KR101842750B1

KR101842750B1 - Realtime simulator for brainwaves training and interface device using realtime simulator

Info

Publication number: KR101842750B1
Application number: KR1020150166175A
Authority: KR
Inventors: 권장우; 정해성; 이상민
Original assignee: 인하대학교 산학협력단
Priority date: 2015-11-26
Filing date: 2015-11-26
Publication date: 2018-03-27
Also published as: KR20170061317A

Abstract

A real-time simulator for brain wave training and an interface device using the same are disclosed. At least one program loaded memory; And at least one processor, wherein the at least one processor acquires brain wave data measured by a user through an EEG measuring device after presenting an imagining operation for EEG training according to the control of the program; And extracting feature information corresponding to the imaging operation from the brain wave data to learn the characteristics of the user's brain wave with respect to the imaging operation.

Description

TECHNICAL FIELD [0001] The present invention relates to a real-time simulator for training an EEG, and an interface device using the simulator.

The following description relates to a simulator for EEG training and an interface device using the same.

This study was conducted as a result of the study of the following projects.

- IITP-2015-H8501-15-1019 (IITP-2015-H8501-15-1019) The University ICT Research Center of Future Creation Science and Information and Communication Technology Promotion Center

- The basic research project (2010-0020163), funded by the government (Ministry of Education) and funded by the Korea Research Foundation,

In the era of information explosion, "how to deliver" information rather than "how to generate" is important, which results in interface problems. As interfaces and system paradigm shifts to human-centricity become more and more necessary to develop underlying technologies, interfaces will be an effective means of reducing gaps between products and users.

Recently, research on interfaces has been studied with various approaches other than mouse and keyboard in the past. In the course of various studies, the bio-signal-based interface technology is a human-friendly interface technology that utilizes artificial bio-signals such as electrocardiograms and brain waves, and is being studied as a next generation user interface after text / voice / gestures. Although bio-signal-based interface technology has been studied extensively in Korea and abroad, it has yet to show practical technology that is commercially available.

Recently, BCI (Brain Computer Interface), which is actively studied, is an interface technology using brain waves. It is introduced interface technology to control cursor through brain wave and BCI model to control automobile operation in 3D virtual environment.

Korean Patent Application No. 1993-0021335 (entitled " Brake automatic control device using brain wave and method thereof "), Korean Patent Application No. 1996-057464 (invented) as an example of a technique for controlling an automobile brake using brain waves Name of the brake control device by detecting the brain waves of the driver).

However, the biggest problem of interface technology using EEG is that individual brain waves have different characteristics and the ability to generate brain waves also has individual differences. Therefore, in order to apply the human brain signal to the BCI, it is necessary to study the EEG signal training that the user can smoothly generate the EEG signal.

The present invention provides a rehabilitation system and method for training a user to generate an EEG signal smoothly through an EEG signal training.

The present invention provides a rehabilitation system and method for training a user's ability to generate brain waves for efficient use of an EEG-based interface device.

At least one program loaded memory; And at least one processor, wherein the at least one processor acquires brain wave data measured by a user through an EEG measuring device after presenting an imagining operation for EEG training according to the control of the program; And extracting feature information corresponding to the imaging operation from the brain wave data to learn the characteristics of the user's brain wave with respect to the imaging operation.

According to an aspect of the present invention, the at least one processor is configured to process a process of providing a guide or a comment related to the imaginary operation when the feature information is out of a critical range, under the control of the program, The acquisition process and the learning process may be repeatedly performed.

According to another aspect of the present invention, the learning process includes separating frequency bands necessary for EEG analysis from the EEG data using Fast Fourier Transform (FFT), and then extracting EEG data from the EEG data of the separated frequency band It is possible to extract the corresponding feature information and store the information according to the imaging operation through the machine learning.

According to another aspect of the present invention, the learning process includes analyzing a coefficient of regression analysis for the EEG data using a k-nearest neighbors algorithm, The learning model of the k-nearest neighbor algorithm can be generated by extracting feature information corresponding to the imaginary operation in the frequency band of the channel.

According to another aspect of the present invention, the learning process includes: a first step of analyzing a regression analysis decision coefficient for the acquired EEG data; A second step of selecting a channel and a frequency band for EEG learning according to the regression analysis decision coefficient analyzed in the first step; A third step of extracting feature information corresponding to the imaging operation in a frequency band of the selected channel in the second step; A fourth step of generating a learning model in which the feature information extracted in the third step is a training set; A fifth step of analyzing a regression analysis decision coefficient for EEG data acquired after the learning model is generated; A sixth step of selecting a channel and a frequency band for EEG learning according to the regression analysis decision coefficient analyzed in the fifth step; A seventh step of extracting feature information corresponding to the imaging operation in a frequency band of the channel selected in the sixth step; And an eighth step of comparing the characteristic information extracted in the seventh step with the training set and outputting a comparison result.

There is provided an interface device that uses the system of the above-described contents to set the user's brain wave characteristic as an interface variable for inputting a computer.

According to the embodiment of the present invention, the user's brain wave power can be improved by providing a training simulation for user's brain wave generation.

According to the embodiment of the present invention, BCI technology based on EEG can be realized by applying trained EEG as an interface variable for computer input.

1 is a view for explaining an example of the configuration of a rehabilitation system according to an embodiment of the present invention.
FIG. 2 is a diagram for explaining an example of an EEG training process of an EEG training apparatus according to an embodiment of the present invention.
3 is a diagram showing an example of the internal configuration of an EEG training apparatus according to an embodiment of the present invention.
4 and 5 are diagrams illustrating an example of an interface screen in which an EEG signal can be confirmed in an embodiment of the present invention.
6 is a diagram illustrating an example of an EEG signal learning process of the EEG training apparatus according to an embodiment of the present invention
7 is a block diagram for explaining an example of the internal configuration of a computer system according to an embodiment of the present invention.

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

The present invention relates to a technique for providing a training simulation of a user's brain wave, and in order to use software based on an EEG signal, a simulation service including generation practice, measurement, judgment, etc. for a user's EEG signal .

The present invention aims to develop a simulator for training the user's ability to generate brain waves for efficient use of an EEG signal based interface device. In order to apply the human brain signal to the BCI, it is necessary to develop an EEG signal trainer to train the user to generate the EEG signal smoothly.

First, a background study related to the present invention will be described.

BCI (Brain Computer Interface)

BCI is defined as an interface technology that connects a human brain with a computer and controls the computer through brain waves, and widely belongs to HCI (Human Computer Interface) technology. In addition, BCI technology is sometimes referred to as BMI (Brain Machine Interface) because brain waves can manipulate machines such as wheelchairs and robots. The implementation of BCI technology involves measuring brain waves through a device that recognizes EEG stimuli, analyzing EEG through signal processing, and then issuing commands to input / output devices.

EEG brainwave )

Electroencephalogram (EEG) refers to a minute electrical activity of a brain cell population induced by attaching an electrode to the scalp, amplified by an electroencephalogram system, and recorded with the potential as the vertical axis and the time as the horizontal axis. In other words, EEG is a measure of electrical activity that occurs in the cerebral cortex. It changes in time and space according to brain activity, state at the time of measurement, and brain function. It has a frequency of 0 ~ 50Hz and an amplitude of 10 ~ 200uV. The EEG is classified into a delta wave, a delta wave, a theta wave, an alpha wave, an alpha wave, a beta wave, and a beta wave depending on the frequency range . Table 1 shows the characteristics of the EEG according to frequency.

designation frequency Primary site amplitude characteristic Delta wave
(delta wave) 0.1 to 3 Hz Variable 100 ~ 20㎶ Appears mainly in deep sleep Setta
(? wave) 4 ~ 7㎐ occipital,
Temporal 10 to 50 pounds Attention to sleep or emotional state, attention to concentration Al Papa
(alpha wave) 8-12Hz However,
Occipital lobe 10 to 150 pounds Tension appears mainly in a relaxed state and is closely related to brain development Beta wave
(beta wave) 13 ~ 30㎐ Frontal 5 to 10 pounds In all wakeful conscious activity, especially in anxiety / tension state, it occurs with visual, auditory, tactile, taste, olfactory Gamma wave
(? Wave) 30-50 Hz Frontal lobe,
Occipital lobe 2 ~ 120 It is mainly caused by arousal and excitement.

Measurement position of EEG

Electroencephalograms measure the electrical activity of the brain by attaching electrodes to the cerebral cortex. At this time, attach the electrode according to International10-20System which specifies the attachment position of the electrode on the scalp. In International10-20 System, C means Central, F means Frontallobe, P means Parietallobe, O means Occipitallobe, and T means Temporallobe. The numbers after the alphabet mean the right hemisphere, which is the odd number, and the right hemisphere, which is the even number.

The position of the electrode is indicated by the center line connecting the division point (Naison) between the eye and the forehead and the division point (Inion) between the neck and the back of the head, and the middle part is called Cz. Based on the Cz point, the location of the EEG was determined by dividing 20% of the length in the direction of Naison and Inion, 20% in front of it, and 10% in front of it again. The name attached to it is International10-20 System.

EEG analysis and learning

In the present invention, it is possible to extract and train frequencies necessary for EEG analysis through Fourier transform, fast Fourier transform, and k-nearest neighbor algorithm of the above-described description for brain waves acquired from a user.

(1) Fourier Transform

Fourier transform refers to the decomposition of a waveform such as a voice into a fundamental frequency (base tone) and an angular frequency of each of the fundamental frequencies (each harmonic). In short, It is a method to calculate. All the waveforms can be represented by the sum of the fundamental wave and the high frequency wave through this Fourier transform. Therefore, if one waveform is continued, it is possible to know through which Fourier transform the waveform is composed.

(2) Fast Fourier Transform (Fast Fourier Transform)

Fast Fourier transform is abbreviated as FFT as one of algorithms for performing discrete Fourier transform (DFT).

When a computer performs discrete Fourier transform (DFT) of N data strings, it is necessary to multiply N2 times by calculating the Fourier coefficients of each frequency independently. However, if N is decomposed into a prime factor, the data string is divided and subjected to discrete Fourier transform (DFT) on a prime number group corresponding to each prime number, and finally N discrete Fourier transforms (DFT ), The number of operations is reduced. This effect is most significant when N is a power of 2, which requires only Nlog2N [times] multiplication. Therefore, the calculation time is significantly shortened when two-dimensional discrete Fourier transform (DFT) of an image, which is a two-dimensional data stream, is performed.

(3) k-nearest neighbors algorithm

The training data is a vector in a multidimensional feature space each having an item category name. The training phase of the algorithm is only to store the feature vector and item category name of the training specimen.

In the classification step, k is a user-defined constant and the vector without the classification (query or verification point) is classified by assigning the most frequent category names among the k training samples.

The most commonly used distance measure in continuous variables is the Euclidean distance. For discrete variables such as character classification, other measures such as the overlap distance (or hamming distance) may be used. For example, in the case of gene expression microarray data, k-NN uses correlation coefficients such as Pearson and Spearman only. Often, a distance measure with a large margin can improve the accuracy of k-NN classification significantly if it is learned by a special algorithm such as nearest neighbors or neighborhood component analysis.

The disadvantage of the classification according to the "majority vote" appears when the item distribution is biased. In other words, data of more frequent items tend to dominate the prediction of new data. This is because the data of the more frequent items tend to be the majority and become the majority of the k closest neighbors. One way to solve this problem is to weight the classification by considering the distance between each verification point and k nearest neighbors. Multiply the k items of the nearest neighbors (in the case of regression problems) by a weight inversely proportional to the reciprocal of the distance from that point to the verification point. Another way to solve the bias is to abstract the data representation. For example, in a self-organizing map (SOM), each node is a cluster representative (center) of similarities regardless of the original training data density. k-NN can also be applied to SOM.

In addition, in the present invention, an independent component analysis (ICA) may be used to remove unnecessary noise of the measured EEG, and a power spectrum analysis may be applied to acquire frequency information of brain waves And a band pass filter can be used to extract a frequency band necessary for EEG analysis.

(6) Independent Component Analysis

Independent component analysis is a method of extracting original signals by separating signals from the result of mixing multiple signals and minimizing interference. Independent component analysis is one of the methods of Blind Source Separation (BSS) that finds the source of the signal through analysis of the signal in the absence of information about the generation position or generation path of linear signals.

Independent component analysis minimizes correlations and dependencies in signals of various dimensions, and can be classified into the most independent signals probabilistically by maximizing entropy. Electroencephalogram (EEG) is also a multidimensional signal that is a combination of linear signals measured from several electrodes. When the independent component analysis is applied to this signal, the brain waves obtained from the electrodes in the EEG signal can be separated into independent signals of the electrodes, so that data closest to the original EEG signals with minimal interference can be extracted. Thus, independent component analysis can be applied to isolate and remove unwanted noise measured with EEG such as eye flicker, eye movement, and muscle movement.

(7) Band Pass Filter

The EEG signal acquired from the user is composed of various frequency components even if the unnecessary noise is removed. In this integrated signal, only the frequency domain necessary for analysis should be extracted. For this purpose, frequency filtering is used. The types of the filters include a low pass filter which passes only a frequency signal lower than a specific frequency band according to the frequency domain, a high pass filter which passes only a frequency signal higher than a specific frequency band And a band pass filter for filtering a specific frequency band by passing a specific frequency band. Therefore, a frequency corresponding to brain waves can be extracted using a band-pass filter of a specific band (for example, 0 to 50 Hz).

(6) Power Spectrum Analysis

Although it is convenient to observe the EEG pattern as an analog time series data that changes with time, it is not enough to analyze by analyzing the specific information.

Power spectral analysis converts the time series analogue signal, EEG, into frequency domain through FFT. The signal transformed into the frequency domain can confirm the distribution and the density of the frequency components. In this spectrum analysis, the power value is mainly used, which is expressed as the square of the amplitude (㎶ ² ). It is possible to calculate the power value by spectral analysis with very short time EEG measurement. However, in order to better represent the electrical activity of the brain, the power value is calculated for the refined EEG, which is removed from the noise of about 20-30 seconds to 3 minutes per frequency band.

1 is a view for explaining an example of the configuration of a rehabilitation system according to an embodiment of the present invention. Referring to FIG. 1, the rehabilitation system according to the present invention includes an EEG device 101 and an EEG device 100.

The EEG device 101 may include an electrode for attaching to a subject's scalp and measuring brain waves. At this time, the electrode may be a wet electrode or a dry electrode which does not need to use a separate conductive gel. The position of the electrodes to measure the user's brain waves can be determined according to the International Standard (International10-20System).

The EEG device 101 may be configured as one system with the EEG training device 100 or may be configured as a separate system and interlocked with the EEG training device 100. The EEG device 101 may be connected to the EEG training device 100 through wired communication or wireless communication. For example, the EEG device 101 may be connected to the EEG training device 100 including a communication module for short-range wireless communication (e.g., WiFi, Bluetooth, ZigBee, etc.). In other words, the EEG device 101 acquires real-time EEG data generated in the cerebral cortex of the user and transmits EEG data to the EEG training device 100, which is an apparatus for EEG training through short-range wireless communication .

The EEG training apparatus 100 plays a role of providing a training simulation of the user's brain wave generation using the EEG signal acquired in real time through the EEG 101. The EEG training apparatus 100 may be implemented as an application that can be used in a PC environment as well as a mobile environment, and may be implemented as an independently operating program or an in-app form of a specific application And can be implemented to be able to operate on the specific application.

As can be seen from the characteristics of the EEG signal according to frequency division, the EEG signal has unique characteristic information for each imagined motion. As a simple example, in the case of logical actions such as mathematical calculations, the left brain wave is activated compared to the right brain wave, and the right brain wave can be activated compared to the left brain wave in the case of emotional actions such as reminiscent of good memory or sick memory. When the user imagines a specific operation, an EEG signal including feature information corresponding to the operation is generated. Since the characteristics of each user's brain waves are different, it is very difficult to distinguish the imaginary motion from the brain waves, Therefore, it is difficult to acquire the EEG signal containing the accurate feature information when there is a concern or difficulty concentrating in the process of operation imagination. However, repeating the regular practice of the operation imagination can improve the user's ability to generate EEG by learning the user's EEG characteristics and increasing the concentration for EEG signal generation. To this end, the EEG training apparatus 100 according to the present invention can provide an exercise environment for generating the EEG signal of the user.

FIG. 2 is a diagram for explaining an example of an EEG training process of an EEG training apparatus according to an embodiment of the present invention.

1. An EEG training apparatus 100 provides an imaginary operation for brainwave generation training. For example, the brain-wave training apparatus 100 can present information about the operation so that the user can imagine a 'forward movement'. The brain-wave training apparatus 100 can present an imaginary motion by explaining an operation with text or voice, or displaying an image or a moving image representing the corresponding operation on the screen.

2. The EEG training apparatus 100 can acquire measured EEG data from the user through the EEG apparatus 101. [ The EEG training apparatus 100 can acquire the EEG measured in real time while the user imagines the 'forward movement' after presenting the 'forward movement' as the imaginary motion to the user.

3. The EEG training apparatus 100 can extract the EEG characteristics corresponding to the imaginary motion presented to the user by analyzing the EEG acquired from the EEG 101, Feature can be learned. For each imaginary motion, channels and frequency bands for EEG analysis can be defined in advance. For example, the EEG apparatus 100 can extract the activation level of the left EEG for the 'forward movement' and learn the EEG characteristics of the user regarding the 'forward movement' have.

4. The EEG training apparatus 100 can provide feedback information according to the results of the EEG analysis to improve the user's ability to generate EEG. For example, when the user's brain wave characteristic extracted from the imaginary motion presented to the user is out of a predetermined threshold range for the motion, the EEG training apparatus 100 may focus on the motion or help the EEG signal generation You can provide a guide or comment that can be made. Guides and comments that can aid in the generation of EEG signals can be provided in a variety of forms, such as text and images, and can be defined in advance for each imagination action presented to the user. For example, if the activation level of the left EEG is lower than the threshold value in the measured EEG while the user imagines the 'forward movement', the motion or image that can be activated by the left EEG in relation to the imagination motion presented to the user Can be presented.

5. The EEG training apparatus 100 may provide a training simulation for the user's brain wave generation with respect to the imaginary motion presented to the user while repeating the above-mentioned 2 to 4 steps. At this time, the EEG training apparatus 100 goes through the above-described steps 1 to 5 for each imaginary motion, and thus can learn the characteristics of the user's EEG through the EEG signal generation training according to the imaginary motion.

6. The EEG training device 100 can transmit the trained EEP analysis result to the interface device 102. [ In other words, the EEG training device 100 may transmit the analysis result of the EEG signal acquired through the EEG signal generation training to the input of the interface device 102 for controlling other devices. At this time, the interface device 102 can be utilized as a game input / output device based on brain waves, an automobile control system, an input / output device that can use a disabled person or a patient in a coma state, or the like.

Accordingly, in the present invention, the user can imagine the action suggested by the brain training apparatus 100 and directly watch the simulation results and the feedback information about the generated brain waves, thereby repeating the training for the EEG signal generation more effectively and conveniently have. Accordingly, the EEG training apparatus 100 can learn the characteristics of the user's EEG signal through the EEG signal generation training for the user, and can improve the user's ability to generate EEG by providing feedback. Further, by applying the EEG trained in the EEG 100 as an interface variable for computer input, it is possible to implement a more accurate EEG-based BCI technology.

3 is a diagram showing an example of the internal configuration of an EEG training apparatus according to an embodiment of the present invention.

3, the EEG apparatus 100 includes an EEG receiving unit 310, a signal separating unit 320, a learning unit 330, a guessing unit 340, a user defining unit 350, (360).

The EEG signal generated by the user's imagination may be measured by the EEG 101 and transmitted to the EEG training apparatus 100. At this time, the EEG receiving unit 310 may measure the EEG signal measured by the EEG 101, And performs a preprocessing process on the received EEG data after receiving it in real time. For example, the EEG receiving unit 310 includes a pre-processing filter unit for removing low-frequency components of the EEG signal received from the EEG apparatus 101, an instrumentation amplifier for first amplifying the EEG signal filtered by the low- A post-processing filter unit for removing an excess frequency band of the EEG signal amplified by the measurement amplifier, and an output amplifying unit for performing a secondary amplification of the EEG signal filtered by the post-processing filter unit. At this time, the post-processing filter unit may include a low-pass filter for removing a frequency band exceeding 30 Hz, and a band elimination filter for further filtering a band of 30 Hz or more that can not be removed from the EEG signal filtered by the low-pass filter.

The signal separator 320 separates the EEG signal according to the frequency range and separates the EEG signal of the band required for EEG analysis from the EEG signal received through the EEG receiver 310. [ For example, the signal separator 320 can convert a frequency band into a frequency band through a Fast Fourier Transform, wherein a theta pass filter for extracting an EEG signal in the 4-7 Hz band, an EEG signal in the 7-13 Hz band An alpha-pass filter for extracting the signal; And a beta pass filter for extracting EEG signals in the 13-30 Hz band.

The learning unit 330 extracts the minutiae points of the EEG signal separated by the signal separator 320, applies the machine learning, and then stores information according to the motion imagery presented to the user. At this time, the learning unit 330 can learn the EEG signal through power analysis using the frequency spectrum of the EEG signal. For example, the learning unit 330 may include a feature extraction unit 131 for extracting feature points from a waveform of the EEG signal through a feature point extraction algorithm stored therein, and a feature extraction unit And a storage unit 133 for storing the machine learning data input through the machine learning processing unit 132. The machine learning processing unit 132 may include a machine learning processing unit 132, At this time, the feature extraction unit 131 may calculate the feature set of the vector form by sequentially applying Equations (1) to (4) to the EEG signal.

In this case, DAMV means the absolute difference average value of the signal, which is a value obtained by integrating the absolute value of the signal for a predetermined time. Where x is the measured signal, i is the order of the samples, and N is the number of samples. This condition applies equally to the calculation of other feature points. DAMV is also referred to as average amplitude change (AAC) in other words.

Here, MAV denotes an absolute value of the absolute value of each signal amplitude as an absolute integral value.

Here, VAR is used to estimate the power density of the signal. In general, VAR is called signal variance and is defined as the mean square value of the signal deviation.

Here, RMS means the root-mean-square (RMS) of the signal, which is a variable with high magnitude and relevance of the power over time. It has an advantage that it can be easily calculated by a mathematical method without modifying the original signal of the signal.

The estimator 340 extracts feature points of the EEG signal measured after the execution of the learning unit 330, applies the machine learning, and stores information according to the motion imagery presented to the user. At this time, the estimator 340 can learn the EEG signal through power analysis using the frequency spectrum of the EEG signal. For example, the estimating unit 340 may include a feature extracting unit 341 for extracting feature points from a waveform of the EEG signal through a feature point extraction algorithm stored therein, a feature set including the extracted feature points, And a storage unit 343 for storing the machine learning data inputted through the machine learning processing unit 342. [ At this time, the feature extracting unit 341 may calculate the feature set of the vector form for the EEG signal using Equations (1) to (4).

The specific operation of the learning unit 330 and the speculative unit 340 are similar, but the result of learning based on the feature vector of the initially measured EEG signal is stored in the learning unit 330, and for the EEG signal measured thereafter The guessing unit 340 learns and stores the same in the same manner as the learning unit 330. [

The user-defined unit 350 can receive a channel and a frequency band with high accuracy in the analysis result of the EEG signal by the user through the decision coefficient analysis through the machine learning algorithm in the learning unit 330. For example, if the similarity between the learning information of the learning unit 330 and the learning information of the estimating unit 340 is lower than the threshold value, the channel and frequency for the EEG signal may be selected by the user. The learning model may be regenerated and stored in the learning unit 330. Therefore, the similarity can be expected to increase as the learning is repeated based on the model generated again for the user's EEG signal.

The determining unit 360 compares the learning information stored in the learning unit 330 with the learning information stored in the guessing unit 340 with respect to the EEG signal of the user. (K-NN model) is generated based on the generated feature vector after extracting the feature points from the initially measured signal. The generated model is stored in the learning unit 330, and the signal measured thereafter is estimated (340). &Lt; / RTI > Accordingly, the determination unit 360 can calculate the degree of similarity by testing the feature vector generated by the estimating unit 340 on the model stored in the learning unit 330. FIG.

At this time, the determination unit 360 may provide a result of the comparison between the learning information of the learning unit 330 and the learning information of the inferring unit 340 as feedback information on the result of the brain wave analysis to the user who is the EEG training object. For example, the determination unit 360 can provide the user with the similarity between the learning information of the learning unit 330 and the learning information of the inferring unit 340. At this time, the user can confirm the similarity between the learning information, As well as information on a predefined threshold value.

The EEG training apparatus 100 can display the measured EEG waveform and EEG analysis results to the user so that the user can confirm the results. For example, as shown in FIGS. 4 and 5, raw data (Raw EEG) of the EEG signal received from the EEG 101 and channel and frequency band (Channel & Frequency) of the EEG signal are displayed on the screen in real time can do. The user can practice realization of the EEG signal with the EEG training simulation and check the EEG generation level in real time through the feedback provided by the EEG training device 100. [

The EEG training method according to the present invention may be composed of at least two operations based on the details described with reference to FIG. For example, the EEG training method includes an EEG receiving step of receiving an EEG signal generated by a user's motion imagery; A signal separation step of separating the received EEG signal according to a frequency range; A learning step of extracting minutiae points of the separated EEG signal and applying machine learning and storing information according to the motion imagery; Extracting characteristic points of the EEG signal measured after the learning step and applying machine learning to store information according to the operation imagery; And a user-defined step of determining a channel and a frequency band with high accuracy through analysis of decision coefficients in a machine learning algorithm; And a determination step of comparing information stored in the learning step with information stored in the guessing step.

6 is a diagram illustrating an example of an EEG signal learning process of the EEG training apparatus according to an embodiment of the present invention. Each step of the EEG signal learning process can be performed by the EEG training apparatus 100 described with reference to FIG.

The EEG training apparatus 100 requests a measurement of EEG data in a rest state (Step 601) when EEG data in a rest state is not inputted in advance (Step 602), and then transmits the EEG data to the EEG data received in response to the request An FFT (Fast Fourier Transform) filter can be applied to the Fourier transform (step 603).

The EEG training apparatus 100 receives EEG data in real time (601) when EEG data of a rest state is input in advance from the EEG apparatus 101 (604), performs FFT (Fast Fourier Transform) on the received EEG data, A filter may be applied (605).

The EEG training apparatus 100 can analyze the regression analysis decision coefficient r ² according to the EEG learning received in real time from the EEG 101 if the training set is not input in advance (606) (607), and a channel and a frequency band with high accuracy can be selected according to the regression analysis decision coefficient (r ² ) (608). At this time, the brain-training apparatus 100 extracts feature points from the frequency band of the selected channel in step 608, generates a graph of the selected channel and frequency band in step 608, A training model can be created with the training set of NN (611).

The EEG training apparatus 100 may analyze the regression analysis decision coefficient r ² according to EEG learning received in real time from the EEG 101 if the training set is input in advance (606) , And the regression analysis coefficient (r ² ), it is possible to select the channel and the frequency band with high accuracy (613). At this time, the EEG 100 extracts feature points from the frequency band of the selected channel in step 613, and generates a graph of the selected channel and frequency band in step 613 (step 615) . Then, the EEG 100 compares the training set inputted through the training set and the test set generated through the steps 612 to 615 through the k-NN (step 616), and outputs the comparison result to the user in an identifiable form (617). The EEG training apparatus 100 may generate 611 a training set of k-NN generated through steps 612 through 615 if the user desires to create a new model 618. Thereafter, when the user inputs a test set (619), the EEG apparatus 100 stores the test set, and then returns to step 604 to repeat the above steps (604 to 618).

The process of analyzing the regression analysis decision coefficient (r ² ) (607, 612) is as follows.

(1) where x is a set 1 and y is a set 2, a regression analysis decision coefficient (r ² ) can be defined as shown in Equation (5).

here,

In other words,

Covariance

Lt; / RTI >

Each variance

.

The learning model for the EEG signal is generated based on the result calculated for each channel and frequency through the above equation. The higher the regression analysis decision coefficient (r ² ), the better the channel and frequency to distinguish the two data sets. For example, the EEG apparatus 100 can provide the analysis result (similarity) of the regression analysis decision coefficient r ² through the channel and frequency band (Channel & Frequency) tab of FIG. 5, it is possible to select a channel and a frequency which are helpful in distinguishing the rest state from the operation state.

Accordingly, the present invention provides a simulator for learning the characteristics of the user's brain waves and training the user's ability to generate brain waves for efficient use of the EEG signal-based interface device.

7 is a block diagram for explaining an example of the internal configuration of a computer system in an embodiment of the present invention. The computer system 700 includes at least one processor 710, a memory 720, a peripheral interface 730, an input / output subsystem 740, A power circuit 750, and a communication circuit 760. [ At this time, the computer system 700 may correspond to an EEG training apparatus.

The memory 720 may include, for example, a high-speed random access memory, a magnetic disk, a SRAM, a DRAM, a ROM, a flash memory, or a non-volatile memory. have. The memory 720 may include software modules, a set of instructions, or various other data required for operation of the computer system 700. At this point, accessing memory 720 from other components, such as processor 710 or peripheral device interface 730, may be controlled by processor 710.

Peripheral device interface 730 may couple the input and / or output peripheral devices of computer system 700 to processor 710 and memory 720. The processor 710 may execute a variety of functions and process data for the computer system 700 by executing a software module or set of instructions stored in the memory 720.

The input / output subsystem 740 may couple various input / output peripheral devices to the peripheral interface 730. For example, input / output subsystem 740 may include a controller for coupling a peripheral device such as a monitor, keyboard, mouse, printer, or a touch screen or sensor, as needed, to peripheral interface 730. According to another aspect, the input / output peripheral devices may be coupled to the peripheral device interface 730 without going through the input / output subsystem 740.

The power circuitry 750 may provide power to all or a portion of the components of the terminal. For example, the power circuitry 750 may include one or more power sources, such as a power management system, a battery or alternating current (AC), a charging system, a power failure detection circuit, a power converter or inverter, And may include any other components for creation, management, distribution.

The communication circuitry 760 may enable communication with other computer systems using at least one external port. Or as described above, the communication circuitry 760 may also communicate with other computer systems by sending and receiving RF signals, also known as electromagnetic signals, including RF circuits.

7 is merely an example of a computer system 700, and the computer system 700 may have additional components that are omitted from FIG. 7, or that are not shown in FIG. 7, Lt; RTI ID = 0.0 > components. &Lt; / RTI > For example, in addition to the components shown in FIG. 7, a computer system for a mobile communication terminal may further include a touch screen, a sensor, etc., and may be connected to various communication methods (WiFi, 3G, LTE , Bluetooth, NFC, Zigbee, etc.). The components that may be included in computer system 700 may be implemented in hardware, software, or a combination of both hardware and software, including one or more signal processing or application specific integrated circuits.

The apparatus described above may be implemented as a hardware component, a software component, and / or a combination of hardware components and software components. For example, the apparatus and components described in the embodiments may be implemented as a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit, a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For ease of understanding, the processing apparatus may be described as being used singly, but those skilled in the art will recognize that the processing apparatus may have a plurality of processing elements and / As shown in FIG. For example, the processing unit may comprise a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as a parallel processor.

The software may include a computer program, code, instructions, or a combination of one or more of the foregoing, and may be configured to configure the processing device to operate as desired or to process it collectively or collectively Device can be commanded. The software and / or data may be in the form of any type of machine, component, physical device, virtual equipment, computer storage media, or device , Or may be permanently or temporarily embodied in a transmitted signal wave. The software may be distributed over a networked computer system and stored or executed in a distributed manner. The software and data may be stored on one or more computer readable recording media.

The method according to an embodiment may be implemented in the form of a program command 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 to be recorded on the medium may be those specially designed and configured for the embodiments or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. For example, it is to be understood that the techniques described may be performed in a different order than the described methods, and / or that components of the described systems, structures, devices, circuits, Lt; / RTI > or equivalents, even if it is replaced or replaced.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims

At least one program loaded memory; And
At least one processor
Lt; / RTI >
Wherein the at least one processor, under control of the program,
A process of acquiring measured EEG data from a user through an EEG measurement device after presenting imaginary motion for EEG training;
Extracting feature information corresponding to the imaging operation from the brain wave data to learn the characteristics of the user's brain wave with respect to the imaging operation;
Providing an additional operation or image according to the feature information as a guide or a comment related to the imaginary operation when the feature information is out of a critical range;
Providing a training simulation for the user's brain wave generation with respect to the imaginary motion while repeating the acquisition, the learning, and the providing; And
A process of transmitting an analysis result of an EEG trained through the user's brain wave generation training to an input of an interface device for controlling another device
Lt; / RTI >
The learning process includes:
analyzing a regression analysis decision coefficient for the EEG data using a k-nearest neighbors algorithm, and calculating a feature corresponding to the imaginary motion in a frequency band of the selected channel according to the regression analysis decision coefficient And generating a learning model of the k-nearest neighbor algorithm
Lt; / RTI >

delete

The method according to claim 1,
The learning process includes:
Extracting a frequency band necessary for EEG analysis from the EEG data using Fast Fourier Transform, extracting feature information corresponding to the imaging operation from brain wave data of the separated frequency band, Storing information according to the imaginary operation
Lt; / RTI >

delete

At least one program loaded memory; And
At least one processor
Lt; / RTI >
Wherein the at least one processor, under control of the program,
A process of acquiring measured EEG data from a user through an EEG measurement device after presenting imaginary motion for EEG training;
Extracting feature information corresponding to the imaging operation from the brain wave data to learn the characteristics of the user's brain wave with respect to the imaging operation;
Providing an additional operation or image according to the feature information as a guide or a comment related to the imaginary operation when the feature information is out of a critical range;
Providing a training simulation for the user's brain wave generation with respect to the imaginary motion while repeating the acquisition, the learning, and the providing; And
A process of transmitting an analysis result of an EEG trained through the user's brain wave generation training to an input of an interface device for controlling another device
Lt; / RTI >
The learning process includes:
A first step of analyzing a regression analysis decision coefficient for the obtained EEG data;
A second step of selecting a channel and a frequency band for EEG learning according to the regression analysis decision coefficient analyzed in the first step;
A third step of extracting feature information corresponding to the imaging operation in a frequency band of the selected channel in the second step;
A fourth step of generating a learning model in which the feature information extracted in the third step is a training set;
A fifth step of analyzing a regression analysis decision coefficient for EEG data acquired after the learning model is generated;
A sixth step of selecting a channel and a frequency band for EEG learning according to the regression analysis decision coefficient analyzed in the fifth step;
A seventh step of extracting feature information corresponding to the imaging operation in a frequency band of the channel selected in the sixth step; And
An eighth step of comparing the feature information extracted in the seventh step with the training set and outputting a comparison result;
&Lt; / RTI >

The interface device as claimed in claim 1, 3, or 5, wherein the user's brain wave characteristic is an interface variable for inputting a computer.