KR20140009635A - Extracting erd/ers feature using pso - Google Patents

Abstract

An EEG analysis device and an EEG analysis method are disclosed. The disclosed EEG analysis device measures the EEG and analyzes the measured EEG using the PSO algorithm. The EEG analyzer extracts the feature points of the EEG using the PSO algorithm. The feature point may be determined as the position where the amplitude of the EEG is greatest among the EEGs shown on the time-frequency plane. The extracted feature points may be used to estimate the user's intention.

Description

Extraction of ERD / ERS characteristics using POSO {EXTRACTING ERD / ERS FEATURE USING PSO}

The present invention relates to a technique for analyzing human brain and intention by analyzing brain waves, and more specifically, to detect human psychology and intention by searching for a point (feature point) having the largest amplitude of brain waves on a time-frequency plane. It's about technology.

BCI (Brain Computer Interface) is a technology that can operate a computer only by human thoughts or psychological actions. The technique is to read what you think in your head and send it to your computer to do the work accordingly. This method is especially applicable to people with disabilities and free limbs, such as the disabled.

Univ. Of Australia. At Tech of Sydeny, we conducted a study to control the on / off of electric appliances only with the thought of people for the purpose of overcoming physical obstacles and future housing environment. According to this, when people close their eyes, alpha waves appear in a safe state, and when they open their eyes, alpha waves decrease. Therefore, the experiment was used to turn the switch on and off.

In the United States, IBVA researched interactive Biofeedback Control for application to Image / MIDI Control VR. According to this study, users can use EEG to control the speed of a car in a racing game with EEG. there was.

However, in order to use EEG signals in BCI, it is necessary to extract features of each movement or imagination from collected EEG signals and classify them in real time. Currently known classification methods are Fourier transformation or AR (frequency analysis). A method such as autoregressive is used, and a classification method such as SVM (Support Vector Machine) is used to classify it.

This method requires analysis of all areas for the user to obtain data. Therefore, it takes a long time and is not suitable for the real-time processing of EEG data.

The purpose of the following embodiments is to determine the user's intention using brain waves

Another object of the following examples is to extract the feature points of the EEG.

According to an exemplary embodiment, an EEG measuring unit for measuring the brain waves of the user, a Fourier transform unit for performing a Fourier transform for the measured brain waves, the time-frequency space with respect to the Fourier transformed brain waves, the amplitude of the largest feature point An EEG analysis apparatus including a feature point search unit for searching for is provided.

Here, the feature point search unit is a particle setting unit for setting the first position in the time-frequency space to the position of the particle, the amplitude of the brain wave at the first position and the second wave in the time-frequency space of the It may include a comparison unit for comparing the amplitude and the update unit for updating the position of the particle to the second position in accordance with the comparison result.

The feature point search unit may further include a controller configured to determine the feature point of the brain wave as the position of the particle when the amplitude of the brain wave at the position of the particle is greater than a predetermined threshold value.

In addition, the number of particles set by the particle setting unit may be plural.

Here, the brain wave may include an alpha wave and a beta wave, and the Fourier transform unit may perform a Fourier transform on the beta wave.

The filter unit may further include a filter unit for extracting the beta wave from the brain waves.

The apparatus may further include a pseudo estimator for estimating the intention of the user according to the position of the feature point.

According to another exemplary embodiment, a method of measuring brain waves of a user, performing a Fourier transform on the measured brain waves, and searching for a feature point having the largest amplitude in time-frequency space with respect to the Fourier transformed brain waves Provided are an EEG analysis method comprising the steps of:

The searching for the feature point may include setting a first position in the time-frequency space to a position of a particle, the amplitude of the brain wave at the first position, and the second position in the time-frequency space. Comparing the amplitude of the EEG and updating the position of the particle to the second position according to the comparison result.

The searching of the feature point may determine the feature point of the brain wave as the position of the particle when the amplitude of the brain wave at the position of the particle is greater than a predetermined threshold.

In the setting of the particle position, the position of the particles may be set.

Here, the brain wave includes an alpha wave and a beta wave, and the step of Fourier transform may be Fourier transform the beta wave.

The method may further include extracting the beta wave from the brain waves using a filter.

The method may further include estimating the intention of the user according to the position of the feature point.

According to the following embodiments, it is possible to determine the user's intention using the brain wave.

According to the following embodiments, it is to extract the feature points of the EEG.

1 is a view showing the concept of the present invention for analyzing the brain waves.
Fig. 2 is a block diagram showing the structure of an EEG analyzing apparatus according to an exemplary embodiment.
3 is a diagram illustrating a concept of searching for a characteristic point of the measured EEG.
4 is a block diagram illustrating in detail a structure of a feature point search unit according to an exemplary embodiment.
5 is a diagram illustrating a concept of searching for a feature point using a PSO algorithm.
Fig. 6 is a flowchart illustrating step by step an EEG analysis method according to an exemplary embodiment.
7 is a flowchart illustrating a feature point searching method according to an exemplary embodiment.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a view showing the concept of the present invention for analyzing the brain waves.

According to an exemplary embodiment, the EEG analysis device 130 may measure the EEG response of the user. EEG shows different waveforms and patterns according to the user's intention. Therefore, if the user's brain wave response is measured and analyzed, the user's intention can be estimated.

According to one side, the EEG analysis device 130 may measure the EEG response of the user using an invasive method or a non-invasive method. The invasive method is a method of directly measuring the electrical activity of the cerebral cortex by inserting the needle-shaped electrodes (121, 122) in the scalp, non-invasive method is a method of measuring the EEG response by attaching the electrode to the scalp.

When the EEG response is measured using a non-invasive method, when the user surrounds the head with a plurality of electrodes attached to the head, the EEG may be measured at each part of the head.

The EEG analyzing apparatus 130 may analyze the measured EEG of the user. According to one side, the EEG analysis unit 130 may analyze the EEG using a Fourier transform and Particle Swarm Optimization (PSO) algorithm. The EEG analyzing apparatus 130 may extract the feature points of the EEG using the PSO algorithm, and estimate the user's intention according to the extracted feature points. Here, the feature point of the EEG may mean a coordinate of time and frequency in which the amplitude of the EEG is greatest on the time-frequency plane.

According to one side, the EEG analysis device 130 may be provided with a database. The database may store information about the user's brain waves or characteristic points of each brain wave measured in various situations. The EEG analyzing apparatus 130 may estimate the user's intention by comparing the information on the feature points of each EEG stored in the database with the feature points extracted from the measured EEG.

According to one side, the EEG analysis device 130 may control an object, such as a car 140, according to the estimated user's intention. In FIG. 1, only the vehicle 140 is illustrated as an example of an object that the EEG analysis device 130 can control. According to another embodiment, the EEG analysis device 130 uses a control signal such as a wheelchair for a disabled person or an electronic product. You can control all controllable things.

Fig. 2 is a block diagram showing the structure of an EEG analyzing apparatus according to an exemplary embodiment.

The EEG analyzing apparatus 200 according to an exemplary embodiment may include an EEG measuring unit 210, a filter unit 220, a Fourier transform unit 230, a feature point searching unit 240, and a pseudo estimating unit 250. have.

The EEG measuring unit 210 measures the EEG of the user. According to one side, the EEG measurement unit 210 may measure the EEG response of the user using an invasive, non-invasive method. According to one side, the EEG measurement unit 210 may measure the EEG response of the user at a plurality of points of the head of the user. According to one side, the EEG measured by the EEG measurement unit 210 is between 0 ~ 40Hz It can be a range. The measured brain waves may include alpha waves and beta waves. Here, alpha waves are brain waves between 8 and 13 Hz, and beta waves are brain waves between 13 and 30 Hz.

The filter unit 220 extracts beta waves included in the measured EEG by performing filtering on the measured EEG.

The Fourier transform unit 230 performs Fourier transform on the filtered EEG. According to one side, the Fourier transform unit 230 may perform a short time Fourier transform (STFT) on the filtered brain waves. The short time Fourier transform divides the measured EEG into time intervals having a predetermined length of time, and performs Fourier transform for each divided time interval. That is, when the short-time Fourier transform is performed on the EEG, Fourier coefficients are calculated for each time interval and each frequency band, and the Fourier coefficient is the strength of the EEG in the corresponding time interval and the corresponding frequency band. It is proportional to the amplitude or power of.

The feature point search unit 240 searches for the feature point having the largest amplitude in the time-frequency space with respect to the Fourier transformed brain wave. Hereinafter, a configuration of searching for a feature point in time-frequency space will be described in detail with reference to FIG. 3.

3 is a diagram illustrating a concept of searching for a characteristic point of the measured EEG.

In FIG. 3A, the horizontal axis represents time, and the vertical axis represents frequency. In FIG. 3, the EEG intensity of the portion represented by a dot is a low value, and the EEG intensity of the portion represented by a dark color is a high value.

The feature point search unit 240 searches for the feature point 310 having the largest amplitude of the EEG on the time-frequency plane shown in FIG. According to one side, the searched feature points may be used to estimate the user's intention.

In FIG. 3B, the horizontal axis represents time, and the vertical axis represents power of brain waves. In general, since the user has a specific intention, after a predetermined time 320, the power of the EEG becomes maximum, and the timing at which the power of the EEG becomes maximum according to the user's intention may vary. In addition, according to the user's intention, the frequency of the EEG at which the EEG amplitude is maximum may also vary. Therefore, if the feature point corresponding to the time-frequency at which the amplitude of the EEG becomes maximum can be searched, the user's intention can be estimated.

If the user has a specific intention, if the EEG decreases, it is called ERD (Event Related Desynchronization). If the EEG increases, this is called ERS (Event Related Synchronization). The feature point searcher may calculate an ERD / ERS for each time-frequency plane and search for a feature point with respect to the calculated ERD / ERS.

[Equation 1]

는

시간,

주파수 에서 뇌파의 전력을 나타내고,

는

주파수에서,

시간에서

시간 동안의 평균 파워를 나타내며 수학식 2와 같이 결정된다.
here,

The

time,

Represents the power of the EEG at the frequency,

The

At the frequency,

In time

It represents the average power over time and is determined as in Equation 2.

&Quot; (2) "

The pseudo estimator 250 estimates the user's intention according to the position of the feature point. According to one side, the doctor estimator 250 may estimate a user's intention using a database. The database may store information about the user's brain waves or characteristic points of each brain wave measured in various situations. The pseudo estimator 250 may estimate the user's intention by comparing information on the feature points of each brain wave stored in the database with the feature points extracted from the measured brain waves.

4 is a block diagram illustrating in detail a structure of a feature point search unit according to an exemplary embodiment. The feature point searcher 240 according to the exemplary embodiment may include a particle setter 410, a comparator 420, an updater 430, and a controller 440.

Hereinafter, a specific example in which the feature point search unit 240 searches for a feature point will be described with reference to the EEG in time-frequency equality illustrated in FIG. 5.

5 is a diagram illustrating a concept of searching for a feature point using a PSO algorithm.

In FIG. 5A, the horizontal axis is a time axis, the vertical axis is a frequency axis, and the coordinates of time and frequency are values corrected around a feature point.

The particle setting unit 410 sets the first position to the positions of the particles 511, 512, 513, 514, and 515 in the time frequency space. According to one side, the particle setting unit 410 may set a plurality of particles (511, 512, 513, 514, 515).

According to one side, the particle setting unit 410 may arbitrarily set the first position, in particular, the first position of each particle may be set according to the uniform distribution (uniform distribution).

The comparator 420 may compare the amplitude of the EEG at the first position of each particle with the amplitude of the EEG at the second position in the time-frequency space.

The updater 430 may update a position where the amplitude of the EEG is greater among the first position and the second position to the position of each particle. According to one side, the second position may be determined based on the first position.

Among the particles 511, 512, 513, 514, and 515, an object having optimal location information is called pbest, and an object representing most optimal location information among all objects of pbest is called gbest. The second position may be determined using Equation 3 below.

&Quot; (3) "

는 각 파티클의 제1 위치,

는 각 파티클의 제2 위치이며,

는 계산에 의해 업데이트될 속도를 의미한다.

는 하기 수학식 4와 같이 결정될 수 있다.
here,

Is the first position of each particle,

Is the second position of each particle,

Means the speed to be updated by the calculation.

May be determined as in Equation 4 below.

&Quot; (4) "

는 가중치,

는 상수이며,

는 임의의 실수이다.
here,

Is the weight,

Is a constant,

Is any real number.

That is, the updater 430 determines the positions of the particles 511, 512, 513, 514, and 515 with respect to each particle as a position where the amplitude of the EEG is larger among the first position and the second position. If the comparator 420 and the updater 430 repeatedly update the position of each particle 511, 512, 513, 514, 515, each particle 511, 512, 513, 514, 515 is time- In the frequency space, the amplitude of the EEG converges to a larger position.

5B is a diagram illustrating that each particle converges to a location where the amplitude of the EEG is large. In FIG. 5A, many particles 511, 512, 513, 514, and 515 are set. However, as the update is repeated, the particles converge to one point 521 having the largest amplitude of the EEG.

The controller 440 may compare the amplitude of the EEG at a location of each particle with a predetermined threshold. If the amplitude of the EEG at the location of each particle is greater than a predetermined threshold, the controller 440 may determine that each particle has converged, and determine the feature point of the EEG as the location of the particle.

That is, by using the PSO algorithms described in Equations 3 to 4 and FIG. 5, it is possible to easily search for the feature point having the largest amplitude of the EEG. Since the calculation amount of the PSO algorithm is very small compared to other algorithms, the EEG analyzer can quickly analyze the user's EEG and search for a feature point, and can quickly control the object according to the user's intention.

Fig. 6 is a flowchart illustrating step by step an EEG analysis method according to an exemplary embodiment.

In operation 610, the EEG analyzer measures the EEG of the user. According to one side, the EEG analysis device can measure the EEG of the user using an invasive, non-invasive method. The measured brain waves may include alpha waves between 8 and 13 Hz and beta waves between 13 and 30 Hz.

In operation 620, the EEG analyzer may extract beta waves from the EEG by performing filtering on the measured EEG.

In operation 630, the EEG analyzer may perform a Fourier transform on the filtered EEG. According to one side, the EEG analyzing apparatus may classify the EEG by a predetermined time unit by using a window (window). The EEG analyzer may convert an EEG into a time-frequency space by performing a Short Time Fourier Transform (STFT) on the EEGs divided into predetermined time units.

In operation 640, the EEG analyzer searches for feature points of the EEG using the PSO algorithm. The characteristic point of an EEG means the point where amplitude, power, and energy of an EEG are largest in time-frequency space.

Hereinafter, a method of searching for a feature point of an EEG using the PSO algorithm will be described with reference to FIG. 7.

7 is a flowchart illustrating a feature point searching method according to an exemplary embodiment.

In step 710, the EEG apparatus sets the first position to the position of the particle. In this case, the EEG analyzer may individually set the first position with respect to the plurality of particles. According to one side, the EEG apparatus may arbitrarily set a first position for each particle, and in particular, may set the first position according to a uniform distribution.

In operation 720, the EEG analyzer may compare, for each particle, the amplitude of the EEG at the first position and the amplitude of the EEG at the second position.

In operation 730, the EEG analyzing apparatus may update a position where the amplitude of the EEG is greater among the first position and the second position to the position of each particle.

In operation 740, the EEG analyzing apparatus determines whether the feature points have converged. According to one side, the EEG analyzer may determine that the particles have converged if the amplitude of the EEG at the location of each particle is greater than a predetermined threshold. In this case, the EEG analyzing apparatus may determine the feature point of the EEG as the position of the particle in operation 750.

In operation 650, the EEG apparatus may estimate a user's intention according to the position of the feature point. According to one side, the EEG analysis device may have a database. The database may store information about the user's brain waves or characteristic points of each brain wave measured in various situations. The EEG analyzer may estimate the user's intention by comparing information on the characteristic points of each EEG stored in the database with the characteristic points extracted from the measured EEG.

According to one side, the EEG analysis device may generate a control signal for controlling the object according to the user's intention.

The methods according to embodiments of the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined by the equivalents of the claims, as well as the claims.

120: EEG analysis device
121, 122: electrode
130: car

Claims (15)

EEG measuring unit for measuring the brain wave of the user;
A Fourier transform unit for performing a Fourier transform on the measured brain waves;
A feature point search unit for searching for a feature point having the largest amplitude in the time-frequency space with respect to the Fourier transformed EEG.
EEG analysis device comprising a. The method of claim 1,
The feature point search unit
A particle setting unit configured to set a first position as a position of a particle in the time-frequency space;
A comparison unit comparing the amplitude of the brain waves at the first position with the amplitude of the brain waves at a second position in the time-frequency space; And
An update unit to update the position of the particle to the second position according to the comparison result;
EEG analysis device comprising a. 3. The method of claim 2,
The feature point search unit
A control unit for determining a feature point of the brain wave as the position of the particle when the amplitude of the brain wave at the position of the particle is larger than a predetermined threshold value
EEG analysis device further comprising. 3. The method of claim 2,
EEG analysis apparatus having a plurality of particles set by the particle setter. The method of claim 1,
The brain waves include alpha waves and beta waves,
The Fourier transform unit is a brain wave analysis device for Fourier transform the beta wave. The method of claim 5,
Filter unit for extracting the beta wave from the brain waves
EEG analysis device further comprising. The method of claim 1,
Pseudo estimator for estimating the intention of the user according to the position of the feature point
EEG analysis device further comprising. Measuring brain waves of a user;
Performing a Fourier transform on the measured brain waves;
Searching for a feature point having the largest amplitude in time-frequency space with respect to the Fourier transformed EEG;
EEG analysis method comprising a. 9. The method of claim 8,
Searching for the feature point,
Setting a first position in the time-frequency space to the position of a particle;
Comparing the amplitude of the brain waves at the first position with the amplitude of the brain waves at a second position in the time-frequency space; And
Updating the position of the particle to the second position according to the comparison result;
EEG analysis method comprising a. 10. The method of claim 9,
Searching for the feature point,
And when the amplitude of the EEG at the position of the particle is greater than a predetermined threshold, determining a feature point of the EEG as the position of the particle. 10. The method of claim 9,
The setting of the particle position may include setting a position with respect to a plurality of particles. 9. The method of claim 8,
The brain waves include alpha waves and beta waves,
The Fourier transform step is Fourier transform the beta wave method. The method of claim 12,
Extracting the beta wave from the brain waves using a filter
EEG analysis method further comprising a. 9. The method of claim 8,
Estimating the intention of the user according to the position of the feature point
EEG analysis method further comprising a. A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 8-14.

Images