KR102571116B1 - Method and device of brain-computer interface based on optimal channel selection algorithm using nirs - Google Patents

Abstract

The present invention relates to a brain-computer interface device and method based on an optimal channel selection algorithm using a near-infrared spectroscopy technique, wherein N sensors - the sensor includes a light emitting unit generating light signals corresponding to near-infrared rays with respect to the body and A data acquisition module for obtaining NIRS data for each of a plurality of brain regions of a user corresponding to a plurality of channels by using a plurality of channels formed by a pair of light receivers for detecting optical signals corresponding to near-infrared rays; and and a control module for selecting at least one channel among the plurality of channels based on the NIRS data obtained from the data acquisition module, wherein the control module extracts feature data for the NIRS data and A brain-computer interface device based on an optimal channel selection algorithm using a near-infrared spectroscopy technique for selecting at least one of the plurality of channels in consideration of the number (N) of the sensors may be provided.

Description

Brain-computer interface method and device based on optimal channel selection algorithm using near-infrared spectroscopy technique

The present application relates to a near-infrared spectroscopy based brain-computer interface method and apparatus, and more particularly, near-infrared spectroscopy, measured from a sensor attached around the prefrontal cortex while performing a mental arithmetic task. A brain-computer interface method and apparatus for selecting an optimal channel to improve the practicality of a brain-computer interface using NIRS) data.

A brain-computer interface (BCI) measures brain activity to help people with paralysis, such as amyotrophic lateral sclerosis or locked-in syndrome, drive external equipment. means technology.

Although there are various methods for measuring brain activity, such as electroencephalography (EEG), near-infrared spectroscopy is widely used due to its strong resistance to electrical noise or motion noise.

In the brain-computer interface based on near-infrared spectroscopy, the performance tends to increase as the number of channels used increases. However, as the number of channels increases, the experiment preparation time increases and the weight of the equipment increases, causing user fatigue and resulting in performance degradation. there is. Also, the inclusion of unnecessary or noisy channels can degrade performance.

In order to solve this problem, an appropriate channel selection method that can improve practicality by reducing the number of sensors used for channel formation and the calculation cost for driving the system as well as maintaining or improving performance in the brain-computer interface is required. It is necessary.

One problem to be solved is to provide a brain-computer interface method and apparatus based on an optimal channel selection algorithm using a near-infrared spectroscopy technique that can not only select an optimal channel but also reduce the number of attached sensors.

The problem to be solved in the present application is not limited to the above-described problem, and problems not mentioned will be clearly understood by those skilled in the art from this specification and the accompanying drawings. .

According to one aspect of the present application, in a brain-computer interface device based on an optimal channel selection algorithm using a near-infrared spectroscopy technique, N sensors-the sensors include a light emitting unit generating light signals corresponding to near-infrared rays with respect to the body and the body A plurality of body parts of the user corresponding to the plurality of channels using a plurality of channels formed by the plurality of body parts of the user including a brain region -including a data acquisition module for acquiring NIRS data for each and a control module for selecting at least one channel among the plurality of channels based on the NIRS data obtained from the data acquisition module, wherein the The control module extracts feature data for the NIRS data and selects an optimal channel using a near-infrared spectroscopy technique for selecting at least one channel among the plurality of channels in consideration of the extracted feature data and the number (N) of the plurality of sensors. An algorithm-based brain-computer interface device may be provided.

In one embodiment, the control module calculates classification accuracy using the extracted feature data, and selects at least one channel among the plurality of channels based on the calculated classification accuracy and the number (N) of the plurality of sensors. can choose

In one embodiment, the control module calculates a channel selection reference value based on the classification accuracy and the number (N) of the plurality of sensors, and performs sequential forward selection (SFS) on the channel selection reference value. At least one channel among the plurality of channels may be selected by application.

In one embodiment, the channel selection reference value may be determined by an algorithm based on Equations 1 and 2 below.

(Equation 1)

(Equation 2)

In one embodiment, the control module optimizes an algorithm for the channel selection reference value, and the channel selection reference value may be determined by Equation 3 below.

(Equation 3)

In one embodiment, the control module may divide the NIRS data into a plurality of time intervals, and the feature data may include a plurality of divided feature data extracted from the divided plurality of time intervals.

In one embodiment, the plurality of division feature data may be pre-processed to remove noise by applying a band pass filter.

In one embodiment, the plurality of segmented feature data may be subjected to baseline adjustment pre-processing using at least a part of the immediately preceding time interval.

In one embodiment, the NIRS data may include first NIRS data related to activation of brain activity during task performance and second NIRS data related to stabilization of brain activity after task performance.

According to another aspect of the present application, in the optimal channel selection algorithm-based brain-computer interface method using a near-infrared spectroscopy technique, N sensors—the sensors include a light emitting unit for generating light signals corresponding to near-infrared rays with respect to the body and the Forming a plurality of channels by including a light receiver for detecting an optical signal corresponding to near-infrared rays from the body, a plurality of body parts of the user corresponding to the plurality of channels, the plurality of body parts of the user including a brain region -obtaining NIRS data for each and selecting at least one channel among the plurality of channels based on the obtained NIRS data, wherein the selecting step corresponds to the NIRS data Optimal channel selection algorithm-based brain using near-infrared spectroscopy, including the step of extracting feature data for and selecting at least one channel among the plurality of channels in consideration of the extracted feature data and the number (N) of the plurality of sensors. - A computer interface method may be provided.

In one embodiment, the selecting step calculates classification accuracy using the extracted feature data, and selects at least one of the plurality of channels based on the calculated classification accuracy and the number (N) of the plurality of sensors. It may include selecting a channel.

In one embodiment, the step of selecting the plurality of channels by calculating a channel selection reference value based on the classification accuracy and the number (N) of the plurality of sensors, and applying sequential forward selection to the channel selection reference value It may include selecting at least one channel from among.

In one embodiment, the channel selection reference value may be determined by an algorithm based on Equations 1 and 2 below.

(Equation 1)

(Equation 2)

In one embodiment, the selecting step includes optimizing an algorithm for the channel selection reference value, and the channel selection reference value may be determined by Equation 3 below.

(Equation 3)

In one embodiment, the selecting may include dividing the NIRS data into a plurality of time intervals, and the feature data may include a plurality of divided feature data extracted from the divided plurality of time intervals. .

In an embodiment, the selecting may include pre-processing the plurality of division feature data to remove noise by applying a band pass filter.

In an embodiment, the selecting may include pre-processing the plurality of segmented feature data for baseline adjustment using at least a portion of an immediately preceding time interval.

In one embodiment, the NIRS data may include first NIRS data related to activation of brain activity during task performance and second NIRS data related to stabilization of brain activity after task performance.

The solutions to the problems of the present invention are not limited to the above-described solutions, and solutions not mentioned will be clearly understood by those skilled in the art from this specification and the accompanying drawings. You will be able to.

According to an embodiment of the present application, it is possible to not only select an optimal channel but also reduce the number of attached sensors, thereby improving the performance of the brain-computer interface, reducing user fatigue due to a reduction in measurement preparation time, lightening the weight of measurement equipment, etc. Through this, it is possible to improve the convenience of use for patients who have lost motor function.

The effects of the present application are not limited to the above-mentioned effects, and effects not mentioned will be clearly understood by those skilled in the art from this specification and the accompanying drawings.

1 is a diagram for explaining a brain-computer interface device based on an optimal channel selection algorithm using a near-infrared spectroscopy technique according to an embodiment.
2 is an example of a NIRS data acquisition method according to an embodiment.
3 shows a channel arrangement diagram according to an embodiment.
4 is a flowchart of a brain-computer interface method based on an optimal channel selection algorithm using a near-infrared spectroscopy technique according to an embodiment.
5 to 9 are experimental results of a brain-computer interface device based on an optimal channel selection algorithm using a near-infrared spectroscopy technique according to an embodiment.

The foregoing objects, features and advantages of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings. However, the present invention can apply various changes and can have various embodiments. Hereinafter, specific embodiments will be illustrated in the drawings and described in detail.

In the drawings, the thickness of layers and regions is exaggerated for clarity, and elements or layers may be "on" or "on" other elements or layers. What is referred to includes all cases where another layer or other component is intervened in the middle as well as immediately above another component or layer. Like reference numerals designate essentially like elements throughout the specification. In addition, components having the same function within the scope of the same idea appearing in the drawings of each embodiment are described using the same reference numerals.

If it is determined that a detailed description of a known function or configuration related to the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted. In addition, numbers (eg, first, second, etc.) used in the description process of this specification are only identifiers for distinguishing one component from another component.

In addition, the suffixes "module" and "unit" for the components used in the following description are given or used together in consideration of ease of writing the specification, and do not have meanings or roles that are distinguished from each other by themselves.

1 is a diagram for explaining a brain-computer interface device based on an optimal channel selection algorithm using a near-infrared spectroscopy technique according to an embodiment.

Referring to FIG. 1, a brain-computer interface device 10 (hereinafter, may be abbreviated as a brain-computer interface device) based on an optimal channel selection algorithm using a near-infrared spectroscopy technique includes a data acquisition module 100 and a sensor 200. , the output module 300 and the control module 400 may be included.

The brain-computer interface device 10 may refer to a device that measures brain activity of a user and helps users drive external equipment. For example, the brain-computer interface device 10 may be a device that directly connects a user's brain and a computer, measures NIRS data from the user, and analyzes the user's motion imagination as the user's intention. As an example, the brain-computer interface device 10 may be a device that helps paralyzed patients such as amyotrophic lateral sclerosis or confined syndrome to maneuver a wheelchair through interaction between the brain and an external device, and the brain and the external device. Any device capable of performing direct or indirect interaction between the devices may be used.

The data acquisition module 100 may acquire data related to physiological signals and/or behavioral signals of the user. For example, the data acquisition module 100 may acquire NIRS data for the user, but system electroencephalography (EEG) system, electrocardiogram (ECG) system, electromyography (EMG) system, electrochemical sensing system, and It may include, but is not limited to, an eye tracking system.

In some implementations, for example, the data acquisition module 100 may include physiological sensors such as NIRS, EEG, ECG, EMG, electrical signals coupled to signal acquisition devices, such as analog or digital amplifiers coupled to memory. may include chemical or other types of sensor devices.

In one embodiment, the data acquisition module 100 uses a plurality of channels formed by the sensor 200 to perform NIRS for each of a plurality of body parts (a plurality of brain regions) of the user corresponding to the plurality of channels. data can be obtained.

The NIRS data may be data related to the user's brain activity. For example, the data acquisition module 100 may acquire NIRS data for the user from the sensor 200 attached around the prefrontal cortex.

In one embodiment, the data acquisition module 100 may acquire NIRS data related to the user's task performance. For example, the NIRS data may include first NIRS data related to activation of brain activity during task performance and second NIRS data related to stabilization of brain activity after task performance.

2 is an example of a NIRS data acquisition method according to an embodiment.

Referring to FIG. 2 , the data acquisition module 100 according to an embodiment may obtain first NIRS data and second NIRS data from a user performing and resting on a task. In performing the task, the user may perform a mental arithmetic task (MA task) and a relaxation task (baseline task, BL task). In the rest, the user may take a break so that the brain activity changed during the task is stabilized. While the user performs the task and settles down, the data acquisition module 100 may collect NIRS data from a sensor attached to the user's prefrontal cortex.

For example, the user repeats a mental arithmetic task of repeatedly subtracting a single digit from an arbitrary three-digit number according to instructions displayed on the output module 300 and a relaxation task of taking a break in a comfortable posture 30 times for 10 seconds each. can do.

The sensor 200 may detect a user's brain activity or neuron activity signal. For example, the sensor 200 may detect a signal related to a user's brain activity based on NIRS. The sensors 200 are composed of a plurality (N) and may be located in a plurality of body parts (a plurality of brain regions) of the user, and the plurality of sensors 200 are formed by using a plurality of channels formed by the plurality of sensors 200. A user's brain activity or neuron activity signal for each of a plurality of brain regions of the user corresponding to the channel may be detected.

3 shows a channel arrangement diagram according to an embodiment.

Referring to FIG. 3 , the sensor 200 may form 52 channels through 24 light emitting units 210 and 28 light receiving units 220 .

NIRS may refer to a technique of projecting near-infrared rays into the brain from the surface of the scalp and measuring optical changes at various wavelengths as the light is refracted and reflected from the surface. For example, NIRS specifically describes the case of monitoring hemodynamics changes caused by neural activity occurring in the brain. It can penetrate to a depth of about 1 cm to 3 cm from the skull. By irradiating these near-infrared rays to the brain tissue of the human head and detecting the near-infrared rays that are reflected, scattered, or transmitted therefrom, in the human cerebral cortex Occurring hemodynamics (eg, the concentration of oxygen (ie, oxidized hemoglobin) in the blood, etc.) can be monitored. Such NIRS can be used to effectively measure cerebral hemodynamics, detect localized blood volume and oxygenation changes, or create functional maps of brain activity.

The sensor 200 may include a light emitting unit 210 and a light receiving unit 220 . For example, the sensor 200 may form a channel through the light emitting unit 210 and the light receiving unit 220 . Furthermore, the plurality of sensors 200 may form a plurality of channels through the plurality of light emitting units 210 and the plurality of light receiving units 220 .

The light emitting unit 210 may generate an optical signal to the body. For example, the light emitting unit 210 may generate near-infrared rays to the user's brain. The light emitting unit 210 may include, but is not limited to, a laser or at least one light emitting diode.

The light receiving unit 220 may detect an optical signal generated by the light emitting unit 210 with respect to the body. For example, the light receiving unit 220 may receive near infrared rays generated by the light emitting unit 210 toward the brain, but not absorbed by the brain and reflected or induced into the brain. The light receiving unit 220 may include at least one photodiode, but is not limited thereto.

The light emitting unit 210 and the light receiving unit 220 may be disposed on a user's body part. For example, pairs of the light emitting unit 210 and the light receiving unit 220 may be disposed at predetermined intervals at various points on the user's head. The sensor 200 may detect a user's brain activity or neuron activity signal through an optical signal generated by the light emitting unit 210 and sensed by the light receiving unit 220 .

Of course, the sensor 200 includes not only a NIRS-based sensor, but also electric, fluid, chemical, and magnetic sensors such as EEG (electroencephalography), Doppler ultrasound, and MEG (magnetoencephalography) (not limited to these) It is also possible to include or be configured through a combination thereof.

The output module 300 may output information related to the brain-computer interface device 10 . For example, the output module 300 may include an indicator that externally indicates information related to the state of the brain-computer interface device 10 in various ways. For example, the indicator may be implemented as (but not limited to) an LED lamp, a speaker, a motor, a haptic device, a vibrator, a signal output circuit, and the like. The indicator may visually or audibly output a switch operation notification, a low battery notification, and the like.

In addition, the output module 300 may include a configuration for providing a screen such as LCD, OLED, AMOLED, etc. (not limited thereto). Here, when the output module 300 is provided as a touch screen, the output module 300 may also function as an input sensor. In this case, a separate input sensor may not be provided according to selection. For example, a screen for a user to perform a task may be provided to the output module 300 .

The control module 400 may control each component of the brain-computer interface device 10 or process and calculate various types of information. For example, the control module 400 includes the data acquisition module 100 and the sensor 200 so that the brain-computer interface device 10 selects at least one channel among a plurality of channels based on the NIRS data acquired for the user. ), the output module 300, etc. can be controlled.

The control module 400 may be implemented in software, hardware, or a combination thereof. For example, in terms of hardware, the control module 400 may be implemented as a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a semiconductor chip, or other various types of electronic circuits. For example, in terms of software, the control module 400 may be implemented in a logic program executed according to the above-described hardware or various computer languages, etc. In addition, the control module 400 may be implemented by a microprocessor, a microcontroller, a digital signal processing device (DSP) ) or any combination thereof, but is not limited thereto.

4 is a flowchart of a brain-computer interface method based on an optimal channel selection algorithm using a near-infrared spectroscopy technique according to an embodiment.

Referring to FIG. 4, the brain-computer interface method based on the optimal channel selection algorithm using the near-infrared spectroscopy technique according to an embodiment includes forming a plurality of channels by N sensors (S100), obtaining NIRS data (S200), and At least one channel selection (S300) from among a plurality of channels may be included.

The brain-computer interface device 10 may form a plurality of channels by N sensors (S100).

The sensors 200 may be composed of a plurality (N) and may be located in a plurality of brain regions of the user. Using a plurality of channels formed by the plurality of sensors 200, the user's response corresponding to the plurality of channels is used. A user's brain activity or neuron activity signal for each of a plurality of brain regions may be detected.

The light emitting unit 210 and the light receiving unit 220 may be disposed on a user's body part. For example, pairs of the light emitting unit 210 and the light receiving unit 220 may be disposed at predetermined intervals at various points on the user's head. The sensor 200 may detect a user's brain activity or neuron activity signal through an optical signal generated by the light emitting unit 210 and sensed by the light receiving unit 220 .

The brain-computer interface device 10 may acquire NIRS data (S200).

The data acquisition module 100 may obtain NIRS data for each of a plurality of brain regions of a user corresponding to a plurality of channels by using a plurality of channels formed by N sensors.

In one embodiment, the acquired NIRS data may include first NIRS data related to activation of brain activity during task performance and second NIRS data related to stabilization of brain activity after task performance.

The brain-computer interface device 10 may select at least one channel among a plurality of channels based on the NIRS data (S300).

The control module 400 may select at least one channel among a plurality of channels based on the NIRS data acquired from the data acquisition module 100 .

In one embodiment, the control module 400 may extract feature data for NIRS data. For example, the control module 400 may extract the average of NIRS data as feature data.

In one embodiment, the control module 400 may divide the NIRS data into a plurality of time intervals, and the feature data may include a plurality of divided feature data extracted from the divided plurality of time intervals. For example, the control module 400 may divide the NIRS data into three sections (0 to 5, 5 to 10, and 10 to 15 seconds) and extract an average for each section as a feature. Here, the control module 400 may extract three feature data for each channel.

In one embodiment, the control module 400 may pre-process NIRS data.

The control module 400 may preprocess noise removal by applying a band pass filter to the NIRS data. For example, a plurality of segmentation feature data may be pre-processed to remove noise by applying a band pass filter.

The control module 400 may perform baseline adjustment preprocessing on the NIRS data by using at least a part of the immediately preceding time interval. For example, the plurality of segmented feature data may be subjected to baseline adjustment preprocessing using at least a part of the immediately preceding time interval.

In one embodiment, the control module 400 may select at least one channel among the plurality of channels in consideration of the extracted characteristic data and the number N of the plurality of sensors.

The control module 400 may calculate classification accuracy using the extracted feature data. Classification accuracy may refer to a ratio between the total number of test signals and the number of test signals accurately classified through the classifier.

The control module 400 may select at least one channel among the plurality of channels based on the calculated classification accuracy and the number N of the plurality of sensors.

The control module 400 may calculate a channel selection reference value based on the classification accuracy and the number (N) of the plurality of sensors.

For example, the channel selection reference value may be determined by an algorithm based on Equations 1 and 2 below.

(Equation 1)

(Equation 2)

The control module 400 may select at least one channel among a plurality of channels by applying sequential forward selection (SFS) to a channel selection reference value.

Sequential forward selection is a method of extracting only features with optimal performance from high-dimensional features and reducing the dimensions. By applying it to channel selection, the optimal channel can be selected.

In detail, the sequential forward selection method process of the brain-computer interface device 10 will be described. Here, it is assumed that there are 5 channels.

The brain-computer interface device 10 may calculate classification accuracy for each of channels 1 to 5. The brain-computer interface device 10 may select a channel showing the highest classification accuracy. It is assumed that channel 3 has the highest classification accuracy. The brain-computer interface device 10 may calculate classification accuracy for channel combinations [1, 3], [2, 3], [3, 4], and [3, 5] by fixing channel 3.

The brain-computer interface device 10 may select a channel combination showing the highest classification accuracy. Similarly, we assume that the [1, 3] combination was the highest. The brain-computer interface device 10 calculates the classification accuracy for the combinations [1, 3, 2], [1, 3, 4], and [1, 3, 5] with the [1, 3] channel combination fixed. can do.

The brain-computer interface device 10 may repeat the above process for all channels until a section in which classification accuracy does not increase any more.

The control module 400 may select channels not only by sequential forward selection but also by other selection methods such as sequential forward floating selection (SFFS), but is not limited thereto.

The control module 400 may optimize an algorithm for channel selection reference values. For example, the channel selection reference value may be determined by Equation 3 below.

(Equation 3)

The brain-computer interface device 10 may optimize the algorithm by adding a parameter α to the channel selection criterion in order to reduce the number of used sensors 200 as much as possible. α is a value between 0.1 and 0.9 and can be increased by 0.1 units, and the larger the value of α, the greater weight can be applied to classification accuracy.

The brain-computer interface device 10 may further include a configuration (not shown) for optimizing an algorithm for a channel selection reference value. For example, the brain-computer interface device 10 may further include a learning module for algorithm optimization for channel selection reference values, which may be controlled by the control module 400 .

5 to 9 are experimental results of a brain-computer interface device based on an optimal channel selection algorithm using a near-infrared spectroscopy technique according to an embodiment. Experimental results for the brain-computer interface device 10 will be described with reference to FIGS. 5 to 9 .

The results of measuring the offline/online performance of the brain-computer interface device 10 will be described.

In the offline performance measurement, the mental arithmetic task and the relaxation task were classified through a linear discriminant analysis (LDA) using data extracted from 15 subjects as input data. The performance of the channel selection method was calculated through classification accuracy and sensor ratio. In the case of optimization, the number of channels in methods 1 and 2 was fixed to 8, which showed no difference in performance compared to when all channels were used, and θ (α) was added. was calculated as θ(α) is calculated as the ratio of the classification accuracy and the sensor, and was used to set the optimal value of α by reflecting the rate of change of the classification accuracy and the ratio of the used sensor.

All indicators were calculated using 5x5 fold cross-validation, and for performance comparison, a channel selection method that does not consider the sensor ratio (hereinafter, method 1) and a channel selection method that considers the sensor ratio (hereinafter, method 2), respectively Performance was calculated.

In the online performance measurement (verification of practicality), a simulation was conducted under the assumption that an actual brain-computer interface device 10 was used to prove the practicality of the algorithm. The analysis was carried out by sequentially importing 12, 24, 36, and 48 data after listing the previously used data in chronological order, and was performed based on the values set through the optimization previously performed. Through the simulation, the difference with the offline performance was confirmed, and the practicality was verified by quantifying the reproducibility of the algorithm and the amount of required data.

As can be seen in Figure 5, the classification accuracy was 76.1 ± 7.7% when all channels were used. When method 1 was used, it was from 64.44 ± 7.11 % to 74.67 ± 6.55%, and when method 2 was used, it was from 64.44 ± 7.11 % to 75.56 ± 6.10 %. There was no statistical difference (Wilcoxon signed-rank test, p > 0.05). In particular, in the case of method 2, there was only a difference of 0.5% compared to when all channels were used.

As can be seen in FIG. 6, as the number of selected channels increases, the ratio of sensors used decreases from 0.98 ± 0.02 to 0.75 ± 0.02 in the case of method 1, and from 0.75 ± 0.00 to 0.56 ± 0.02 in the case of method 2. ratio was reduced. Method 2 significantly reduced the ratio of sensors used compared to Method 1 by about 33.93% (Wilcoxon signed rank test, p < 0.05). In addition, when actually selected channels are observed through FIG. 7 , it can be seen that the distribution of channels and the number of sensors are significantly reduced.

8 shows the classification accuracy, the ratio of the sensors used, and θ(α) according to the value of α. Classification accuracy tended to be statistically lower when α was less than 0.3 compared to when all channels were used. In the case where the ratio of the used sensors was 0.5 or less, there was no statistical difference between each other, and all showed statistically lower values than those in the case of 0.6 or more. There was no statistical difference between θ(α) when it was less than 0.4, and all showed statistically lower values than when it was 0.5 or more. The optimal α value was set to 0.4 through θ(α) considering the classification accuracy and the ratio of the used sensors together.

9 is an online analysis result for verifying the practicality of the brain-computer interface device 10. There was no statistical difference in classification accuracy between the methods, but in the case of method 1, at least 48 data and in the case of method 2, at least 36 data should be used to show no difference with offline performance. As for the ratio of sensors used, methods 1 and 2 showed statistical differences in all cases, and in case of method 1, at least 12 data and in case of method 2, at least 48 data should be used to show no difference with offline performance. Statistical differences were seen between methods 1 and 2, such as the percentage of sensors used previously, and in the case of method 1, a statistical difference with offline performance was observed in all cases, and in the case of method 2, it was confirmed that at least 36 data should be used. there is. Considering the three indicators, the optimal number of data cannot be determined, but considering classification accuracy and θ(α) calculated through the number of sensors used, it is judged appropriate to use at least 36 data.

Through the foregoing results, the optimal channel selection algorithm-based brain-computer interface device 10 using the near-infrared spectroscopy technique according to an embodiment can efficiently reduce the number of channels and the number of sensors while maintaining performance.

As described above, although the embodiments have been described with limited examples and drawings, those skilled in the art can make various modifications and variations from the above description. For example, the described techniques may be performed in an order different from the method described, and/or components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved.

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

Brain-computer interface device based on optimal channel selection algorithm using near-infrared spectroscopy technique 10
Data Acquisition Module 100
sensor 200
light emitting part 210
light receiver 220
output module 300
control module 400

Claims (18)

In the brain-computer interface device based on the optimal channel selection algorithm using the near-infrared spectroscopy technique,
Using a plurality of channels formed by N sensors - the sensor includes a light emitting unit generating light signals corresponding to near infrared rays with respect to the body and a light receiving unit detecting light signals corresponding to near infrared rays from the body - a data acquisition module for obtaining NIRS data for each of a plurality of body parts of the user corresponding to the plurality of channels, wherein the plurality of body parts include a plurality of brain regions of the user; and
A control module for selecting at least one channel among the plurality of channels based on the NIRS data obtained from the data acquisition module; includes,
The control module,
Extracting feature data for the NIRS data and selecting at least one channel among the plurality of channels in consideration of the extracted feature data and the number (N) of the sensors;
The control module calculates classification accuracy using the extracted feature data, calculates a channel selection reference value based on the calculated classification accuracy and the number (N) of the sensors, and sequentially advances the channel selection reference value ( Selecting at least one channel among the plurality of channels by applying sequential forward selection (SFS),
A brain-computer interface device based on an optimal channel selection algorithm using near-infrared spectroscopy.
delete delete According to claim 1,
The channel selection reference value is determined by an algorithm according to Equations 1 and 2 below
(Equation 1)
(Equation 2)
A brain-computer interface device based on an optimal channel selection algorithm using near-infrared spectroscopy.
According to claim 1,
The channel selection reference value is determined by Equation 3 below,
(Equation 3)
In Equation 3, α is a value set between 0.1 and 0.9, and the sensor ratio used is determined by Equation 4 below.
(Equation 4)
A brain-computer interface device based on an optimal channel selection algorithm using near-infrared spectroscopy.
According to claim 1,
The control module divides the NIRS data into a plurality of time intervals,
The feature data includes a plurality of divided feature data extracted from the plurality of divided time intervals.
A brain-computer interface device based on an optimal channel selection algorithm using near-infrared spectroscopy.
According to claim 6,
The plurality of split feature data is pre-processed to remove noise by applying a band pass filter.
A brain-computer interface device based on an optimal channel selection algorithm using near-infrared spectroscopy.
According to claim 6,
The plurality of split feature data is pre-processed for baseline adjustment using at least some of the immediately preceding time intervals.
A brain-computer interface device based on an optimal channel selection algorithm using near-infrared spectroscopy.
According to claim 1,
The NIRS data is
First NIRS data related to activation of brain activity during task performance and
Containing second NIRS data related to stabilization of brain activity after task performance
A brain-computer interface device based on an optimal channel selection algorithm using near-infrared spectroscopy.
In the optimal channel selection algorithm-based brain-computer interface method using the near-infrared spectroscopy technique performed by the optimal channel selection algorithm-based brain-computer interface device using the near-infrared spectroscopy technique,
By the data acquisition module of the brain-computer interface device based on the optimal channel selection algorithm, N sensors - the sensors include a light emitting unit generating light signals corresponding to near infrared rays with respect to the body and light signals corresponding to near infrared rays from the body Including a light receiver for detecting a plurality of body parts of the user corresponding to a plurality of channels formed by the plurality of body parts including a plurality of brain regions of the user Acquiring NIRS data for each ; and
Selecting, by a control module of the optimal channel selection algorithm-based brain-computer interface device, at least one channel among the plurality of channels based on the NIRS data obtained from the data acquisition module;
The selection step is
Extracting, by the control module, feature data for the NIRS data and selecting at least one channel among the plurality of channels in consideration of the extracted feature data and the number (N) of the sensors;
The selection step is
The control module calculates classification accuracy using the extracted feature data, calculates a channel selection reference value based on the calculated classification accuracy and the number (N) of the sensors, and sequentially advances with respect to the channel selection reference value. Selecting at least one channel among the plurality of channels by applying a selection,
Brain-computer interface method based on optimal channel selection algorithm using near-infrared spectroscopy technique.
delete delete According to claim 10,
The channel selection reference value is determined by an algorithm according to Equations 1 and 2 below
(Equation 1)
(Equation 2)
Brain-computer interface method based on optimal channel selection algorithm using near-infrared spectroscopy technique.
According to claim 10,
The channel selection reference value is determined by Equation 3 below,
(Equation 3)
In Equation 3, α is a value set between 0.1 and 0.9, and the sensor ratio used is determined by Equation 4 below.
(Equation 4)
Brain-computer interface method based on optimal channel selection algorithm using near-infrared spectroscopy technique.
According to claim 10,
The selecting step includes dividing the NIRS data into a plurality of time intervals by the control module, and the feature data includes a plurality of divided feature data extracted from the divided plurality of time intervals.
Brain-computer interface method based on optimal channel selection algorithm using near-infrared spectroscopy technique.
According to claim 15,
The selecting step includes pre-processing noise removal by applying a band pass filter to the plurality of divided feature data by the control module.
Brain-computer interface method based on optimal channel selection algorithm using near-infrared spectroscopy technique.
According to claim 15,
The selecting step includes preprocessing the plurality of divided feature data for baseline adjustment using at least a part of an immediately preceding time interval by the control module.
Brain-computer interface method based on optimal channel selection algorithm using near-infrared spectroscopy technique.
According to claim 10,
The NIRS data is
First NIRS data related to activation of brain activity during task performance and
Containing second NIRS data related to stabilization of brain activity after task performance
Brain-computer interface method based on optimal channel selection algorithm using near-infrared spectroscopy technique.