System and method for device control using a brain-computer interface

Abstract

The present invention relates to a device control system and method using a brain-computer interface, and more specifically, the system is characterized by comprising: a brainwave measurement unit that measures a user's brainwave data while a stimulus is presented; a brainwave analysis unit that presents a stimulus to acquire the measured brainwave data, analyzes the measured brainwave data to determine the user's intention, and controls a user terminal based on the user's intention; and a cloud server that stores the brainwave data measured by the brainwave measurement unit and the results of analyzing the brainwave data by the brainwave analysis unit.

Description

Technology Field

[0001] The present invention relates to a device control system and method using a brain-computer interface, and more specifically, to a system and method that determines a user's intention by analyzing brainwaves through a P300-based DSP algorithm and an SVM classifier, and controls a device based on said user intention. Background Technology

[0002] Device control using brain-computer interfaces (BCI) is evolving in various forms, and in particular, systems utilizing P300 signals—responses to rare stimuli observed in brainwaves measured via electroencephalography (EEG)—are being actively researched.

[0003] Existing P300-based BCI systems are generally designed to use fixed displays to present stimuli, but fixed displays pose a problem in that they restrict users from using the BCI while moving or in various environments.

[0004] Some recent studies have developed BCI systems that present stimuli using Virtual Reality (VR) or Augmented Reality (AR) equipment, but they have the disadvantage of requiring additional VR or AR equipment.

[0005] Meanwhile, the penetration rate of smartphones in Korea has reached 95%, and globally, it has recorded a penetration rate of over 77%, so it can be said that the use of smartphones has become commonplace.

[0006] To address the aforementioned problems, stimulus presentation and BCI operation using smartphones are proposed as a more practical and accessible solution; however, BCI operation using smartphones has the drawback of the complex usage environment of the BCI and the presence of internal and external noise.

[0007] Therefore, BCI systems require adaptive updates to operate stably even amidst diverse environments and internal and external noise; consequently, there is a need for research and development of BCI systems that can enhance their adoption and practical usability.

[0008] The matters described above as background technology are intended only to enhance understanding of the background of the present invention and should not be construed as an acknowledgment that they constitute prior art already known to those skilled in the art. Prior art literature

[0009] Korean Patent Publication No. 10-2537689 (May 24, 2023) Korean Patent Publication No. 10-2646182 (March 6, 2024) Korean Patent Publication No. 10-2446499 (September 20, 2022) The problem to be solved

[0010] Accordingly, the present invention is proposed to resolve the aforementioned conventional problems and aims to provide a system and method that determines a user's intention by analyzing brainwaves through a P300-based DSP algorithm and an SVM classifier, and controls a device based on said user intention.

[0011] Furthermore, the present invention is proposed to resolve the aforementioned conventional problems and aims to provide a system and method for presenting a stimulus to acquire brainwave data using a user terminal or device such as a smartphone, and for determining user intent by analyzing the acquired brainwave data.

[0012] Furthermore, the present invention is proposed to resolve the aforementioned conventional problems and aims to provide a system and method that adaptively updates feature extraction parameters and classifiers for brainwave data according to the user's situation even in complex usage environments, and improves the accuracy of brainwave data classification and user identification through said updates. means of solving the problem

[0013] A device control system using a brain-computer interface for achieving the above-mentioned purpose is characterized by comprising: a brainwave measurement unit (100) that measures the user's brainwave data while a stimulus is presented; a brainwave analysis unit (200) that presents a stimulus to acquire the measured brainwave data, analyzes the measured brainwave data to determine the user's intention, and controls the user terminal based on the user's intention; and a cloud server (300) that stores the brainwave data measured by the brainwave measurement unit (100) and the results of analyzing the brainwave data by the brainwave analysis unit (200).

[0014] In one embodiment of the present invention, the brainwave analysis unit (200) comprises: a stimulus presentation unit (210) that presents a stimulus to the user according to a preset initial setting value; an initial learning unit (220) that analyzes the brainwave data measured while the stimulus is presented to extract a feature extraction parameter and learns the extracted feature extraction parameter and a classifier; a user intention determination unit (230) that analyzes the brainwave data measured while the stimulus is presented using the feature extraction parameter and classifier learned in the initial learning unit (220) to determine the user's intention and controls the device based on the user's intention; and a server communication unit (240) that transmits the brainwave data, the feature extraction parameter, and the classifier parameter received from the brainwave measurement unit (100) to the cloud server (300) through communication with the cloud server (300).

[0015] Herein, the above-mentioned preset initial setting value is characterized by including at least one of the stimulus type, stimulus presentation time, stimulus non-presentation time, and stimulus repetition count, and the stimulus is characterized by being an image having at least one shape, and is characterized by being matched with function control command data of a device to the one image.

[0016] In one embodiment of the present invention, the initial learning unit (220) comprises: a first data preprocessing module (221) that performs preprocessing on the measured brainwave data; a feature extraction learning module (222) that extracts the feature extraction parameters from the preprocessed brainwave data using a feature extraction algorithm and learns the extracted feature extraction parameters; and a classifier learning module (223) that learns the classifier using the learned feature extraction parameters.

[0017] In one embodiment of the present invention, the first data preprocessing module (221) is characterized by preprocessing the brainwave data using a frequency filter and a noise removal technique to improve the quality of the brainwave data measured while the stimulus is presented, and dividing the preprocessed brainwave data into specific sections based on the stimulus presentation time.

[0018] Here, the feature extraction algorithm is characterized by being a discriminative spatial pattern, and the feature extraction learning module (222) defines brainwave data for stimuli selected by the user as the target class and brainwave data for stimuli not selected by the user as the nontarget class, and uses the feature extraction algorithm to obtain a data distribution (S) within the same class. w ), data distribution between classes (S B ), S w -1 S B The method is characterized by extracting the eigenvector (W) and the average value for each class as the feature extraction parameters, and learning the extracted feature extraction parameters.

[0019] In one embodiment of the present invention, the feature extraction learning module (222) maximizes the value of J(W) of the following mathematical formula using a feature extraction algorithm that applies the learned feature extraction parameters, the S w -1 S B It is characterized by calculating the eigenvector (W) of the above brainwave data to extract the characteristics of the above brainwave data, wherein the characteristics of the above brainwave data are a preset number of S w -1 S B It is characterized by being the result of multiplying the eigenvector (W) of the above-mentioned divided brainwave data.

[0020]

[0021] (J(W): characteristic discrimination index, W: S w -1 S B eigenvector of, S w : Intra-class data dispersion matrix, S B : Data dispersion matrix between classes)

[0023] In one embodiment of the present invention, the user intention determination unit (230) comprises: a second data preprocessing module (231) that performs preprocessing on the measured brainwave data; a user intention determination module (232) that determines the user intention by analyzing the brainwave data measured while the stimulus is presented using the feature extraction parameters and classifier learned in the initial learning unit (220); a device control module (233) that controls the device based on the user intention; and an update module (234) that performs updates on the feature extraction parameters and classifier based on the user intention determination result.

[0024] In one embodiment of the present invention, the user intention determination module (232) calculates the SVM score value of the brainwave data measured at the time of presentation of the preprocessed stimulus using the following mathematical formula, using the feature extraction parameter and classifier learned in the initial learning unit (220), and determines the stimulus corresponding to the brainwave data with the largest average value of the SVM score as the stimulus selected by the user.

[0026]

[0027] (N c : Number of times stimulus c is presented, f : Classifier function that calculates the SVM score, V n,c : Brainwave characteristic data for the nth presented stimulus c)

[0029] In one embodiment of the present invention, the device control module (233) is characterized by controlling the function of the device based on function control command data of the device that matches the stimulus determined by the user intention determination module (232) as the stimulus selected by the user.

[0030] In one embodiment of the present invention, the update module (234) is characterized by updating the feature extraction parameter and classifier parameter when the degree of separation between the SVM score value of the brainwave data for the stimulus determined to be the stimulus selected by the user and the SVM score value of the brainwave data not selected by the user is greater than or equal to a first reference value, and transmitting the updated feature extraction parameter and classifier parameter to the cloud server (300).

[0031] In one embodiment of the present invention, the update module (234) is characterized by retraining the feature extraction parameter and classifier if the number of times the update has not been performed is greater than or equal to a preset number.

[0032] The method for controlling a device using a brain-computer interface according to the present invention comprises: a brainwave measurement step of measuring a user's brainwave data while a stimulus is presented at a brainwave measurement unit (100); a brainwave analysis step of presenting a stimulus to acquire the measured brainwave data at a brainwave analysis unit (200), analyzing the measured brainwave data to determine the user's intention, and controlling the device based on the user's intention; and a storage step of storing the brainwave data measured at the brainwave measurement unit (100) and the results of analyzing the brainwave data at the brainwave analysis unit (200) in a cloud server (300).

[0033] In one embodiment of the present invention, the brainwave analysis step comprises: a stimulus presentation step in which a stimulus is presented to the user according to an initial setting value set in a stimulus presentation unit (210); an initial learning step in which a brainwave data measured while the stimulus is presented is analyzed in an initial learning unit (220) to extract a feature extraction parameter and learn the extracted feature extraction parameter and a classifier; a user intention determination step in which a user intention determination unit (230) analyzes the brainwave data measured while the stimulus is presented using the feature extraction parameter and classifier learned in the initial learning step to determine the user intention and controls the device based on the user intention; and a step in which the brainwave data received from the brainwave measurement unit (100), the feature extraction parameter, and the classifier parameter are transmitted to the cloud server (300) through communication with the cloud server (300) in a server communication unit (240).

[0034] In one embodiment of the present invention, the initial learning step comprises: a first data preprocessing step in which preprocessing is performed on the measured brainwave data in a first data preprocessing module (221); a feature extraction learning step in which a feature extraction learning module (222) extracts a feature extraction parameter from the preprocessed brainwave data using a feature extraction algorithm and learns the extracted feature extraction parameter; and a classifier learning step in which a classifier is learned using the learned feature extraction parameter in a classifier learning module (223).

[0035] In one embodiment of the present invention, the user intention determination step comprises: a second data preprocessing step in which preprocessing is performed on the measured brainwave data in a second data preprocessing module (231); a user intention determination step in which the user intention is determined by analyzing the brainwave data measured while the stimulus is presented using the feature extraction parameters and classifier learned in the initial learning unit (220) in a user intention determination module (232); a device control step in which the device is controlled based on the user intention in a device control module (233); and an update step in which the feature extraction parameters and classifier are updated based on the user intention determination result in an update module (234). Effects of the invention

[0036] According to the device control system and method using a brain-computer interface as described above, the following effects are achieved.

[0037] First, the system according to the present invention has the effect of solving the problem of presenting stimuli in a conventional fixed environment by presenting stimuli to a user using a mobile user terminal such as a smartphone.

[0038] Second, the system (1) and method according to the present invention have the effect of increasing the accuracy of user intent determination of the feature extraction algorithm by adaptively updating parameters for extracting features of brainwave data.

[0039] Third, the system (1) and method according to the present invention have the effect of increasing the accuracy of classification and user intent determination of the classifier by adaptively updating parameters for characteristic classification of brainwave data.

[0040] Fourth, the system (1) and method according to the present invention can implement a BCI system without additional equipment by using a smartphone to present stimuli and process brainwaves, thereby greatly improving user accessibility.

[0041] Fifth, the system (1) and method according to the present invention have the effect of maintaining high performance in various usage environments by adjusting parameters and updating the classifier according to the user's situation.

[0042] Sixth, the system and method according to the present invention have the effect of enabling stable device control by overcoming internal and external noise occurring in complex and diverse environments through the use of an adaptive update mechanism.

[0043] The effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the present invention belongs from the description below. Brief explanation of the drawing

[0044] FIG. 1 is a configuration diagram of a device control system using a brain-computer interface according to an embodiment of the present invention. FIG. 2 is a data flow diagram of a device control system using a brain-computer interface according to an embodiment of the present invention. FIG. 3 is a configuration diagram of a brainwave analysis unit (200) according to an embodiment of the present invention. FIG. 4 is a configuration diagram of an initial learning unit (220) according to an embodiment of the present invention. FIG. 5 is a configuration diagram of a user intention determination unit (230) according to an embodiment of the present invention. FIG. 6 is an exemplary diagram of a brainwave analysis unit (200) according to an embodiment of the present invention. FIG. 7 is a flowchart of a welfare-type demand-responsive public transportation operation method reflecting the characteristics of rural users according to an embodiment of the present invention. FIG. 8 is a flowchart of the brainwave analysis step according to an embodiment of the present invention. Figure 9 is an experimental result of a system according to an embodiment of the present invention. Specific details for implementing the invention

[0045] With reference to the attached drawings, a welfare-type demand-responsive public transportation operating system (1) and method reflecting the characteristics of rural users according to embodiments of the present invention will be described in detail. Since the present invention is subject to various modifications and may take various forms, specific embodiments are illustrated in the drawings and described in detail in the text. However, this is not intended to limit the present invention to specific disclosed forms, and it should be understood that it includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the present invention. Similar reference numerals have been used for similar components in the description of each drawing. In the attached drawings, the dimensions of the structures are shown enlarged from the actual size to ensure clarity of the present invention, or reduced from the actual size to understand the schematic configuration.

[0046] Additionally, terms such as "first," "second," etc., may be used to describe various components, but said components should not be limited by said terms. These terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. Meanwhile, unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0048] The present invention relates to a device control system and method using a brain-computer interface, and more specifically, to a system and method that determines a user's intention by analyzing brainwaves through a P300-based DSP algorithm and an SVM classifier, and controls a device based on said user intention.

[0049] Accordingly, the present invention is proposed to resolve the aforementioned conventional problems and aims to provide a system and method that determines a user's intention by analyzing brainwaves through a P300-based DSP algorithm and an SVM classifier, and controls a device based on said user intention.

[0050] Furthermore, the present invention is proposed to resolve the aforementioned conventional problems and aims to provide a system and method for presenting a stimulus to acquire brainwave data using a user terminal or device such as a smartphone, and for determining user intent by analyzing the acquired brainwave data.

[0051] Furthermore, the present invention is proposed to resolve the aforementioned conventional problems and aims to provide a system and method that adaptively updates feature extraction parameters and classifiers for brainwave data according to the user's situation even in complex usage environments, and improves the accuracy of brainwave data classification and user identification through said updates.

[0052] A Brain-Computer Interface (BCI) is a technology that controls computers or external devices using human brainwaves, executing commands by interpreting brain signals in real time. It non-invasively measures brainwaves based on EEG (Electroencephalogram) data and is utilized in various fields such as medicine, rehabilitation, and gaming. It is an innovative interface that controls devices or transmits information by interpreting the user's intentions as neural signals.

[0053] In addition, the system according to the present invention is characterized by controlling a device by determining the user's intention after measuring and analyzing a P300 signal, which represents a response to a stimulus among the user's brain signals. Here, the P300 signal is a brainwave response that appears at a point of approximately 300ms after a stimulus is presented, and occurs in response to a stimulus of interest to the user. It mainly appears as a response to rare or important stimuli and is frequently used to determine selection intentions in brain-computer interfaces. The P300 is measured based on EEG data and is classified, in particular, as one of the Event-Related Potentials (ERPs).

[0054] Accordingly, the brainwave data analyzed by the brainwave analysis unit (200) according to one embodiment of the present invention is characterized as being P300 signal data.

[0056] FIG. 1 is a configuration diagram of a device control system using a brain-computer interface according to an embodiment of the present invention, and FIG. 2 is a data flow diagram of a device control system using a brain-computer interface according to an embodiment of the present invention.

[0057] As illustrated in FIG. 1, the system according to the present invention is characterized by including a brainwave measurement unit (100), a brainwave analysis unit (200), and a cloud server (300). More specifically, the system according to the present invention is characterized by including a brainwave measurement unit (100) that measures a user's brainwave data while a stimulus is presented, a brainwave analysis unit (200) that presents a stimulus to acquire the measured brainwave data, analyzes the measured brainwave data to determine the user's intention, and controls a user terminal based on the user's intention, and a cloud server (300) that stores the brainwave data measured by the brainwave measurement unit (100) and the results of analyzing the brainwave data by the brainwave analysis unit (200).

[0058] As illustrated in FIG. 2, the brainwave measurement unit (100) measures brainwave data according to the stimulus provided by the stimulus presentation unit (210) of the brainwave analysis unit (200), and the measured brainwave data is transmitted to the brainwave analysis unit (200) and the cloud server (300). The brainwave analysis unit (200) analyzes the received brainwave data and transmits the user intent determination result, which is the result of the analysis of the brainwave data, to the cloud server (300). Specifically, it is characterized by extracting feature extraction parameters and classifier parameters of a feature extraction algorithm for analyzing the brainwave data and transmitting them to the cloud server (300), and also transmitting data with updated feature extraction parameters and classifier parameters to the cloud server (300). The cloud server (300) is characterized by storing all data measured and acquired from the brainwave data and data extracted, calculated, and judged by the brainwave measurement unit (100).

[0059] As described above, the cloud server (300) is characterized by including a communication unit for transmitting and receiving data with the brainwave measurement unit (100) and the brainwave analysis unit (200), and a storage unit for storing the data. More specifically, the communication unit is characterized by including a wireless communication module, wherein the wireless communication module is characterized as being Wi-Fi, Bluetooth, and beacon, but is not limited to the method described above.

[0060] In addition, the brainwave measuring unit (100) is characterized by measuring brainwaves (EEG, Electroencephalography) by attaching electrodes to the forehead or scalp of the user, and in another embodiment, it is characterized in that it is in the form of a wearable device such as an earphone, helmet, headset, headband, and glasses having electrodes inside. More preferably, it is characterized in that it has a helmet shape capable of measuring the user's electroencephalogram.

[0062] FIG. 3 is a configuration diagram of a brainwave analysis unit (200) according to an embodiment of the present invention.

[0063] As illustrated in FIG. 3, the brainwave analysis unit (200) according to the present invention is characterized by including a stimulus presentation unit (210), an initial learning unit (220), a user intention determination unit (230), and a server communication unit (240).

[0064] In one embodiment of the present invention, the brainwave analysis unit (200) is characterized as being an application (APP) installed within a mobile user terminal such as a smartphone, smartwatch, or tablet. In another embodiment, the brainwave analysis unit (200) is characterized as being implemented in the form of a website accessible to the user within the mobile user terminal. That is, the device controlled by the user's intention is characterized as being at least one of the mobile user terminal, a computer, and a home IoT device, and more preferably, the device is characterized as being a mobile user terminal. Furthermore, the home IoT device is characterized as being a home appliance capable of wirelessly connecting with the brainwave measurement unit (100), such as a TV, refrigerator, washing machine, or speaker. Specifically, the brainwave analysis unit (200) is characterized by determining the user's intention and controlling various functions of the mobile user terminal according to the user's intention. The function of the above-described mobile user terminal is characterized by having at least one of a phone sending / receiving function, a text message sending / receiving function, a power ON / OFF function, a music ON / OFF function, and a camera ON / OFF function, and is characterized by including another function within the mobile user terminal in addition to the functions described above.

[0065] The stimulus presenting unit (210) according to the present invention is characterized by presenting a stimulus to the user according to a preset initial setting value.

[0066] More specifically, the stimulus presentation unit (210) presents a stimulus to the user according to the preset initial setting value as shown in FIG. 6, and the brainwave measurement unit (100) measures brainwave data for the stimulus presented by the stimulus presentation unit (210).

[0067] Here, the aforementioned preset initial setting value is characterized by including at least one of the stimulus type, stimulus presentation time, stimulus non-presentation time, and stimulus repetition count. The stimulus is characterized by being an image having at least one shape, and is characterized by being matched with the function of a single device to the single image.

[0068] Preferably, in the above preset initial setting value, the type of stimulus is 4, the stimulus presentation time is 75ms, the stimulus non-presentation time is 75ms, and the number of stimulus repetitions is 10.

[0069] The above stimulus is characterized as being an image such as an icon, and the icon is characterized as being in the form of a geometric shape such as a circle, triangle, or square, or in various other shapes, and the shape of the icon is characterized as being modifiable and set in advance.

[0070] For example, the stimulus presentation unit (210) presents a total of four stimuli to the user, including a circular icon, a triangle icon, a square icon, and a star-shaped icon, and each icon appears for 75ms and disappears for 75ms, and each icon is presented repeatedly a total of 10 times. At this time, the color of each stimulus is characterized by being different during the stimulus presentation time and during the stimulus non-presentation time.

[0071] Furthermore, the user is characterized by staring intently at a stimulus that matches the stimulus the user wishes to select or the function of the device the user wishes to control. Here, the function of the device is characterized by being pre-configurable by the user. Specifically, if the device is a smartphone, the function matching the circular icon is the phone call sending / receiving function, the function matching the triangle icon is the text message sending / receiving function, the function matching the square icon is the music ON / OFF function, and the function matching the star-shaped icon is the camera ON / OFF function. For example, if the function the user wishes to control is the music ON / OFF function, the user is characterized by staring intently whenever the square icon is presented among the four presented stimuli.

[0072] Accordingly, the user's intention is characterized as a function of the device that matches the stimulus the user wishes to select.

[0073] The initial learning unit (220) according to the present invention is characterized by analyzing the brainwave data measured while the stimulus is presented to extract feature extraction parameters and learning the extracted feature extraction parameters and classifier, and the user intention determination unit (230) is characterized by analyzing the brainwave data measured while the stimulus is presented using the feature extraction parameters and classifier learned in the initial learning unit (220) to determine the user intention and controlling the device based on the user intention, and the server communication unit (240) is characterized by transmitting the brainwave data, the feature extraction parameters, and the classifier parameters received from the brainwave measurement unit (100) to the cloud server (300) through communication with the cloud server (300).

[0075] FIG. 4 is a configuration diagram of an initial learning unit (220) according to an embodiment of the present invention.

[0076] As illustrated in FIG. 4, the initial learning unit (220) according to the present invention is characterized by including a first data preprocessing module (221), a feature extraction learning module (222), and a classifier learning module (223). More specifically, the initial learning unit (220) is characterized by including a first data preprocessing module (221) that performs preprocessing on the measured brainwave data, a feature extraction learning module (222) that extracts feature extraction parameters using a feature extraction algorithm from the preprocessed brainwave data and learns the extracted feature extraction parameters, and a classifier learning module (223) that learns the classifier using the learned feature extraction parameters.

[0077] The first data preprocessing module according to the present invention is characterized by receiving brainwave data measured by the brainwave measurement unit (100) while the stimulus is being presented by the stimulus presentation unit (210) and performing a preprocessing process on the received brainwave data. More specifically, the first data preprocessing module (221) is characterized by preprocessing the brainwave data using a frequency filter and a noise removal technique to improve the quality of the brainwave data measured while the stimulus is being presented, and dividing the preprocessed brainwave data into specific sections based on the stimulus presentation time.

[0078] Specifically, the stimulus presentation unit (210) is characterized by presenting a total of 40 stimuli to the user according to a preset initial setting value. The first data preprocessing module (221) is characterized by performing a preprocessing process on the brainwave data using at least one of the Wavelet Neural Network (WNN) technique, which is a data filtering method using a frequency filter for noise removal of the acquired brainwave data, and the Artifact Subspace Reconstruction (ASR) technique, which is a noise removal technique.

[0079] The first data preprocessing module (221) according to the present invention is characterized by completing a preprocessing process on the brainwave data and then dividing the preprocessed brainwave data into specific intervals based on the stimulus presentation time. Specifically, the specific interval is characterized as being from 200ms before the stimulus presentation time to 600ms after the stimulus presentation time, and this is characterized as a interval setting for obtaining the P300 signal, since the P300 signal is a brainwave signal that appears at the 300ms point after the stimulus is presented.

[0080] In addition, the first data preprocessing module (221) is characterized by performing a time point synchronization process before segmenting the preprocessed brainwave data. Synchronization between the stimulus presentation time and the brainwave measurement time is essential to accurately classify the stimulus selected by the user. To synchronize the stimulus presentation time and the brainwave measurement time, the standard time (UTC) of the brainwave data measurement start time and the standard time (UTC) of the stimulus presentation time are recorded, and the brainwave data corresponding to the stimulus presentation time is synchronized based on the recorded standard time. Through such time point synchronization, brainwave data for the stimulus selected by the user can be accurately measured and extracted.

[0081] The feature extraction learning module (222) according to the present invention is characterized by extracting feature extraction parameters from the preprocessed brainwave data using a feature extraction algorithm and learning the extracted feature extraction parameters. Here, the feature extraction algorithm is characterized as being a discriminative spatial pattern (DSP).

[0082] The above feature extraction learning module (222) is characterized by defining brainwave data for stimuli selected by the user as the target class and brainwave data for stimuli not selected by the user as the nontarget class, and using the above feature extraction algorithm, the data distribution (S) within the same class w ), data distribution between classes (S b ), S w -1 S b The method is characterized by extracting the eigenvector (W) and the average value for each class as the feature extraction parameters, and training the extracted feature extraction parameters.

[0083] To explain in detail, the feature extraction algorithm is an algorithm that extracts feature values ​​from segmented brainwave data after undergoing the preprocessing process, and the feature extraction module is characterized by training feature extraction parameters for extracting the brainwave data feature values ​​in the feature extraction algorithm, and applying the trained feature extraction parameters to the feature extraction algorithm to extract the brainwave data feature values.

[0084] The above feature extraction parameters extracted at this time are the data distribution within the same class (S w ), data distribution between classes (S b ), S w -1 S b It is characterized by the eigenvector (W) and the average value for each class.

[0085] The above data distribution within the same class (S w ) is a Within-class scatter matrix and refers to the data distribution or data dispersion matrix within the same class, and is characterized by being calculated through the following mathematical formula (1).

[0087] Mathematical formula (1)

[0088] (S w : Data distribution within the same class, S j : Variance of class j)

[0090] Here, the variance of class j (S j ) can be calculated using the following mathematical formula (2).

[0092] Mathematical formula (2)

[0093] (X j (i) : i-th data of class j, : Average of class j, n j : Number of data items in class j)

[0095] Data distribution between the above classes (S b ) is a between scatter matrix, signifying the difference in center between classes or the data dispersion between classes, and is characterized by indicating the clarity of class distinction. The above data distribution between classes (S b ) is characterized by being calculated through the following mathematical formula (3).

[0097] Mathematical formula (3)

[0098] (X j (i) : i-th data of class j, : Average of class j, n j : Number of data items in class j)

[0100] In addition, the feature extraction learning module (222) uses a feature extraction algorithm that applies the learned feature extraction parameters to maximize the value of J(W) of the following mathematical formula (4) S w -1 S b It is characterized by calculating the eigenvector (W) of the above brainwave data to extract the characteristics of the above brainwave data, wherein the characteristics of the above brainwave data are a preset number of S w -1 S b It is characterized by being the result of multiplying the eigenvector (W) of the above-mentioned divided brainwave data.

[0102] Mathematical formula (4)

[0103] (J(W): characteristic discrimination index, W: S w -1 S B eigenvector of, S w : Intra-class data dispersion matrix, S B : Data dispersion matrix between classes)

[0105] At this time, the above S w -1 S bThe eigenvector (W) is a feature extraction parameter used to extract discriminable characteristics of brainwave data, and the characteristics of the brainwave data are a set number of S w -1 S b It is characterized by being the result of multiplying the eigenvector (W) of and the segment-divided preprocessed brainwave data. More preferably, the preset number is the S w -1 S B It is characterized by having the top 2 to 5 eigenvectors (W) in order of magnitude.

[0106] The above J(W) is S w -1 S b For the eigenvector (W) of , the between-class variance (S B Maximize ) and within-class variance (S w It refers to the ratio that minimizes ). More specifically, it refers to the variance between classes (S B As the value of ) increases, the distinction between classes becomes clearer; therefore, the eigenvector (W) is set in a direction that maximizes this, and the variance within the class (S) w Since a smaller value indicates that the data inside the class is more densely packed, the above eigenvector (W) is set in a direction that minimizes this.

[0107] It is characterized by maximizing the distinction between clags by maximizing the value of J(W) through the above mathematical formula (4), and thereby extracting a feature value with high discriminative power from the above preprocessed brainwave data.

[0109] The classifier learning module (223) according to the present invention is characterized by training the classifier using the learned feature extraction parameters, and the classifier is characterized as being a Support Vector Machine (SVM).

[0111] FIG. 5 is a configuration diagram of a user intention determination unit (230) according to an embodiment of the present invention.

[0112] As illustrated in FIG. 5, the user intent determination unit (230) according to the present invention is characterized by including a second data preprocessing module (231), a user intent determination module (232), a device control module (233), and an update module (234).

[0113] The initial learning unit (220) and the user intent determination unit (230) undergo the same process of performing a preprocessing process on the brainwave data measured by the brainwave measurement unit (100). If the initial learning unit (220) is a component that extracts the feature extraction parameters from the preprocessed brainwave data using the feature extraction algorithm and trains the feature extraction parameters and the classifier, then the user intent determination unit (230) is a component that determines the user intent from the preprocessed brainwave data using the feature extraction parameters and classifier trained by the initial learning unit (220), and controls the device according to the user intent.

[0114] To explain in detail, the user intent determination unit (230) is characterized by including a second data preprocessing module (231) that performs preprocessing on the measured brainwave data, a user intent determination module (232) that determines the user intent by analyzing the brainwave data measured while the stimulus is presented using feature extraction parameters and a classifier learned in the initial learning unit (220), a device control module (233) that controls the device based on the user intent, and an update module (234) that performs updates on the feature extraction parameters and classifier based on the user intent determination result.

[0115] The user intent determination unit (230) determines the user intent using the feature extraction parameters and classifier learned in the initial learning unit (220), which indicates that the process in the user intent determination unit (230) proceeds after the process in the initial learning unit (220) is performed. The second data preprocessing module (231) according to the present invention is characterized as a component that performs the same process as the first data preprocessing module (221) described above.

[0116] Therefore, it can be seen that the first data preprocessing module (221) and the second data preprocessing module (231) preprocess brainwave data obtained from the brainwave measurement unit (100), and only the process of segmenting after time synchronization is the same, but the obtained brainwave data is not the same.

[0118] The user intention determination module (232) according to the present invention is characterized by determining the stimulus corresponding to the preprocessed brainwave data with the largest average SVM score value as the stimulus selected by the user using the feature extraction parameters and classifier learned in the initial learning unit (220), and the average SVM score value is calculated using the following mathematical formula (5).

[0120] Mathematical formula (5)

[0121] (N c : Number of times stimulus c is presented, f : Classifier function that calculates the SVM score, V n,c : Brainwave characteristic data for the nth presented stimulus c)

[0123] More specifically, the user intent determination module (232) according to the present invention calculates an SVM score value, which is the result of classifying the preprocessed brainwave data using a classifier to which the learned feature extraction parameter is applied, determines the stimulus corresponding to the brainwave data in which the average value of the SVM score appears high as the stimulus selected by the user, and determines the function of the device matched to the stimulus selected by the user as the user's intent.

[0125] The device control module (233) according to the present invention is characterized by controlling the function of the device based on the function of the device that matches the stimulus determined by the user intention determination module (232) as the stimulus selected by the user.

[0126] Specifically, when the device is a smartphone, the function matching the circular icon is the phone sending / receiving function, the function matching the triangle icon is the text sending / receiving function, the function matching the square icon is the music ON / OFF function, and the function matching the star-shaped icon is the camera ON / OFF function, the device control module (233) controls the text sending / receiving function, which is the function matching the triangle icon, when the stimulus selected by the user determined by the user intent determination unit (230) is the triangle icon.

[0128] The update module (234) according to the present invention is characterized by updating the feature extraction parameter and classifier parameter when the degree of separation between the SVM score value of the brainwave data for the stimulus determined to be the stimulus selected by the user and the SVM score value of the brainwave data not selected by the user is greater than or equal to a first reference value, and transmitting the updated feature extraction parameter and classifier parameter to the cloud server (300).

[0129] And the update module (234) is characterized by not updating the feature extraction parameter and classifier parameter when the degree of separation between the SVM score value of the brainwave data for the stimulus determined to be the stimulus selected by the user and the SVM score value of the brainwave data not selected by the user is less than the first reference value.

[0130] The degree of separation between the SVM score value of brainwave data for a stimulus determined to be selected by the user and the SVM score value of brainwave data for a stimulus not selected by the user may be the difference in eigenvector (W) values ​​for the stimulus selected by the user and the stimulus not selected by the user, and in another example, it is characterized by being the difference in the average SVM scorer value for the stimulus selected by the user and the stimulus not selected by the user.

[0131] Here, preferably, the first reference value is preferably at least one of 0.5, 1.0, or 1.5, and the preset number of updates not performed is 5 to 10 times, and more preferably 5 times.

[0132] Additionally, the update module (234) is characterized by retraining the feature extraction parameter and classifier if the number of times the update has not been performed is greater than or equal to a preset number.

[0133] As illustrated in FIG. 8, when the number of stimulus presentations and processing cycles is denoted as N, the update module (234) is characterized by performing an update if the degree of separation is greater than or equal to a first reference value. Conversely, if the degree of separation is less than the first reference value, no update is performed, and the value of N is increased by 1. At this time, if the value of N exceeds a preset number of cycles, the feature extraction parameter and classifier are retrained using data stored in the cloud server (300).

[0134] To explain in more detail, the update of the feature extraction parameter is characterized by being performed by updating the eigenvector (W) of the feature extraction algorithm DSP. The update of the eigenvector (W) is characterized by being performed using a recursive least square type algorithm for Generalized eigendecomposition, and the updated eigenvector (W) is characterized by being calculated through the following mathematical formula (6).

[0136] Mathematical formula (6)

[0137] (S W (n): S calculated using new brainwave data W, S B (n): S calculated using new brainwave data B )

[0139] In addition, S in the above mathematical formula (6) W (n) and S B (n) Characterized by using the above mathematical formula (2) and mathematical formula (3) to obtain the value, and the above S W (n) and S B (n) The average value used for calculation is characterized by being calculated using the following mathematical formula (7). That is, the average value calculated using the following mathematical formula (7) is characterized as being an updated average value.

[0141] Mathematical formula (7)

[0142] ( : Average value of existing class j data, : Update weight, N : Number of data points)

[0144] In the above mathematical formula (7) The average value of the data of the existing class j is characterized as being, more specifically, a value calculated by the feature extraction module of the initial learning unit (220) prior to the update. The update weight is characterized as being set by the user, but more preferably 0.2.

[0145] The above classifier update is characterized by being performed by adding a support vector to the above SVM. More specifically, if the SVM score value, which is the result of the classifier based on newly input EEG data, is 1 or less, an update is performed using the corresponding data as a new support vector, and at this time, the new support vector is identical to the nearest existing support vector It is characterized by assigning a value. At this time, the SVM score value, which is the result of the classifier based on the newly input brainwave data, is calculated using the following mathematical formula (8).

[0147] Mathematical formula (8)

[0148] (y(x) : SVM score value, tm : label of the m-th data, K : kernel function, b : bias, : A value indicating whether the m-th data point is determined to be a support vector)

[0150] Here, the newly input brainwave data refers to brainwave data additionally acquired whenever the stimulus presented by the stimulus presentation unit (210) is repeated. The system and method according to the present invention are characterized by acquiring brainwave data for repeated stimuli and adaptively learning and updating the feature extraction parameters and classifiers to more accurately determine the user's intention and perform stable device control.

[0152] The cloud server (300) according to the present invention is characterized by storing at least one of the brainwave data, the extracted feature extraction parameter, the classifier parameter, the updated feature extraction parameter, the updated classifier parameter, the brainwave data feature value, the user intention, and the function of the device matched to the stimulus.

[0153] FIG. 7 is a flowchart of a welfare-type demand-responsive public transportation operation method reflecting the characteristics of rural users according to an embodiment of the present invention.

[0154] The method for controlling a device using a brain-computer interface according to the present invention is characterized by comprising: a brainwave measurement step of measuring a user's brainwave data while a stimulus is presented at a brainwave measurement unit (100); a brainwave analysis step of presenting a stimulus to acquire the measured brainwave data at a brainwave analysis unit (200), analyzing the measured brainwave data to determine the user's intention, and controlling the device based on the user's intention; and a storage step of storing the brainwave data measured at the brainwave measurement unit (100) and the results of analyzing the brainwave data at the brainwave analysis unit (200) in a cloud server (300).

[0155] In one embodiment of the present invention, the brainwave analysis step comprises: a stimulus presentation step in which a stimulus is presented to the user according to an initial setting value set in the stimulus presentation unit (210); an initial learning step in which the brainwave data measured while the stimulus is presented is analyzed in the initial learning unit (220) to extract feature extraction parameters and learn the extracted feature extraction parameters and a classifier; a user intention determination step in which the brainwave data measured while the stimulus is presented is analyzed in the user intention determination unit (230) using the feature extraction parameters and classifier learned in the initial learning step to determine the user intention and control the device based on the user intention; and a step in which the brainwave data received from the brainwave measurement unit (100), the feature extraction parameters, and the classifier parameters are transmitted to the cloud server (300) through communication with the cloud server (300) in the server communication unit (240).

[0156] The above-mentioned preset initial setting value is characterized by including at least one of the stimulus type, stimulus presentation time, stimulus non-presentation time, and stimulus repetition count, and the stimulus is characterized by being an image having at least one different form, and the single image is characterized by being matched with the function of a single device.

[0157] In one embodiment of the present invention, the initial learning step is characterized by comprising: a first data preprocessing step in which preprocessing is performed on the measured brainwave data in a first data preprocessing module (221); a feature extraction learning step in which a feature extraction learning module (222) extracts the feature extraction parameters from the preprocessed brainwave data using a feature extraction algorithm and learns the extracted feature extraction parameters; and a classifier learning step in which a classifier is learned using the learned feature extraction parameters in a classifier learning module (223).

[0158] The first data preprocessing step according to the present invention is characterized by preprocessing the brainwave data using a frequency filter and a noise removal technique to improve the quality of the brainwave data measured while the stimulus is presented in the first data preprocessing module (221), and dividing the preprocessed brainwave data into specific sections based on the stimulus presentation time.

[0159] The above feature extraction algorithm is characterized by being a discriminative spatial pattern.

[0160] The feature extraction learning step according to the present invention defines brainwave data for a stimulus selected by the user in the feature extraction learning module (222) as the target class and brainwave data for a stimulus not selected by the user as the nontarget class, and uses the feature extraction algorithm to determine the data distribution (S) within the same class. w ), data distribution between classes (S b ), S w -1 S b The method is characterized by extracting the eigenvector (W) and the average value for each class as the feature extraction parameters, and training the extracted feature extraction parameters.

[0161] The above feature extraction learning step utilizes a feature extraction algorithm applying the learned feature extraction parameters to maximize the value of J(W) in the following mathematical formula, the above S w -1 S B It is characterized by calculating the eigenvector (W) of the above brainwave data to extract the characteristics of the above brainwave data, wherein the characteristics of the above brainwave data are a preset number of S w -1 S B It is characterized by being the result of multiplying the eigenvector (W) of the above-mentioned divided brainwave data.

[0163]

[0164] (J(W): characteristic discrimination index, W: S w -1 S B eigenvector of, S w : Intra-class data dispersion matrix, S B : Data dispersion matrix between classes)

[0166] In one embodiment of the present invention, the user intention determination step comprises: a second data preprocessing step in which preprocessing is performed on the measured brainwave data in a second data preprocessing module (231); a user intention determination step in which the user intention is determined by analyzing the brainwave data measured while the stimulus is presented using the feature extraction parameters and classifier learned in the initial learning unit (220) in a user intention determination module (232); a device control step in which the device is controlled based on the user intention in a device control module (233); and an update step in which the feature extraction parameters and classifier are updated based on the user intention determination result in an update module (234).

[0167] The user intent determination step according to the present invention is characterized in that the user intent determination module (232) determines the stimulus corresponding to the preprocessed brainwave data with the largest average SVM score value as the stimulus selected by the user by using the feature extraction parameters and classifier learned in the initial learning unit (220), and the average SVM score value is calculated using the following mathematical formula.

[0169]

[0170] (N c : Number of times stimulus c is presented, f : Classifier function that calculates the SVM score, V n,c : Brainwave characteristic data for the nth presented stimulus c)

[0172] The device control step according to the present invention is characterized by controlling the function of the device based on function control command data of the device that matches the stimulus determined by the user intention determination module (232) as the stimulus selected by the user in the device control module (233).

[0173] The update step according to the present invention is characterized by updating the feature extraction parameter and classifier parameter when the degree of separation between the SVM score value of the brainwave data for the stimulus determined to be the stimulus selected by the user in the update module (234) and the SVM score value of the brainwave data not selected by the user is greater than or equal to a first reference value, and transmitting the updated feature extraction parameter and classifier parameter to the cloud server (300).

[0174] The above update step is characterized by retraining the feature extraction parameters and classifier if the number of times the above update has not been performed is greater than or equal to a preset number.

[0176] Figure 9 is an experimental result of a system according to an embodiment of the present invention.

[0177] Figure 9 is a graph showing the results of an experiment in which a device control system using a brain-computer interface according to the present invention was applied to a user.

[0178] The above experiment is characterized by satisfying the following experimental conditions.

[0180] (1) The subject was presented with a total of 160 stimuli in 5 sets, with 4 stimuli repeated 5 times as 1 set.

[0181] (2) Device control is performed in accordance with the following three conditions regarding the above stimulus.

[0182] - Speak Performance Condition: The subject controls the device while speaking freely on a given topic for 20 seconds.

[0183] - MM (Mouth Movement) Performance Condition: The subject moves their mouth according to a given mouth shape while simultaneously controlling the device and performing the given task.

[0184] - None Performance Condition: The subject controls the device without a separate task while moving their mouth according to a given mouth shape.

[0185] (3) After training the feature extraction parameters and classifier using the 160 samples obtained in step (1) above, the feature extraction parameters and classifier are sequentially updated according to the execution conditions of step (2) above.

[0187] The graph in Figure 9 above compares the accuracy before and after training the feature extraction parameters and classifier, where the x-axis of the graph represents the three execution conditions, and the y-axis of the graph represents the accuracy of the system according to the present invention. In the graph, "Fixed" refers to the case where the feature extraction parameters and classifier are applied as is without updating after training, and "Adaptive" refers to the case where the feature extraction parameters and classifier are continuously updated after training under the three execution conditions.

[0188] To explain the results of the graph above in detail, it can be seen that the accuracy of the Adaptive method increased significantly compared to the Fixed method under all three conditions. The experimental result graph indicates that when only the initial learned parameters are maintained without updating, as in the Fixed method, a discrepancy with new brainwave data arises, leading to a decrease in the system's accuracy. On the other hand, as in the Adaptive method, continuously updating the initial learned parameters to match new brainwave data enables the reflection of the user's changing state and external tasks, and it can be seen that the system's accuracy also increases significantly.

[0189] In addition, observing that the increase in accuracy of the Adaptive method is the greatest under Speak execution conditions, it can be seen that the system according to the present invention is very effective even when processing multiple tasks.

[0190] Therefore, through the above experiments, it can be seen that adaptively updating parameters according to changes in the user's usage environment, as in the system and method according to the present invention, is a highly effective method, and that a device control system using a brain-computer interface that performs adaptive updates, as in the system according to the present invention, is highly effective even in situations involving external interference. Furthermore, it can be seen that the system and method according to the present invention is excellent in terms of determining user intent and improving device control accuracy by adaptively updating parameters according to changes in the user's usage environment.

[0192] According to the device control system and method using a brain-computer interface as described above, the following effects are achieved.

[0193] First, the system according to the present invention has the effect of solving the problem of presenting stimuli in a conventional fixed environment by presenting stimuli to a user using a mobile user terminal such as a smartphone.

[0194] Second, the system (1) and method according to the present invention have the effect of increasing the accuracy of user intent determination of the feature extraction algorithm by adaptively updating parameters for extracting features of brainwave data.

[0195] Third, the system (1) and method according to the present invention have the effect of increasing the accuracy of classification and user intent determination of the classifier by adaptively updating parameters for characteristic classification of brainwave data.

[0196] Fourth, the system (1) and method according to the present invention can implement a BCI system without additional equipment by using a smartphone to present stimuli and process brainwaves, thereby greatly improving user accessibility.

[0197] Fifth, the system (1) and method according to the present invention have the effect of maintaining high performance in various usage environments by adjusting parameters and updating the classifier according to the user's situation.

[0198] Sixth, the system and method according to the present invention have the effect of enabling stable device control by overcoming internal and external noise occurring in complex and diverse environments through the use of an adaptive update mechanism.

[0199] The effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the present invention belongs from the description below.

[0200] Although the technical concept of the present invention described above has been specifically described in preferred embodiments, it should be noted that the aforementioned embodiments are for illustrative purposes only and are not intended to be limiting. Furthermore, those skilled in the art will understand that various embodiments are possible within the scope of the technical concept of the present invention. Accordingly, the true scope of technical protection of the present invention should be determined by the technical concept of the appended claims. Explanation of the symbols

[0201] 1: Device control system using a brain-computer interface 100 : Brainwave measurement unit 200 : Brainwave Analysis Unit 300 : Cloud Server 210: Stimulus presentation section 220 : Initial learning section 230 : User Intent Determination Unit 240 : Server Communications Unit 221 : 1st Data Preprocessing Module 222 : Feature Extraction Training Module 223 : Classifier Training Module 231 : 2nd Data Preprocessing Module 232 : User Intent Determination Module 233 : Device control module 234 : Update Module

Claims

Claim 1 A brainwave measurement unit that measures brainwave data including the user's P300 signal while a stimulus is presented; a brainwave analysis unit that is implemented in the form of an application installed on a mobile user terminal, presents the stimulus through the user terminal, analyzes the brainwave data to determine the user's intention, and controls the user terminal based on the user's intention; and a cloud server storing the brainwave data, feature extraction parameters, and classifier parameters; wherein the brainwave analysis unit comprises: a stimulus presentation unit that repeatedly presents images as stimuli having multiple different forms according to preset initial settings, each matching one function control command data of the user terminal; an initial learning unit that, in the first brainwave data measured in response to the stimulus presented for initial learning, defines the brainwave data for the stimulus selected by the user as the target class and the brainwave data for the stimulus not selected by the user as the nontarget class, and extracts and learns the data distribution within the same class (Sw), the data distribution between classes (Sb), the eigenvector (W) of Sw^-1Sb, and the average value for each class as the feature extraction parameters using a Discriminative Spatial Pattern (DSP) algorithm, and learns the Support Vector Machine (SVM) classifier and the classifier parameters using the learned feature extraction parameters; and the learned feature extraction parameters and the A user intent determination module that applies an SVM classifier, calculates the average value of SVM score values ​​for each repeatedly presented stimulus, and determines the stimulus with the largest average value as the stimulus selected by the user; and a device control module that controls the functions of the user terminal based on function control command data that matches the stimulus selected by the user.A device control system using a brain-computer interface, characterized by comprising: an update module that, whenever new brainwave data for a repeated stimulus is acquired, calculates the degree of separation between the average value of the SVM score for the stimulus selected by the user and the average value of the SVM score for the stimulus not selected by the user, updates the feature extraction parameter and the classifier parameter if the degree of separation is greater than or equal to a first reference value, does not perform the update and increases the number of times the update is not performed if the degree of separation is less than the first reference value, and retrains the feature extraction parameter and the SVM classifier using brainwave data stored in the cloud server if the number of times the update is not performed is greater than or equal to a preset number; and a server communication unit that transmits the first brainwave data, the second brainwave data, the learned or updated feature extraction parameter, and the learned or updated classifier parameter to the cloud server. Claim 2 delete Claim 3 A device control system using a brain-computer interface according to claim 1, wherein the preset initial setting value includes at least one of a stimulus type, a stimulus presentation time, a stimulus non-presentation time, and a number of stimulus repetitions. Claim 4 A device control system using a brain-computer interface according to claim 1, wherein the initial learning unit comprises: a first data preprocessing module that performs preprocessing on the first brainwave data; a feature extraction learning module that extracts the feature extraction parameters from the preprocessed first brainwave data using the DSP algorithm and learns the extracted feature extraction parameters; and a classifier learning module that learns the SVM classifier using the learned feature extraction parameters. Claim 5 A device control system using a brain-computer interface according to claim 4, wherein the first data preprocessing module preprocesses the first brainwave data using a frequency filter and a noise removal technique to improve the quality of the first brainwave data measured while the stimulus for the initial learning is presented, and divides the preprocessed first brainwave data into specific sections based on the stimulus presentation time. Claim 6 delete Claim 7 delete Claim 8 delete Claim 9 delete Claim 10 delete Claim 11 delete Claim 12 delete Claim 13 An initial learning stimulus presentation step in which a stimulus presentation unit repeatedly presents images having multiple different forms according to preset initial settings, each matching a single function control command data of a mobile user terminal, as stimuli for initial learning; a first brainwave measurement step in which a brainwave measurement unit measures first brainwave data including the user's P300 signal while the stimulus for initial learning is being presented; an initial learning step in which an initial learning unit defines brainwave data for the stimulus selected by the user in the first brainwave data as a target class and brainwave data for the stimulus not selected by the user as a nontarget class, and extracts and learns the data distribution within the same class (Sw), the data distribution between classes (Sb), the eigenvector (W) of Sw^-1Sb, and the average value for each class as feature extraction parameters using a DSP algorithm, and learns an SVM classifier and classifier parameters using the learned feature extraction parameters; a user intention determination stimulus presentation step in which, after the initial learning is completed, the stimulus presentation unit repeatedly presents the multiple images again as stimuli for determining user intention; and the brainwave measurement unit determines the user intention A second brainwave measurement step for measuring second brainwave data different from the first brainwave data while a stimulus is presented; a user intent determination step in which a user intent determination module applies the learned feature extraction parameter and the SVM classifier to the second brainwave data, calculates the average value of the SVM score values ​​for each repeatedly presented stimulus, and determines the stimulus with the largest average value as the stimulus selected by the user; and a device control step in which a device control module controls the function of the user terminal based on the function control command data that matches the stimulus selected by the user.A method for controlling a device using a brain-computer interface, characterized by comprising: an update step in which, whenever new brainwave data for a repeated stimulus is acquired by an update module, the degree of separation between the average value of the SVM score for the stimulus selected by the user and the average value of the SVM score for the stimulus not selected by the user is calculated, and if the degree of separation is greater than or equal to a first reference value, the feature extraction parameter and the classifier parameter are updated, and if the degree of separation is less than or equal to the first reference value, the update is not performed and the number of times the update is not performed is increased, and if the number of times the update is not performed is greater than or equal to a preset number, the feature extraction parameter and the SVM classifier are retrained using brainwave data stored in a cloud server; and a storage step in which a server communication unit transmits the first brainwave data, the second brainwave data, the learned or updated feature extraction parameter and the learned or updated classifier parameter to the cloud server, and the cloud server stores the received data. Claim 14 delete Claim 15 In claim 13, the above initial learning step comprises: a first data preprocessing step in which a first data preprocessing module performs preprocessing on the first brainwave data; a feature extraction learning step in which a feature extraction learning module extracts feature extraction parameters from the preprocessed first brainwave data using the DSP algorithm and learns the extracted feature extraction parameters; and a classifier learning step in which a classifier learning module learns the SVM classifier using the learned feature extraction parameters; characterized in that the method for controlling a device using a brain-computer interface is characterized in that the above initial learning step comprises: a first data preprocessing step in which a first data preprocessing module performs preprocessing on the first brainwave data; a feature extraction learning step in which a feature extraction learning module extracts feature extraction parameters and learns the extracted feature extraction parameters; and a classifier learning step in which a classifier learning module learns the SVM classifier using the learned feature extraction parameters. Claim 16 delete

Images