KR20200052205A - Brain-computer interface systems and method for analysing brain wave signals expressed by motor imagery - Google Patents

Abstract

The present invention relates to a brain-computer interface system and a method for analyzing a brainwave signal expressed by motor imagery. The method comprises: (a) a step of acquiring a source brainwave signal expressed in motor imagery during a preset measurement time, and filtering the source brainwave signal in a base frequency band associated with motor imagery to generate a brainwave signal; (b) a step of performing an optimal performance search algorithm for calculating a time range and an optimal expression frequency domain of a brainwave feature pattern for the brainwave signal to divide the brainwave signal into preset frequency intervals to convert the brainwave signal into n sub-band signals and then detect a maximum performance time range for each of the sub-band signals; (c) a step of extracting a brainwave feature based on the maximum performance time range for each sub-band signal, using the extracted brainwave feature to generate a classification model for recognizing a motion imagined or intended by a user, and then calculating the accuracy for a classification result of the classification model to perform classification performance evaluation; and (d) a step of detecting an optimal sub-band signal having the highest performance based on result values of the classification performance evaluation, and a maximum performance time range linked to the optimal sub-band signal to provide the optimal sub-band signal and the maximum performance time range in an optimized motor imagery frequency-time domain for each user. The optimal performance search algorithm performs performance evaluation in a plurality of time ranges generated by setting a time search window having a preset time range and applying the time search window at preset critical time intervals in the measurement time of each sub-band signal.

Description

BRAIN-COMPUTER INTERFACE SYSTEMS AND METHOD FOR ANALYSING BRAIN WAVE SIGNALS EXPRESSED BY MOTOR IMAGERY}

The present invention relates to a brain-computer interface system that provides an optimized frequency domain and a time domain for extracting EEG features related to motion imagination for each user, and a method of analyzing EEG signals expressed by motion imagination.

EEG is an electrical signal that is expressed when information between the body's nervous system and the brain nerve is transmitted. In particular, EEG measured using a non-invasive method can be measured through electrodes attached to the scalp without additional surgical procedures. , It can be an important means to grasp the user's intention by measuring real-time activity of the brain.

In general, it is known that EEG can be classified into various approaches such as temporal characteristics, band-specific characteristics according to frequency, and spatial characteristics. For example, alpha waves (frequency 8-13 Hz, amplitude 20-60 V) expressed in a comfortable state with eyes closed, beta waves (14-30 Hz, amplitude 2-20 V) expressed in a consciously focused state, shallow sleep Theta waves (frequency 4 ~ 7Hz, amplitude 20 ~ 100V) expressed in are widely known, but besides these, mu-band waves (Mu) generated in the brain cortex when intention or thought related to body movement -band) (8 ~ 12Hz) is also known to exist.

The non-invasive brain-machine interface is an interface technology that recognizes the user's intention through a non-invasive brain wave and controls an external machine. It has been used mainly in the medical field, such as, a speller for communication. However, due to the recent development of EEG analysis technology, non-invasive brain-machine interfaces have been applied in various fields through the development of daily life assistance services for the general public.

  In particular, brain-machine interface technology based on EEG analysis according to motion imagery is mainly used because it is easy to recognize a user's intention without a separate external stimulus. EEG analysis according to motion imagination is a technique to grasp the user's intention through EEG expressed only by thinking of moving a specific part of the body without moving the muscle. Motion imagination can activate the primary motor cortex of the brain, and brain waves can be expressed in the form of mu-band waves in the frequency range of 8 to 12 Hz while imagining the motion of each part.

Among the types of features for representative motion imagination analysis that can be found in mu-band waves are Event-Related Desynchronization (ERD) and Event-Related Synchronization (ERS). ERD refers to a phenomenon in which the mu-band wave temporarily decreases in the sensory motor cortex when the user intends or imagines the movement, and ERS temporarily increases the beta wave (18-25 Hz) after the user moves. Refers to the phenomenon.

In order to improve the performance of the brain-machine interface capable of performing high-level tasks, the current technology approach is largely divided into an analysis technique based on the motion imagination and the characteristics of the EEG signal expressed through external stimuli. Lose.

As a BCI system technology based on motion imagination, Republic of Korea Patent Registration No. 10-1293446 (Invention name: brain wave classification method and device for imagining motion), Korea Patent Registration No. 10-1518575 (Invention name: BCI User Intention Cognitive Analysis Method), Republic of Korea Patent Registration No. 10-1619973 (Invention Name: Motion Imagination EEG Determination Method and Discrimination System Using Spectral Analysis and Vector Quantization), Korea Patent Registration No. 10-1446845 (Invention Name: Vehicle control method using EEG, device, and a recording medium recording a program for performing the method).

Also, as an analysis technology using features of EEG signals, Korean Patent Registration No. 10-1680995 (Invention name: Brain-computer interface system based on collected temporal and spatial patterns of biophysical signals), Korean Patent Registration No. 10- No. 1740894 (Invention name: Brain-machine interface system and method using spatial filter), Republic of Korea Patent Registration No. 10-1205892 (Invention name: Method for generating 2-dimensional space-frequency ERD / ERS pattern from EEG signal, A method for classifying human intention using this two-dimensional space-frequency ERD / ERS pattern and a brain-computer interface device using the EEG signal classified by this classification method as an input signal, Korean Patent Registration No. 10-1068017 (Invention) Name: Synthetic common spatial pattern analysis method for brain-computer interface and EEG analysis method using the same.

Patent registration No. 10-1293446 relates to an EEG classification method for imagining motion and its technology, and proposes a technology for classifying a left motion imagination EEG signal and a right motion imagination EEG signal. The EEG classification technology applied to the technology has an advantage of improving classification accuracy using a support vector generation algorithm (GMM) and a classification algorithm (SVM). However, this technique has a limitation in that it is difficult to realize since it can be implemented only when the characteristics of the EEG containing the motion imagination information are accurately extracted from the features of the ERD / ERS pattern that are not clearly revealed.

 Registered Patent No. 10-1518575 uses the digital notch filter and the band pass filter to express brain signals that are expressed when imagining motions, and to characterize brain signals only in the frequency band of mu-band waves. A technique that can be extracted is proposed. However, this technology generally has a limitation in that when using a mu-band wave, a motion imagination is recognized through brain waves, and each user may not reflect brain wave characteristics from various frequency bands.

 Patent No. 10-1619973 has a feature of filtering and using a brain signal related to exercise imagination over the entire 8 ~ 30Hz frequency band. When a brain signal feature is extracted by filtering all time ranges of the entire frequency band, a brain signal including feature patterns and various noises not related to motion imagination or motion intention may be extracted. As a result, there is a limitation in that it is impossible to accurately recognize a brain signal feature pattern expressed when imagining motion.

Patent No. 10-1446845 relates to a vehicle control apparatus using brain waves that recognize a user's intention with at least one brain wave characteristic of motion imagination (left, right hand, and tongue motion imagination). The EEG feature recognizes the user's intention using the frequency band of the SMR (SensoriMotor Rhythm) region. When various motion imagination experiment tasks extract brain signal features using a frequency band of a simple SMR region, it is difficult to extract optimized brain signal features according to each motion imagination experiment task.

 Patent No. 10-1680995 proposes a technique of correlating collected temporal and spatial patterns of biophysical signals associated with a user's brain activity to identify a user's brainwave signal characteristics. That is, the technology discloses a system that connects recorded brain activity in a topologically configured brain region to a step of finally performing input to a computer through a method of correlating with movement of a body position corresponding to each part of the brain. Doing. These systems utilize temporal and spatial features to analyze EEG signals related to motion imagination, but it is practical to detect frequency and time domains specific to the user by inferring effective time intervals by searching for a clear point of time that the user imagines. There is a problem that is difficult.

 Patent No. 10-1740894 relates to a brain-machine interface system and method using a spatial filter, calculates variance between the same operation and different operations of EEG data, and variance between the same operation and different operations of the EEG data A system for classifying a motion intended by a user is proposed by generating a spatial filter using a ratio. To this end, the EEG data is extracted for each operation from the start to the end of the operation for a plurality of operations, and stored for each operation. In this method, the entire time period that is not specialized to the user's state in generating a spatial pattern filter Due to the limitations of using the features extracted from, there is a limit to effective performance.

 In order to stably extract the characteristics of various EEG signals, Patent No. 10-1205892 proposes a method for generating a two-dimensional space-frequency ERD / ERS pattern from EEG signals after measuring the EEG of a subject's stable state. have. The user's intention is classified using this method, and the classified EEG signal may be provided as an input means to the computer system. Since these technologies use the entire frequency band of the brain signal obtained in the user's stable state, it is difficult to obtain brain signal information for a specific number of movements, and the part of optimizing the time interval is not taken into account, so it is not necessary in the feature extraction process. There is a problem with efficiency.

Patent No. 10-1068017 proposes a method of extracting EEG features through a synthetic common spatial pattern analysis method of ERD / ERS patterns, which are representative brain signal features expressed when imagining motion. However, the ERD / ERS pattern has a large variation depending on the user, and it is difficult to extract the characteristics of the EEG because it is not a phenomenon that is clearly seen in the sensory cortical region. To solve this problem, a common spatial pattern analysis method is used, but the common spatial pattern analysis method is difficult to apply to users (BCI illiteracy) who do not clearly exhibit spatial distribution or characteristics of the brain. There may be limitations in which common space patterns can all come out differently.

As described above, the conventional motion imagination-based BCI system filters the brain signal by empirically filtering the raw EEG signal obtained from the user into a specific frequency band such as a mu-band or a frequency band corresponding to an ERD / ERS pattern feature. Although analyzing, the performance of the motion imagination-based BCI system has to be deteriorated and the inefficiency is limited because the EEG characteristics are extracted without considering the exact time interval in which motion imagination is performed for each user.

In addition, even if the performance of a specific motion imagination class is high in a conventional BCI system, performance is measured by including information irrelevant to the corresponding EEG whether the performance is actually derived from the characteristics of the EEG generated through the imagination of the motion. There is a problem in that it is impossible to confirm whether it is reliable. That is, although the EEG characteristic pattern is differently expressed for each person, EEG characteristics are extracted by applying a certain range of frequency bands to EEGs of all users, so that EEG information that enables the most accurately expressed brain signal characteristics for each user's imagination There is a problem of not being considered.

Accordingly, an optimized frequency band capable of accurately extracting motion imagination information is required for each user, and at the same time, time ranges are applied for each optimal frequency band for extracting EEG features because the time intervals at which users voluntarily execute motion imagination are different. Therefore, analyzing the common space pattern can help improve BCI performance.

The brain-computer interface system according to the present invention and the method of analyzing the EEG signal expressed by the motion imagination are optimized for each user to improve the performance of the BCI system using the EEG obtained through the general motion imagination BCI experiment paradigm. I would like to suggest a way to select the time range that gives the highest performance.

Method for analyzing the brain wave signal expressed by the motion imagination according to an embodiment of the present invention, in the method for analyzing the brain wave signal expressed by the motion imagination performed by the brain-computer interface system, a) a predetermined Obtaining a primitive EEG signal expressed during imagination of motion during a measurement time, and filtering the primitive EEG signal into a fundamental frequency band associated with imagination of motion to generate an EEG signal; b) by performing an optimal performance search algorithm for calculating the optimal expression frequency range and time range of the EEG feature pattern for the EEG signal, dividing the EEG signal into predetermined frequency intervals and converting the EEG signal into n subband signals Then detecting a maximum performance time range for each subband signal; c) extracting EEG features based on the maximum performance time range for each sub-band signal, and using the extracted EEG features to generate a classification model that recognizes a user's imagination or intended behavior, and then classifying the classification model Performing classification performance evaluation by calculating accuracy for the; And d) detecting the optimal sub-band signal having the highest performance based on the result of the classification performance evaluation, and the maximum performance time range associated with the optimal sub-band signal, thereby providing an optimized frequency-time domain for each user. The optimum performance search algorithm is generated by setting a time search window having a preset time range and applying the time search window at a predetermined threshold time interval to the measurement time of each subband signal. Performance evaluation is performed in a plurality of time ranges.

A brain-computer interface system according to another embodiment of the present invention includes: a brain-computer interface system for analyzing brainwave signals expressed by motion imagination, comprising: a sensor module that detects a user's brainwave and provides a raw brainwave signal; A memory in which a program for performing a method for analyzing an EEG signal expressed by the motion imagination is recorded; And a processor for executing the program, wherein the processor acquires a raw EEG signal expressed when an operation is imagined for a predetermined measurement time by executing the program, and the basic EEG signal is associated with the motion imagination. Filtering by frequency band generates an EEG signal, and performs an optimal performance search algorithm to calculate an optimal frequency range and time range of the EEG characteristic pattern for the EEG signal, and divides the EEG signal into preset frequency intervals. After converting to n sub-band signals, the maximum performance time range is detected for each sub-band signal, EEG characteristics are extracted based on the maximum performance time range for each sub-band signal, and the extracted EEG features are used for the user. Creates a classification model that recognizes imaginary or intended behavior Then, the classification performance evaluation is performed by calculating the accuracy of the classification result of the classification model, and based on the result of the classification performance evaluation, the optimal subband signal having the highest performance and the maximum performance associated with the optimal subband signal. Detecting a time range and providing an optimized operation for each user in an imaginary frequency-time domain, wherein the optimal performance search algorithm sets a time search window having a preset time range, and is preset to the measurement time of each sub-band signal. The performance evaluation is performed in a plurality of time ranges generated by applying the time search window at a critical time interval.

According to the above-described problem solving means of the present invention, in the operationally based BCI system, it is possible to detect a frequency range optimized for each user and a maximum performance time range in a short section in which the EEG characteristics are most clearly expressed in the frequency domain. , As a result, it is possible to more accurately extract the characteristic patterns of the EEG expressed when imagining the motion, thereby greatly improving the accuracy of the BCI system.

In addition, since the present invention does not perform motion imagination in a short period of time due to the characteristics of the motion imagination, since a strong brain signal is expressed in a short time period in which the user voluntarily executes the motion imagination, the maximum in the motion imagination frequency domain optimized for each user It is possible to operate a stable BCI system with higher performance and faster speed than the previous one by extracting EEG features by specifying a time range that exhibits performance.

In addition, the present invention does not utilize a fixed time range that is a wide time interval (for example, about 3000 ms), that is, the entire time of measuring the EEG, and precisely reduces the time range for extracting EEG features, so in the EEG analysis process, By reducing the computation and noise required, it is possible to increase the reliability and convenience of the BCI system.

1 is a view for explaining the configuration of a brain-computer interface system according to an embodiment of the present invention.
2 is a diagram illustrating a user intention recognition process of a general brain-computer interface system.
3 is a flowchart illustrating a method of analyzing an EEG signal expressed by motion imagination according to an embodiment of the present invention.
4 is a flowchart illustrating an optimal performance search algorithm of FIG. 3.
FIG. 5 is a diagram illustrating a process of applying a time search window for each sub-band and selecting a maximum performance time range of the optimal performance search algorithm of FIG. 3.

Hereinafter, with reference to the accompanying drawings, embodiments of the present invention will be described in detail so that those skilled in the art to which the present invention pertains can easily practice. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. In addition, in order to clearly describe the present invention in the drawings, parts irrelevant to the description are omitted, and like reference numerals are assigned to similar parts throughout the specification.

Throughout the specification, when a part is "connected" to another part, this includes not only "directly connected" but also "electrically connected" with another element in between. . Also, when a part is said to "include" a certain component, this means that other components may be further included rather than excluding other components, unless otherwise stated.

Hereinafter, a user-optimized non-invasive brain-machine interface system according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a view for explaining the configuration of a brain-computer interface system according to an embodiment of the present invention.

Referring to FIG. 1, a brain-computer interface system 100 that analyzes an EEG signal expressed by a motion imagination based on EEG characteristics of an EEG signal obtained from a user includes a communication module 110, a memory 120, It includes a processor 130, a database 140 and a sensor module 150.

The communication module 110 provides a communication interface necessary for providing a transmission / reception signal of the brain-computer interface system 100 and an external device in packet data form in cooperation with a communication network. Here, the communication module 110 may be a device including hardware and software necessary for transmitting and receiving a signal such as a control signal or a data signal through a wired or wireless connection with another network device.

In the memory 120, a program for performing a performance improvement method of a brain-computer interface system for providing optimized frequency information for each user and a time range for feature extraction is recorded. Also, the memory 120 temporarily or permanently stores data processed by the processor 130. Here, the memory 120 may include volatile storage media or non-volatile storage media, but the scope of the present invention is not limited thereto.

The processor 130 controls the entire process of providing a method for improving the performance of the brain-computer interface system to provide a time range that exerts maximum performance in a frequency band optimizing for each user, and the brain-computer interface system ( In order to improve the performance of 100), the frequency domain and time range in which the characteristics of the EEG signal are most clearly expressed are found to provide an optimized imaginary frequency-time domain. Each operation performed by the processor 130 will be described in more detail later.

Here, the processor 130 may include all kinds of devices capable of processing data, such as a processor. Here, the 'processor (processor)', for example, may mean a data processing device embedded in hardware having physically structured circuits to perform functions represented by codes or instructions included in a program. As an example of such a data processing device embedded in hardware, a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, and an application-specific integrated ASIC circuit, a field programmable gate array (FPGA), or the like, but the scope of the present invention is not limited thereto.

The database 140 stores data accumulated while performing a method for improving the performance of the brain-computer interface system. For example, the database 140 stores raw EEG signals, classification performance index information for a classification model, and optimized motion imagination frequency-time domain information for each user.

The sensor module 150 detects a raw EEG signal related to the motion imagination in real time while being worn by the user.

2 is a diagram illustrating a user intention recognition process of a general brain-computer interface system.

Referring to FIG. 2, the brain-computer interface system 100 measures a user's original EEG signal related to real-time motion imagination through the sensor module 150 using a predetermined experimental paradigm. At this time, the experimental paradigm mainly directs the user to a specific task as an image, and the user performs a specific task according to a given queue. For example, a specific task is to think of image data related to imagining motions such as right / left hand grip, right / left hand stretch, and wrist left / right rotation.

The processor 130 fixes the measurement time until the imagination of the motion is finished based on the given cue and collects the raw EEG signal, and filters the raw EEG signal collected during the measurement time into a basic frequency band. In general, the fundamental frequency band includes a mu band (8 to 13 Hz), which is a significant frequency domain in relation to the motion imagination, and the measurement time is fixed to 3000 ms when the experimental paradigm is conducted for 3000 ms.

The processor 130 extracts spatial characteristics using a common spatial pattern (CSP), and performs the classification for various motion imaginations through the classification model 131 using a machine learning technique, so that the user is the first Decoding based on the EEG signal for the imagined motion is performed to output the final classification result.

At this time, the fundamental frequency band and the measurement time are frequency domains and time ranges that have been empirically obtained through experiments. However, the brain-computer interface system has a different optimal frequency domain in which the EEG characteristics are most clearly expressed for each user, and at the same time, the range of motion imagination expression time that the user imagines motion from after the given cue to the end of one experiment process. Is a part of the total measurement time, and the length and time point of the motion imagination expression time range vary from person to person.

Therefore, the method for improving the performance of the brain-computer interface system of the present invention searches for an accurate frequency domain in which a motion imagination is expressed for each user, and a maximum performance time range in which a user voluntarily executes a motion imagination in association with the discovered frequency domain After calculating is, the system performance may be improved by analyzing the common spatial pattern by applying the optimized frequency-time domain of the user's personalized optimized motion.

3 is a flowchart illustrating a method of analyzing an EEG signal expressed by motion imagination according to an embodiment of the present invention, FIG. 4 is a flowchart illustrating the optimal performance search algorithm of FIG. 3, and FIG. 5 is FIG. 3 A diagram for explaining a process of applying a time search window for each sub-band of an optimal performance search algorithm and selecting a maximum performance time range.

The brain-computer interface system 100 measures the user's raw EEG signal in real time through the sensor module 150 (S110), and the preprocessor to minimize the influence of ambient noise sound and power DC noise on the raw EEG signal ( (Not shown) to perform the filtering to the fundamental frequency band of 0.1 ~ 50 kHz to output the EEG signal (S120).

The fundamental frequency band is empirically in the brain-computer interface technology, certain frequency bands (8 ~ 12 Hz) such as mu bands, ERD (Event-Related Desynchronization) in which EEG decreases at a specific frequency according to stimulation, and ERS in which EEG increases The frequency band corresponding to the (Event-Related Synchronization) pattern can be used selectively.

The processor 130 generates n subband signals by dividing the EEG signals filtered by the basic frequency band into random or regular frequency intervals according to a predetermined rule (S130). For example, the processor 130 divides the EEG signal at frequency intervals within 3 to 5 kHz, and 0.1 to 3 kHz, 4 to 7 kHz, 8 to 13 kHz,… , 45 ~ 50Hz sub-band signal is generated.

The processor 130 performs an optimal performance search algorithm that calculates a maximum performance time range exerting maximum performance for each subband signal (S140). As a result of performing the optimal performance search algorithm, an optimal sub-band optimized for imagination of motion for each user and a maximum performance time range associated therewith are detected (S150).

As shown in FIG. 4, the optimal performance search algorithm converts the EEG signal of 0.1-50 Hz to n sub-band signals, and a time search window having a preset time range is displayed for the entire measurement time of each sub-band signal. A plurality of time ranges are generated while applying the set threshold time intervals (S410). For example, when the total measurement time is 3000 ms, the time search window is 1000 ms, and the critical time interval is 200 ms, the sub-band signal is applied with a time search window 0 to 1000 ms, 200 to 1200 ms, 400 to 1400 ms,… , Time ranges of 2000 to 3000 ms can be generated.

The optimal performance search algorithm extracts EEG features using signal power (Frequency Power) or spatial patterns for each time range of each sub-band signal, and then uses the extracted EEG features to classify the user's motion imagination or motion imagination classes for intention The learned classifier is generated (S420, S430). The optimal performance search algorithm evaluates the classification performance of the classifier generated for each sub-band signal and its time range (S440).

That is, the optimal performance search algorithm calculates the accuracy of the classification result of the classifier trained using any of the methods of cross-validation, cross-information calculation, and Fisher score, and calculates a performance index, and based on the calculated performance index Through a probabilistic approach such as Bayesian probability, a maximum performance time range for each subband signal capable of exerting the highest performance for a user is obtained.

For example, when performing performance evaluation by applying a k-fold cross-validation method, the optimal performance search algorithm separates the user's raw EEG signal into k sub-data sets and one of the separated k sub-data sets. The k models are estimated by using k-1 sub-datasets excluding the data set as a training set. Using the average value of the estimated results, the performance of the model can be objectively measured without relying on the data used for training.

As illustrated in FIG. 5, the optimal performance search algorithm detects a maximum performance time range that is an optimal expression time of an optimal subband signal having the highest performance and an EEG characteristic pattern among performance indicators calculated for each subband signal. The optimal performance search algorithm is based on the Bayesian probability theory based on the performance index in the time range candidate group having the maximum performance for each subband signal, and imagines the subband and time range that can continuously show the highest recognition rate for the user. It is selected as a candidate group in the frequency-time domain, and information on a time band and a subband signal having a relatively low recognition rate is removed (S450).

The optimal performance search algorithm optimizes the optimal subband and maximum performance time range for the user if the performance improvement is not made in a subband and time range other than the selected optimal subband and maximum performance time range. It is provided as an area (S460).

Referring to FIG. 3 again, the processor 130 extracts EEG characteristics based on the optimized motion imagination frequency-time domain for the user (S160), and generates a final classification model using the extracted EEG characteristics After that, the performance of the final classification model is repeatedly compared to search for a frequency domain and a time range that show higher performance than the optimized motion imaginary frequency-time domain (S170).

As a result of repeatedly performing the optimal performance search algorithm, the processor 130 stops the optimal performance search algorithm when the performance improvement of the brain-computer interface (BCI) system is no longer performed (S180 to S200). The mathematical or statistical method applied to calculate the performance index in the optimal performance search algorithm can be used without limitation in any method other than cross-validation, mutual information calculation, Fisher score, Bayesian probability, and applied to the mathematical statistical method. Even if the algorithm is changed, the degree of performance improvement of the entire BCI system is not significant.

As described above, the BCI system of the present invention uses the original EEG signal obtained through the motion imagination BCI experiment paradigm, and the optimal performance search algorithm selects the maximum performance time range that shows the highest performance in the motion imagination frequency band optimized for each user. Gives

Previously, the EEG characteristics were extracted by applying a specific frequency band of the EEG known to all users to the EEG signal of the user, but the optimal performance search algorithm is optimized for each user. By precisely reducing the range of the time section, it is possible to obtain a time range of a short section in which the EEG characteristics of the user are most clearly expressed by detecting an accurate time range in which the user's own motion imagination is arbitrarily executed.

Therefore, the present invention can provide an optimized BCI system for each user as compared to the existing BCI system, and precisely reduces the range of time intervals for extracting the EEG feature of the user, thereby reducing the computational speed and reducing unnecessary noise. Can be.

The method of analyzing the EEG signal expressed by the motion imagination according to the embodiment of the present invention described above, may also be implemented in the form of a recording medium including instructions executable by a computer, such as a program module executed by a computer. Can be. Such recording media include computer readable media, which can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. In addition, computer readable media includes computer storage media, which are volatile and nonvolatile implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. , Removable and non-removable media.

The above description of the present invention is for illustration only, and those skilled in the art to which the present invention pertains can understand that the present invention can be easily modified into other specific forms without changing the technical spirit or essential features of the present invention. will be. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined form.

The scope of the present invention is indicated by the following claims rather than the above detailed description, and it should be interpreted that all changes or modified forms derived from the meaning and scope of the claims and equivalent concepts thereof are included in the scope of the present invention. do.

100: brain-computer interface system
110: communication module 120: memory
130: processor 140: database
150: sensor module

Claims (15)

A method of analyzing brainwave signals expressed by an imagination of motion performed by a brain-computer interface system, the method comprising:
a) acquiring a raw EEG signal expressed during imagination of operation during a predetermined measurement time, and filtering the raw EEG signal into a fundamental frequency band related to the imagination of operation to generate an EEG signal;
b) by performing an optimal performance search algorithm for calculating the optimal expression frequency range and time range of the EEG feature pattern for the EEG signal, dividing the EEG signal into predetermined frequency intervals and converting the EEG signal into n subband signals Then detecting a maximum performance time range for each subband signal;
c) extracting EEG features based on the maximum performance time range for each sub-band signal, and using the extracted EEG features to generate a classification model recognizing a user's imagination or intended motion, and then classifying the classification model Performing classification performance evaluation by calculating accuracy for the; And
d) Based on the result of the classification performance evaluation, the optimal sub-band signal having the highest performance and the maximum performance time range associated with the optimal sub-band signal are detected and provided as an optimized imaginary frequency-time domain for each user. Including the steps,
The optimal performance search algorithm sets a time search window having a preset time range, and a plurality of time ranges generated by applying the time search window at predetermined threshold time intervals to the measurement time of each subband signal. A method of analyzing brainwave signals expressed by motion imagination, which is to perform a performance evaluation. According to claim 1,
Step a) is,
The base frequency band is a mu band (Mu-band), ERD (Event-Related Desynchronization) of the frequency band corresponding to the characteristics of the or ERS (Event-Related Synchronization) characteristics of at least one of the frequency band of the frequency band A method of analyzing brainwave signals expressed by motion imagination. According to claim 1,
Step b),
A method of analyzing the EEG signal expressed by imagination of motion by dividing the EEG signal having the basic frequency band into random or predetermined rules at frequency intervals within 3 to 5 Hz. According to claim 1,
The optimal performance search algorithm,
Set a time search window having a preset time range (T _d ),
A plurality of time ranges for subdividing the measurement time (T _total, T _total > T _d > T _t ) are generated by applying the time search window at a predetermined threshold time interval (T _t ) for each subband signal,
EEG characteristics are respectively extracted from a plurality of time ranges generated for each of the sub-band signals,
After generating a classifier learning the motion imaginary classes using the extracted EEG feature, calculating the accuracy of the classification result of the classifier to calculate a performance index,
A method of analyzing an EEG signal expressed by motion imagination, which is to detect an optimal subband signal having the highest performance and a maximum performance time range based on the performance index. The method of claim 4,
The optimal performance search algorithm,
A method of analyzing the EEG signal expressed by the motion imagination, which removes information on the subband signal and the time range having the lowest performance based on the performance index. According to claim 1,
Step c),
A method of analyzing EEG signals expressed by motion imagination, which extracts EEG characteristics for EEG signals for each sub-band signal using a feature extraction algorithm using a common spatial pattern (CSP) or frequency intensity . According to claim 1,
Step c),
Calculating the accuracy of the classification result of the class model by using any one of a cross validation method, a mutual information calculation method, or a Fisher Score method to calculate a performance index A method of analyzing EEG signals expressed by phosphorus and motion imagination. According to claim 1,
Step d),
Performance evaluation is repeatedly performed through the optimal performance search algorithm, and when the result value higher than the result of the classification performance evaluation for the optimal subband signal and the maximum performance time range is not detected, the optimal performance search algorithm is terminated. How to analyze the brain wave signal expressed by the motion imagination. In the brain-computer interface system for analyzing the EEG signal expressed by the motion imagination,
A sensor module that detects a user's EEG and provides a raw EEG signal;
A memory in which a program for performing a method for analyzing an EEG signal expressed by the motion imagination is recorded; And
It includes a processor for executing the program,
The processor, by the execution of the program,
Acquiring a raw EEG signal expressed during imagination of motion during a predetermined measurement time, filtering the raw EEG signal into a fundamental frequency band related to imagination of motion to generate an EEG signal,
After performing an optimal performance search algorithm for calculating an optimal expression frequency region and time range of the EEG feature pattern for the EEG signal, the EEG signal is divided into predetermined frequency intervals, converted into n sub-band signals, and then the The maximum performance time range is detected for each sub-band signal,
EEG characteristics are extracted based on the maximum performance time range for each sub-band signal, and a classification model for recognizing a user's imagined or intended motion is generated using the extracted EEG characteristics, and then the classification result of the classification model is analyzed. Classify performance evaluation by calculating accuracy,
Based on the result of the classification performance evaluation, the optimal sub-band signal having the highest performance and the maximum performance time range associated with the optimal sub-band signal are detected and provided as an optimized imaginary frequency-time domain for each user.
The optimal performance search algorithm sets a time search window having a preset time range, and a plurality of time ranges generated by applying the time search window at predetermined threshold time intervals to the measurement time of each sub-band signal. A brain-computer interface system that performs performance evaluation. The method of claim 9,
The basic frequency band is a mu band (Mu-band), ERD (Event-Related Desynchronization) of the frequency band corresponding to the characteristics or ERS (Event-Related Synchronization) of the frequency band corresponding to at least one of the frequency bands Brain-computer interface system. The method of claim 9,
The processor,
The brain-computer interface system is to generate n sub-band signals by dividing the EEG signal having the basic frequency band into random or predetermined rules at a frequency interval within 3 to 5 Hz. The method of claim 9,
The processor,
A brain-computer interface system that extracts EEG features for EEG signals for each sub-band signal using a feature extraction algorithm using a common spatial pattern (CSP) or frequency intensity. The method of claim 9,
The processor,
The performance index is calculated by calculating the accuracy of the classification result of the trained classification model using any one of a cross validation method, a mutual information calculation method, or a Fisher Score method. The brain-computer interface system. The method of claim 9,
The optimal performance search algorithm,
Set a time search window having a preset time range (T _d ),
A plurality of time ranges for subdividing the measurement time (T _total, T _total > T _d > T _t ) are generated by applying the time search window at a predetermined threshold time interval (T _t ) for each subband signal,
EEG characteristics are respectively extracted from a plurality of time ranges generated for each of the sub-band signals,
After generating a classifier learning the motion imaginary classes using the extracted EEG feature, calculating the accuracy of the classification result of the classifier to calculate a performance index,
A brain-computer interface system that detects an optimal subband signal having the highest performance and a maximum performance time range based on the performance indicator. The method of claim 14,
The optimal performance search algorithm,
A brain-computer interface system that removes information on a sub-band signal having the lowest performance and a time range based on the performance index.

Images