prediction like continuous arm motion. In order to predict the arm motion from the EEG data, the relationship between EEG data and arm motion is mathematically modeled by algorithms such as multiple linear regression or Kalman filter, And applies a mathematical model to predict the velocity or position of each of the x, y, and z axes in real time.

Using this technology, we succeeded in controlling the robot arm in real time by using EEG signals in 2008. In 2012, we will control the robot arm in real-time by using limb paralysis. The experiment was successful. Thus, this technique is a very valuable technology that is absolutely necessary for patients who are uncomfortable or unable to move at all.

However, one of the biggest disadvantages of this technique is that the accuracy is very low. The most prominent study of the Brown University group in the 2012 Nature report shows that despite the long duration of training, the success rate is 20.8 to 62.2% even in the simple task of reaching a single target, . This disadvantage becomes a serious barrier to practical use of the above technology. This problem occurs because we do not know clearly how the brain wave changes with movement, and we can not extract the signal correctly.

One embodiment of the present invention provides a spatial filter generation system of human brain wave data applicable to brain-machine interfaces.

Another embodiment of the present invention provides a brain-machine interface system including a spatial filter application.

Another embodiment of the present invention provides a brain-machine interfacing method using a spatial filter of EEG data.

According to an aspect of the present invention,

EEG data measuring part;

EEG data classifier;

A variance calculating unit of EEG data; And

And a spatial filter generation unit for generating brain wave data.

The EEG data measuring unit may measure the EEG data using electroencephalogram (EEG), magnetoencephalography (MEG), electrocorticogram (ECoG), functional magnetic resonance imaging (fMRI), functional near- : functional near-infrared spectroscopy) or a combination thereof.

The EEG data measuring unit may take the form of a headset or a hairband type.

The spatial filter generating system for EEG data may further include a noise extractor for removing noise from the EEG data measured by the EEG data measuring unit and a feature extracting unit for extracting EEG features from the noise- .

The noise canceller may include a low-pass filter, a high-pass filter, a band-pass filter, a notch filter, or a combination thereof.

The feature extraction unit may use Fourier Transform, Partial Directed Coherence (PDC), Direct Transfer Function (DTF), Independent Component Analysis (ICA), Principle Component Analysis (PCA), or Common Spatial Pattern (CSP).

Wherein the feature extracting unit extracts an intensity or a pattern of a specific frequency band of an EEG that changes according to a movement direction when moving an arm or a foot or imagining a motion; The ratio of the signal from the subject's EEG measured when the eye is open and when the eye is closed (eye open / eyes closed); The response time or size of the P300 component of the EEG measured by the subject when the abnormal stimulus is applied; Brain connectivity of subjects measured at rest; The magnitude or frequency domain of the visual evoked potential (SSVEP) in the occipital lobe when visual stimuli are applied; Or a combination thereof can be extracted as an EEG characteristic.

The EEG data classifier may extract EEG data from the beginning to the end of each operation for each of a plurality of operations, and store EEG data for each of the same operations.

The EEG data measuring unit, the noise removing unit, the feature extracting unit, the EEPR classification unit, the EEPROM dispersion calculating unit, and the EEG data spatial filter generating unit may be configured separately from each other, or two or more may be integrated .

The spatial filter generation system of EEG data may further include a spatial filter storage unit of EEG data for storing a spatial filter of EEP data generated by a spatial filter generation unit of EEG data.

According to another aspect of the present invention,

New EEG data measurement part;

A spatial filter applying unit for applying the spatial filter of the EEG data generated in the spatial filter generating system of the EEG data to the new EEG data measured by the new EEG data measuring unit to amplify the new EEG data; And

And a motion prediction unit for predicting an intended motion of the subject based on the amplified new EEG data.

According to another aspect of the present invention,

Acquiring brain wave data;

Classifying the obtained EEG data;

Calculating two or more variances from the classified brain wave data;

Generating a spatial filter of EEG data from the two or more variances;

Acquiring new EEG data;

Amplifying the new brain wave data by applying a spatial filter of the generated brain wave data to the acquired new brain wave data; And

And a step of predicting the intended action of the subject based on the amplified new EEG data.

The acquiring of the EEG data may include measuring EEG data, removing noise from the measured EEG data, and extracting EEG features from the noise-removed EEG data.

The step of predicting the intended motion of the subject based on the amplified new EEG data may be performed using a regression analysis.

The system for generating a spatial filter of EEG data according to an embodiment of the present invention is a brain-machine interface system in which prediction accuracy of a continuous subject is predicted (for example, . In addition, the spatial filter generation system of the EEG data can be applied to a disabled person or an elderly person who is inconvenienced in the movement of the body, and can be utilized to help them live comfortably. Also, the system can be used as a communication method Do. In addition, it can be used for controlling all electronic devices such as changing a channel of a television, adjusting a temperature of an air conditioner, adjusting a music volume, etc. This has a ripple effect enough to replace the user environment of all currently used electronic devices, The potential marketability is explosive. In addition, it is expected that it can be applied to various entertainment products that can control a computer by thinking instead of a joystick or a button of a game machine.

1 is a diagram illustrating an apparatus for generating spatial filter of brain wave data according to an embodiment of the present invention.
FIG. 2 illustrates a brain-mechanical interface system and a machine connected thereto according to an embodiment of the present invention.
FIG. 3 is a diagram illustrating a brain-machine interfacing method according to an embodiment of the present invention.
4 is a graph showing detected EEG data when the subject imagines an arm motion in four directions.
FIG. 5 is a graph showing an average brain response of nine subjects obtained through the brain-mechanical interface system according to an embodiment of the present invention in color when imagining an arm motion in four directions.
6 is a graph showing prediction accuracy of arm motion of the brain-machine interface system manufactured in Example 1 and Comparative Example 1, respectively.

Hereinafter, a system for generating a spatial filter of brain wave data according to an embodiment of the present invention will be described in detail with reference to the drawings.

1 is a diagram illustrating a system 100 for generating spatial filter of EEG data according to an embodiment of the present invention.

Referring to FIG. 1, a spatial filter generation system 100 for generating EEG data according to an embodiment of the present invention includes an EEG data measuring unit 101, an EEG data classifying unit 102, a EEG dispersion calculator 103, And a spatial filter generating unit 104 of EEG data.

The EEG data measuring unit 101 may take the form of a headset or a hairband type.

The EEG data measuring unit 101 may measure EEG data such as electroencephalogram (EEG), magnetoencephalography (MEG), electrocorticogram (ECoG), functional magnetic resonance imaging (fMRI) Functional near-infrared spectroscopy (fNIRS), or a combination thereof.

The electroencephalogram (EEG) is an examination for recording the electrical activity of the brain by attaching an electrode to the scalp and has advantages of temporal resolution and portability. The time resolution indicates how often the sensor records an image of a particular region, which is 16 days for Landsat Thematic Mapper (TM) 4 and 5.

The EEG data measuring unit 101 includes a plurality of EEG electrodes (not shown) exposed to the inside of the subject, a printed circuit board (not shown) disposed inside the outside, (Not shown), a wireless communication module (not shown), and / or a power supply module (not shown) mounted on the board.

The portion of the brain-wave data measuring unit 101 excluding the portion of the subject's face that is in contact with the forehead and / or the scalp may include a shape memory alloy.

The EEG electrode receives electric power from the power supply module and is electrically connected to the printed circuit board, and contacts the forehead and / or scalp of the subject to detect an electric signal. The EEG electrode may be a plurality of electrodes, and each of the electrodes is also referred to as a channel. That is, if the EEG electrode has 64 electrodes, the number of channels is 64.

The printed circuit board may be a flexible printed circuit board.

The EEG measurement module receives the weak electrical signal detected by the EEG electrode and transmits the weak electrical signal to the wireless communication module without amplifying or amplifying the electrical signal.

The wireless communication module transmits the electric signal (i.e., brain wave data) obtained from the brainwave measurement module to a subsequent EEG / analysis apparatus (not shown). Such a wireless communication module may be a short range communication module, for example, a Bluetooth communication module.

EEG data obtained by the EEG data measuring unit 101 are represented by three kinds of matrices, where the number of rows is the same as the number of channels, and the number of columns is the same as the number of EEG measurements. For example, if the number of channels is 64 and the number of brain wave measurements is 1000 times per second, the brain wave data is represented by a matrix of 64 x 1000 per each position vector (i.e., x, y, or z) of three-dimensional spatial coordinates. When the subject imagines a continuous motion, the EEG data (i.e., the elements of the three kinds of matrixes) change continuously or almost continuously according to the measurement time.

The subsequent EEG processing / analysis apparatus may include a noise removing unit (not shown), a feature extracting unit (not shown) and / or an EEG data classifying unit 102.

The noise removing unit removes noise from the EEG data measured by the EEG data measuring unit (101).

The feature extraction unit extracts EEG features from the noise-canceled EEG data.

The EEG features may include firing rate, power of a specific frequency band, event related potential (ERP), steady state visually evoked potential (SSVEP), event related synchronization, event related desynchronization (ERD), default mode network (DMN), or a combination thereof.

Among the event-related potentials, the P300 potential component is considered to be an endogenous potential since it is a component derived from the decision-making process and its occurrence is related to the individual's response to the stimulus, not to the physical properties of the stimulus. Specifically, when the P300 potential component intermittently presents an abnormal stimulus during presentation of a normal visual, auditory, or tactile stimulus, a brain response that exhibits a maximum peak after about 300 ms based on the abnormal stimulus is presented to the parietal lobe parietal lobe).

The steady state visual evoked potential excites the human retina using visual stimuli between 3.5 and 70 Hz and causes brain activity of the same frequency component as the frequency used in the visual stimulus in the occipital lobe of the brain &Lt; / RTI >

The event-related synchronization refers to a phenomenon in which the intensity of a signal increases at 0.5 to 8 Hz in a sensory motor cortex area when an operation such as an arm or foot motion occurs.

The event-related desynchronization refers to a phenomenon in which the intensity of the signal decreases at 9 to 22 Hz in the sensory motor cortex area when an operation such as arm or foot motion occurs.

The default mode network means the connectivity of the brain in a resting state that consumes little energy.

The event-related potential, the steady state visual evoked potential, the event-related synchronization, the event-related desynchronization, and the default mode network are observed in most subjects and are known to have common characteristics for all subjects.

The EEG data classifier 102 classifies the extracted EEG data.

The brain wave data classifying unit 102 may extract brain wave data from the beginning to the end of each of the operations for each of a plurality of operations by time intervals and store the extracted brain wave data for each of the same operations. For example, the same operation means movement performed several times in the same direction when the arm is moved several times in various directions.

The variance calculation unit 103 of the brain wave data calculates two or more variances from the brain wave data classified from the brain wave data classification unit 102.

Hereinafter, the process of calculating the variance of two or more types of EEG data dispersion calculation unit 103 will be described in detail.

For convenience of explanation, the notation the classified brain wave data to X, and is referred to as a brain wave signal of the k-th channel X _k. Also, the subject imagines that he / she has moved the n-th arm in m directions. In this case, the brain wave of the k-th channel when imagining that the arm moved j-th in the i-th direction can be expressed as X _ijk . At this time, the variance calculation unit 103 of the brain wave data can calculate the variance (Var (X _ik )) for j for each i using the following equation (1).

[Equation 1]

If the variance of the brain wave data obtained when imagining that the same operation is executed a plurality of times is referred to as "within variance" and the variance of the brain wave data obtained when each of two or more different operations are executed plural times is called "between variance" The variance calculation unit 103 may calculate the " within variance " for the k-th channel using the following equation (2).

&Quot; (2) "

The variance calculation unit 103 of the brain wave data can calculate the average of j for each i using the following equation (3).

&Quot; (3) "

Then, the spatial filter generation unit 104 of EEG data is divided "between variance" value for each channel (k) calculated in the dispersion calculation unit 103 of the brain wave data to "within variance" value, and calculates the F _k. That is, the spatial filter generation unit 104 of the brain wave data can calculate F _k using the following equation (4).

&Quot; (4) "

The _Fk value is a value indicating how much the EEG signal is different for two or more different operations in the case of the channel k and how much the same is displayed for the same operation.

The spatial filter generating unit 104 of the brain wave data generates F _k Find the value to get the F value. The F value thus obtained is referred to as a spatial filter of EEG data. That is, the spatial filter of the EEG data is represented by three kinds of matrices for a unit time period (for example, one second), where the number of rows is equal to the number of channels, and the number of columns is equal to the number of EEG measurements Do. For example, if the number of channels is 64 and the number of times of EEG measurement is 1000 times per second, the spatial filter of the brain wave data is a matrix of 64 x 1000 per each position vector (i.e., x, y, or z) Is displayed. When the subject imagines continuous motion, the spatial filter of the brain wave data changes continuously or almost continuously according to the measurement time.

The spatial filter generation system 100 of EEG data may further include a spatial filter storage unit 105 of brain wave data. The spatial filter storage unit 105 of the EEG data stores a spatial filter of the brain wave data generated by the spatial filter generating unit 104 of EEG data.

The EEG data analyzing unit 101, the noise removing unit, the feature extracting unit, the EEG data classifying unit 102, the EEG data dispersion calculating unit 103, the EEG data spatial filter generating unit 104, and / The spatial filter storage units 105 of the data may be configured separately from each other, or two or more of them may be integrated.

FIG. 2 is a diagram illustrating a brain-machine interface system 200 and a machine 300 connected thereto according to an embodiment of the present invention.

The brain-machine interface system 200 includes a new EEG data measuring unit 201, a spatial filter applying unit 202 and an operation predicting unit 203.

The new EEG data measuring unit 201 may be the same as the EEG data measuring unit 101 described above.

The spatial filter applying unit 202 applies the spatial filter of the brain wave data generated by the spatial filter generating system 100 of the brain wave data to the new brain wave data measured by the new brain wave data measuring unit 201, . For example, the spatial filter applying unit 202 receives a spatial filter of brain wave data from the spatial filter storage unit 105 of the brain wave data provided in the spatial filter generating system 100 of the brain wave data, (201). Specifically, the spatial filter applying unit 202 multiplies the new brain wave data (for example, three 64 × 1000 matrices) by a space filter (for example, three kinds of 64 × 1000 matrices) of the brain wave data And amplifies the new brain wave data.

The motion predicting unit 203 predicts the intended motion of the subject based on the amplified new EEG data. Therefore, the motion predicting unit 203 can improve the prediction accuracy of the motion intended by the subject.

The motion predicting unit 203 may perform a regression analysis on the new EEG data amplified by the spatial filter applying unit 202 to predict an intended motion of the subject. Here, regression analysis is a statistical technique capable of estimating the effect of at least one independent variable on dependent variables. In a regression with one independent variable, one equation describes a line passing through the distribution of the points showing the combined distribution of independent and dependent variables. This equation has the form Y _i = a + bX _i + e _i . X _i is the value of the independent variable. a is the point of the regression line passing through the Y axis, b is the slope of the regression line, and e _i is the error of the regression line prediction. The regression line is the best line for scattered points. The most useful value in the regression line is the slope b. b represents the effect of independent variables on dependent variables. The most appropriate regression line is explained by the correlation coefficient. The value is denoted by r. However, in the case of a regression model with one or more independent variables, the correlation coefficient is R. The square of the correlation coefficient r ² or R ² represents the degree of variance in the dependent variable described by the independent variable. If the regression coefficient b describes the effect of X on Y, the correlation coefficient shows how the hypothesis model fits the actual data.

The spatial filter applying unit 202 and the motion predicting unit 203 may be configured separately from each other or may be integrated with each other.

The motion predicting unit 203 is connected to the control unit 301 of the machine 300 (for example, a computer) and provides the motion prediction data to the control unit 301. The control unit 301 controls the operation of the operation unit 302 of the machine 300.

Hereinafter, a brain-machine interfacing method according to an embodiment of the present invention will be described in detail.

FIG. 3 is a diagram illustrating a brain-machine interfacing method according to an embodiment of the present invention.

Referring to FIG. 3, a brain-machine interfacing method according to an embodiment of the present invention includes acquiring EEG data (S10), classifying the obtained EEG data (S20), extracting 2 (S30) of generating a spatial filter of EEG data from at least two kinds of variances (S40), acquiring new EEP data (S50), calculating A step S60 of amplifying the new EEG data by applying a spatial filter of the EEG data, and a step S70 of predicting the intended operation of the subject based on the amplified new EEG data.

The step of acquiring EEG data may include acquiring EEG data (S10), removing noise from EEG data of the subject (S20), extracting EEG features from the EEG data of the subject (S30), and classifying the extracted EEG characteristics (S40).

Step S10 of acquiring EEG data includes steps of measuring EEG data (not shown), removing noise from measured EEG data (not shown), and extracting EEG features from the noise-removed EEG data (Not shown).

The brain-machine interfacing method can precisely control the machine according to the subject's intention by measuring and processing brain wave data of the subject. Specifically, the brain-machine interfacing method may control the operation of the machine by providing the machine with the motion prediction data obtained in the operation predicting step S70 (S80).

Hereinafter, the present invention will be described with reference to the following examples, but the present invention is not limited thereto.

Experimental Example 1: Detection of EEG signal

When the subject imagined the arm motion in all four directions (direction 1), after (direction 2), left (direction 3) and right (direction 4), EEG signals were detected respectively. The detected EEG signals are graphically shown in FIG. The brain wave measuring device used was Neuroscan's Synamp2 model, and there were 64 measurement sensors, and the measurement speed was 1,000 times per second.

Referring to FIG. 4, EEG signals corresponding to arm movements in four directions are completely different from each other. From this result, it can be seen that it is possible to construct the brain-machine interface by using an EEG signal (i.e., brain wave data). Referring to FIG. 4, the distribution of the EEG signals obtained when the same operation is repeated several times is smaller than that of the EEG signals obtained when the two operations are repeated several times. From this result, it can be seen that it is possible to construct a brain-machine interface having a high prediction accuracy of the operation by using the dispersion of the brain wave signal (i.e., brain wave data).

Experimental Example 2: Create a spatial filter of EEG data

EEG data were obtained for each of the nine subjects when each of the subjects imagined continuous arm motion in four directions. The brain wave measuring device used was Neuroscan's Synamp2 model, and there were 64 measurement sensors, and the measurement speed was 1,000 times per second. Thereafter, spatial filters of brain wave data were generated according to Equations 1 to 4 using the EEG data obtained for the nine subjects. Then, the spatial filters of the EEG data are averaged to obtain a spatial filter (F) of the average brain wave data. The spatial filters are displayed in color, and the results are shown in FIG. 5 as brain shapes. In FIG. 5, it is set that the larger the F value indicates the red color, and the smaller the F value indicates the blue color. That is, as the F value in the color indicating bar of FIG. 5 increases, the upper color appears, and as the F value decreases, the lower color appears.

Referring to FIG. 5, it can be seen that the F value is high in the frontal cortex portion related to cognition and the posterior parietal cortex portion related to the position recognition. From this result, it can be seen that it is possible to construct a brain-machine interface having a high prediction accuracy of the motion intended by the subject using the spatial filter of EEG data.

Example 1: Configuration of brain-machine interface system

EEG data were obtained for each of the nine subjects when each of the subjects imagined continuous arm motion in four directions. The brain wave measuring device used was Neuroscan's Synamp2 model, and there were 64 measurement sensors, and the measurement speed was 1,000 times per second. Then, the EEG data obtained by applying the spatial filter of the brain wave data obtained in Experimental Example 2 to the EEG data obtained above, and regression analysis was performed on the amplified EEG data to obtain a predicted operation. Thereafter, a correlation coefficient between the predicted operation and the actual operation was obtained. The correlation coefficient exists for each position vector (i.e., x, y, or z) of the three-dimensional spatial coordinates. Here, the correlation coefficient is an index indicating how linear the distribution of the two variables is. In other words, the correlation coefficient can visually indicate the degree to which a clean straight line is drawn when a combination of two variables is displayed on a two-dimensional coordinate axis. Specifically, assuming that two variables x and y are given as (x ₁ , ..., x _i , ..., x _n ) (y ₁ , ..., y _i , ..., y _n ) Can be expressed by the following equation (5).

&Quot; (5) "

Comparative Example 1: Configuration of brain-machine interface system

First, EEG data were obtained from 9 subjects in the same manner as in Example 1. Then, the spatial filter of the EEG data generated in Experimental Example 2 was applied to the obtained EEG data, and regression analysis was performed on the obtained EEG data itself to obtain a predicted operation. Thereafter, a correlation coefficient between the predicted operation and the actual operation was obtained in the same manner as in the first embodiment.

The correlation coefficients obtained in Example 1 and Comparative Example 1 are shown graphically in FIG. In FIG. 6, the values (1, 2 and 3) on the horizontal axis represent the position vectors (x, y and z) of the three-dimensional spatial coordinates, respectively. "**" shown on the bar graph in FIG. 6 means that the significance value (p value) is smaller than 0.001.

Referring to FIG. 6, the brain-machine interface system constructed in the embodiment 1 has a statistically significant improvement in the prediction accuracy of the operation as compared with the brain-machine interface system constructed in the comparative example 1 in predicting the continuous arm motion.

Although the present invention has been described with reference to the drawings and embodiments, it is to be understood that various changes and modifications may be suggested to those skilled in the art. Accordingly, the true scope of the present invention should be determined by the technical idea of the appended claims.

100: Spatial Filter Generation System for EEG Data
101: EEG data measuring unit 102: EEG data classification unit
103: Variance calculation unit of EEG data 104: Spatial filter generation unit of EEG data
105: Spatial filter storage unit of EEG data
200: brain-machine interface system 201: brain wave data measuring unit
202: Spatial filter applying section 300: Machine
301: control unit 302:

Claims

EEG data measuring part;
EEG data classifier;
A variance calculating unit of EEG data; And
And a spatial filter generating unit of EEG data,
Wherein the variance calculating section calculates the variance between the same operation of the EEG data and the different operation of the EEG data,
Wherein the spatial filter generation unit generates a spatial filter using a ratio of variance between different operations of the brain wave data with respect to variance between the same operations of the brain wave data.

The method according to claim 1,
The EEG data measuring unit may measure the EEG data using electroencephalogram (EEG), magnetoencephalography (MEG), electrocorticogram (ECoG), functional magnetic resonance imaging (fMRI), functional near- : functional near-infrared spectroscopy) or a combination thereof.

The method according to claim 1,
Wherein the EEG data measuring unit has a form of a headset or a hairband type.

The method according to claim 1,
A noise removing unit for removing noise from brain wave data measured by the brain wave data measuring unit; And a feature extraction unit extracting EEG features from the noise-removed EEG data.

5. The method of claim 4,
The noise cancellation unit may include at least one of EEG data including a low-pass filter, a high-pass filter, a band-pass filter, a notch filter, Spatial filter creation system.

5. The method of claim 4,
The feature extraction unit may include a spatial filter generation system of EEG data using Fourier Transform, Partial Directed Coherence (PDC), Direct Transfer Function (DTF), Independent Component Analysis (ICA), Principle Component Analysis (PCA) .

5. The method of claim 4,
Wherein the feature extracting unit extracts an intensity or a pattern of a specific frequency band of an EEG that changes according to a movement direction when moving an arm or a foot or imagining a motion; The ratio of the signal from the subject's EEG measured when the eye is open and when the eye is closed (eye open / eyes closed); The response time or size of the P300 component of the EEG measured by the subject when the abnormal stimulus is applied; Brain connectivity of subjects measured at rest; The magnitude or frequency domain of the visual evoked potential (SSVEP) in the occipital lobe when visual stimuli are applied; Or a combination thereof is extracted as an EEG characteristic.

The method according to claim 1,
Wherein the brain wave data classifier extracts brain wave data from the beginning to the end of each of the plurality of operations on a time-by-time basis, and stores the extracted brain wave data for each of the same operations.

5. The method of claim 4,
The EEG data measuring unit, the noise removing unit, the feature extracting unit, the EEPR classification unit, the EEPROM dispersion calculating unit, and the EEG data spatial filter generating unit may be configured separately from each other, or two or more may be integrated Spatial Filter Generation System for EEG Data.

The method according to claim 1,
And a spatial filter storage unit of EEG data for storing a spatial filter of EEP data generated by a spatial filter generating unit of the EEP data.

New EEG data measurement part;
Applying the spatial filter of the EEG data generated in the spatial filter generation system of EEG data according to any one of claims 1 to 10 to the new EEG data measured by the new EEG data measuring unit, A spatial filter applying unit; And
And a motion prediction unit for predicting an intended motion of the subject based on the amplified new EEG data.

Acquiring brain wave data;
Classifying the obtained EEG data;
Calculating two or more variances from the classified brain wave data;
Generating a spatial filter of EEG data from the two or more variances;
Acquiring new EEG data;
Amplifying the new brain wave data by applying a spatial filter of the generated brain wave data to the acquired new brain wave data; And
And predicting an intended motion of the subject based on the amplified new EEG data,
Wherein the variance calculating step calculates the variance between the same operation of the EEG data and the different operation of the EEG data,
Wherein the spatial filter generating step generates a spatial filter using a ratio of variance between different operations of the brain wave data with respect to variance between the same operations of the brain wave data.

13. The method of claim 12,
Wherein the acquiring of the brain wave data comprises:
Acquiring brain wave data;
Removing noise from the acquired EEG data; And
And extracting EEG characteristics from the noise-canceled EEG data.

13. The method of claim 12,
Wherein the step of predicting the intended behavior of the subject based on the amplified new EEG data is performed using regression analysis.