KR20200008219A - Method for selecting optimized eeg electrodes based on brain machine interfaces and recording medium for performing the method - Google Patents

Abstract

The present invention relates to a method for selecting minimum electroencephalogram (EEG) electrodes for a brain-machine interface (BMI) which is realized to minimize the number of EEG electrodes substantially affecting BMI decoding accuracy, and a storage medium for performing the same. The method comprises: a step of acquiring feature data of brain signals from a plurality of EEG electrodes during all trials of brain signal measurement; a step of calculating discriminability information of the EEG electrodes for each class in the all trials; a step of using the calculated discriminability information to select priorities of the EEG electrodes in accordance with a degree of affecting BMI decoding accuracy; and a step of determining a final electrode set consisting of at least one EEG electrode essential in brain signal measurement in the all trials.

Description

METHOD FOR SELECTING OPTIMIZED EEG ELECTRODES BASED ON BRAIN MACHINE INTERFACES AND RECORDING MEDIUM FOR PERFORMING THE METHOD}

The present invention relates to a method for selecting an EEG minimum electrode for BMI and a storage medium for performing the method. More specifically, the minimum EEG for BMI implemented to minimize the number of EEG electrodes that substantially affects the BMI decoding accuracy. An electrode selection method and a storage medium for performing the method.

'Brain-Machine' Interface (BMI) is a fusion that connects the human brain with a machine and utilizes brain waves, which are brain signals, in real time, or improves human ability by inputting and modulating external information. Technology.

Here, electroencephalography (EEG) is a representative biosignal that captures brain activity spatio-temporal and has been widely used in clinical and brain function research.

In the EEG measurement, an electrode is attached to the user's scalp to measure the electric current generated by the activity of the brain.

In general, measuring EEG with an EEG electrode involves drilling an electrode through a hole in the skull, but according to the development of technology, EEG can be simultaneously detected from as many sites as possible using a non-invasive method using a polyimide-based microelectrode. A method of recording has been proposed.

Conventionally, a hair cap equipped with an EEG electrode is worn to drive a BMI. In addition, a large number of EEG electrodes are required to measure brain waves, and the data measured from the electrodes are processed as the number of electrodes increases. There was a need for a high performance computing device.

In addition, it was difficult to use in a real life environment, and had the disadvantage of being inefficient in implementing BMI using all electrodes.

Korean Patent Publication No. 10-2017-0056295 Korean Patent Publication No. 10-2017-0091247

One aspect of the present invention is to calculate the correlation maximum or covariance minimum between the characteristic properties of the brain signal and the BMI decoding performance, and then to substantially influence the BMI decoding accuracy by prioritizing the EEG electrodes that affect the BMI decoding accuracy. Provided are an EEG minimum electrode selection method for BMI and a storage medium for performing the method.

The technical problem of the present invention is not limited to the technical problem mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the following description.

EEG minimum electrode selection method for BMI according to an embodiment of the present invention, in performing a mental task for driving a brain-machine interface (BMI), a plurality of EEG during every trial of brain signal measurement (trial) obtaining feature property data of a brain signal from electroencephalogram electrodes; Calculating discriminability information of the EEG electrodes for each class in all the trials by using the acquired characteristic data of the brain signal and a target BMI decoding performance; Prioritizing the EEG electrodes according to the degree of influence on the BMI decoding accuracy using the calculated discrimination power information; And determining a final electrode set composed of at least one EEG electrode essential for brain signal measurement in all the above attempts based on priorities of the EEG electrodes.

In one embodiment, the calculating of the discrimination power information may include: a maximum value of a correlation value or a minimum value of a covariance value between BMI decoding performance and feature attribute data of a brain signal obtained from each EEG electrode in all the trials. By calculating the identification information for each electrode for each class of BMI can be calculated.

In an embodiment, the determining of the final electrode set may include adding an i-th order electrode having the highest correlation value to a first electrode set including only the first rank electrode having the highest correlation value and adding the second electrode set. A set expansion step of generating a; Comparing the BMI decoding accuracy of the second electrode set and the first electrode set, if the BMI decoding accuracy of the first electrode set is better than the BMI decoding accuracy of the second electrode set, the first electrode set is replaced with the final electrode set. If the BMI decoding accuracy of the second electrode set is better than the BMI decoding accuracy of the first electrode set, an i + 1 rank electrode having the highest correlation value i + 1 is added to the second electrode set. An accuracy comparison step of generating a third set of electrodes; And repeating the execution of the accuracy comparison step again while sequentially adding the remaining EEG electrodes to the third electrode set in order until the final electrode set is determined.

In an embodiment, the determining of the final electrode set may include generating the second electrode set by adding an i-th rank electrode having the lowest covariance value to the first electrode set including only the first rank electrode having the lowest covariance value. A set expansion step; Comparing the BMI decoding accuracy between the second electrode set and the first electrode set, if the BMI decoding accuracy of the first electrode set is better than the BMI decoding accuracy of the second electrode set, the first electrode set is replaced with the final electrode set. If the BMI decoding accuracy of the second electrode set is better than the BMI decoding accuracy of the first electrode set, an i + 1 rank electrode having the i + 1th lower covariance value is added to the second electrode set. An accuracy comparison step of generating a third electrode set; And repeating the execution of the accuracy comparison step again while sequentially adding the remaining electrodes to the third electrode set in order until the final electrode set is determined.

In one embodiment, the EEG minimum electrode selection method for BMI according to an embodiment of the present invention, the electrode from which similar feature attribute data is obtained among the EEG electrodes constituting the final electrode set determined in the step of determining the final electrode set Clustering the first electrode set consisting of a; Generating a second electrode set by removing a randomly selected electrode from the first electrode set; And comparing the BMI decoding accuracy of the first electrode set and the second electrode set to determine a final electrode set composed of at least one EEG electrode essential for brain signal measurement in all the trials.

In an embodiment, the clustering may include clustering the identification information of all the EEG electrodes into a plurality of groups using at least one of a K-means clustering technique and a similar technique of an unsupervised learning sequence. It is possible to generate a first electrode set composed of electrodes from which feature characteristic data similar to each other is obtained.

EEG minimum electrode selection method for BMI according to another embodiment of the present invention, in the mental task for driving Brain-Machine Interface (BMI), the characteristics similar to each other during every trial (trial) of brain signal measurement Clustering a first electrode set consisting of a plurality of electroencephalogram (EEG) electrodes from which feature data is obtained; Removing a randomly selected EEG electrode from the first electrode set to generate a second electrode set; And comparing the BMI decoding accuracy of the first electrode set and the second electrode set to determine a final electrode set composed of at least one EEG electrode essential for brain signal measurement in all the attempts.

In an embodiment, the determining of the final electrode set may include: comparing accuracy estimating BMI decoding accuracy using only the EEG electrodes of the second electrode set; If the BMI decoding accuracy of the second electrode set is better than the BMI decoding accuracy of the first electrode set, the corresponding EEG electrode is removed from the first electrode set, and the BMI decoding accuracy of the first electrode set is the second electrode. Determining an electrode set which maintains the corresponding EEG electrode in the first electrode set if it is better than the BMI decoding accuracy of the set; And repeatedly performing the electrode set determination step again while removing or maintaining the remaining EEG electrodes in any order from the electrode set determined in the electrode set determination step until the final electrode set is determined.

In one embodiment, the EEG minimum electrode selection method for BMI according to another embodiment of the present invention, the feature property of the brain signal from each of the EEG electrodes constituting the electrode set determined in the step of determining the final electrode set (feature) Acquiring data; Calculating discriminability information of EEG electrodes for each class in all the trials by using the acquired characteristic data of the brain signal and the target BMI decoding performance; Selecting priorities of the EEG electrodes according to the degree of influence on the target BMI decoding accuracy using the calculated discrimination power information; And determining a final electrode set composed of at least one EEG electrode essential for brain signal measurement in all the above attempts based on priorities of the EEG electrodes.

In a computer-readable storage medium according to another embodiment of the present invention, a computer program for performing the EEG minimum electrode selection method for BMI is recorded.

According to one aspect of the present invention, while performing the mental task for driving the BMI, it is possible to minimize the number of EEG electrodes that have a substantial impact on the BMI decoding accuracy, which can contribute to the development of portable EEG-BMI technology, driving the BMI By reducing the amount of computation required, the BMI response time can also be significantly improved.

In addition, by applying the EEG electrode only to a specific position of the scalp, the hair cap can be peeled off to improve the usability of the BMI, and can be easily applied to real life, leading to the popularization of the BMI.

In addition, it can provide scalability to implement an interlocking environment of IoT, TV, refrigerator or robot and BMI, and can implement AR / VR-based BMI in favor of mounting a holo lens.

1 is a view comparing a conventional BMI driving method and a BMI driving method according to the present invention.
2 is a flowchart illustrating a method of selecting an EEG minimum electrode for BMI according to an embodiment of the present invention.
3 is a graph showing feature information calculated at each EEG electrode and an EEG electrode automatically selected by the present invention.
FIG. 4 is a flowchart for describing a step of determining the final electrode set of FIG. 2.
5 is a flowchart illustrating a method of selecting an EEG minimum electrode for BMI according to another embodiment of the present invention.
6 is a flowchart illustrating a method of selecting an EEG minimum electrode for BMI according to another embodiment of the present invention.
Fig. 7 illustrates the EEG electrode selection according to the present invention.
FIG. 8 is a flowchart illustrating an operation of determining a final electrode set of FIG. 6.
9 is a flowchart illustrating a method of selecting an EEG minimum electrode for BMI according to still another embodiment of the present invention.

DETAILED DESCRIPTION OF THE INVENTION The following detailed description of the invention refers to the accompanying drawings that show, by way of illustration, specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It is to be understood that the various embodiments of the invention are different but need not be mutually exclusive. For example, certain shapes, structures, and characteristics described herein may be embodied in other embodiments without departing from the spirit and scope of the invention in connection with one embodiment. In addition, it is to be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. The following detailed description, therefore, is not to be taken in a limiting sense, and the scope of the present invention is defined only by the appended claims, along with the full scope of equivalents to which such claims are entitled. Like reference numerals in the drawings refer to the same or similar functions throughout the several aspects.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings.

As shown in FIG. 1, the present invention relates to BMI driving based on a paradigm optimal channel set generation using automatically selected minimum EEG electrodes, rather than driving the BMI through the existing entire channel configuration.

In order to realize this, the present invention calculates the correlation value peak or the covariance minimum value between the feature characteristic of the brain signal and the BMI decoding performance from each of the EEG electrodes, and then selects an electrode that significantly affects the BMI decoding accuracy. The minimum electrodes are selected in order of priority.

2 is a flowchart illustrating a method of selecting an EEG minimum electrode for BMI according to an embodiment of the present invention.

Referring to FIG. 2, first, in performing a mental task for driving a Brain-Machine® Interface (BMI), brain signals from a plurality of EEG (electroencephalogram) electrodes during every trial of brain signal measurement are performed. Feature property data is acquired (S110).

Discriminability information of the EEG electrodes for each class in all trials is calculated using the feature attribute data of the brain signal acquired in the above-described step S110 and the BMI decoding performance set as the target (S120).

The data obtained from the EEG electrodes include dozens or hundreds of brain signal measurement trials, from which the correlation or covariance between each trial can be calculated to identify the electrode-specific discrimination power for each class of BMI. discriminability (R ² )) information can be calculated.

In a particular trial (ie, one trial), it is not possible to calculate the correlation or covariance between the trial and trial for each class or between trials.

Accordingly, in the present invention, the correlation or covariance between total attempts for any class is to be calculated.

)를 의미하는 것으로, 다음의 수학식 1과 같이 표현될 수 있다.Here, the correlation refers to a point biserial correlation coefficient,

), Which can be expressed by Equation 1 below.

은 양성 이진 값(positive binary value)(i.e. "1")을 받는 전극 집합의 평균(전체 test에 대한)이고,

는 음성 이진 값(negative binary value)(i.e. "0")을 받는 전극 집합의 평균(전체 test에 대한)이며,

은 전체 test의 표준편차이고,

는 "0" 그룹에 속한 사건의 비율이고,

는 "1" 그룹에 속한 사건의 비율이다.here,

Is the mean (for the entire test) of a set of electrodes that receives a positive binary value (ie "1"),

Is the mean (for the entire test) of the electrode set that receives the negative binary value (ie "0"),

Is the standard deviation of the entire test,

Is the proportion of events in group "0",

Is the proportion of events in the "1" group.

Here, the covariance means a normalized LDA based shrinkage value (shrinkage-regularized LDA based) and may be calculated as follows.

After generating the hyper parameter (w) using the electrode data (x) containing all normalized trials and the serial number (y) of the class corresponding to all trials in sequence, each EEG Covariance values can be calculated by weighting and standardizing the electrodes.

The discrimination power information calculated in the above-described step S120 may be represented by respective graphs as shown in FIG. 3, and each electrode because the electrode set of specific EEG electrodes shows a more characteristic pattern than other electrode sets. Sets can be distinguished from each other.

In the calculating of the above-described discrimination power information (S120), the maximum value or the lowest covariance value between the feature attribute data of the brain signal obtained from each EEG electrode and the BMI decoding performance is calculated in every trial. Identification information for each electrode of each class of BMI can be calculated.

The priority of the EEG electrodes is selected based on the degree of influence on the BMI decoding accuracy using the discrimination power information calculated in the above-described step S120 (S130).

The minimum EEG electrode constituting the final electrode set is determined by determining a final electrode set composed of at least one EEG electrode essential for brain signal measurement in all trials based on the priorities of the EEG electrodes selected in step S130 described above (S130). Can be selected automatically.

Here, the method of determining the specific final electrode set in step S130 described above will be described with reference to FIG. 4.

In the method for selecting an EEG minimum electrode for BMI according to an embodiment of the present invention having the steps as described above, after calculating feature information for each class between trials by correlation and covariance, each calculated feature information is used. Distinguish discrimination between channels.

Minimum electrode by automatically selecting the EEG electrodes (i.e., O1, Oz, and O2 of FIG. 3) by various clustering methods as shown in FIG. 3 having similar characteristics with each discriminating power distinguished from other EEG electrodes. Can be selected.

According to the EEG minimum electrode selection method for BMI according to an embodiment of the present invention having the steps as described above, the number of EEG electrodes that substantially affects the BMI decoding accuracy can be minimized during the mental task for driving the BMI. Therefore, it can contribute to the development of portable EEG-BMI technology, and the BMI response time can be drastically reduced by reducing the amount of computation required for driving the BMI.

Conventional ERP and covariance calculations have been limited to raw data and specific paradigms, but the present invention provides an FFT-based frequency component extraction and correlation and covariance between each trial to capture optimal characteristics for the general experimental paradigm. By analyzing, we can overcome the limitations of existing technology.

In addition, the existing technology is a method limited to the electrode device developed only for a specific experiment, it was difficult to apply to a variety of real environments, the present invention is an automated application to all the EEG signals obtained through the electrodes of the general EEG equipment This can overcome the limitations of existing technologies.

In addition, the existing methods based on specific covariance (p-covariance) and clustering (Riemann clustering) techniques have limitations in that they are not optimally guaranteed and difficult to apply to practical BMI techniques. Various correlation and covariance methods and clustering techniques can be freely selected and analyzed to overcome the limitations of existing techniques in determining the best performing electrode set.

FIG. 4 is a flowchart for describing a step of determining the final electrode set of FIG. 2.

Referring to FIG. 4, in an embodiment of determining the final electrode set (S140), first, an i rank electrode having the highest correlation value and a first rank electrode having only the first correlation electrode having the highest correlation value are first formed. In addition to the electrode set, a second electrode set is generated (S141).

By comparing the BMI decoding accuracy of the second electrode set and the first electrode set, if the BMI decoding accuracy of the first electrode set is better than the BMI decoding accuracy of the second electrode set (Yes in S142), If the first electrode set is maintained (S143) (that is, the existing electrode set in which the i rank electrode is excluded) and the above-described accuracy comparison for all the EEG electrodes is completed (Yes in S144), the first electrode set. Is determined as the final electrode set.

However, if the above-described accuracy comparison for all the EEG electrodes has not yet been completed (in the case of No in S144), until the accuracy comparison is completed for all the EEG electrodes (in the case of Yes in S144 or in the case of Yes in S146). The remaining EEG electrodes are sequentially added to the first electrode set according to the ranking (S145), and the accuracy comparison step (S142) is executed again.

However, if the BMI decoding accuracy of the second electrode set is better than the BMI decoding accuracy of the first electrode set (No of S142), and yet, the above-described accuracy comparison for all the EEG electrodes has not been completed (S146 of In case of No), the i-rank electrode is kept as it is and the i + 1 rank with the highest correlation value until the accuracy comparison is completed for all EEG electrodes (Yes in S144 or Yes in S146). The third electrode set is added to the second electrode set to generate the third electrode set (S147), and the accuracy comparison step (S142) is executed again.

That is, the step S140 of determining the final electrode set having the above-described steps corresponds to a method of comparatively analyzing the BMI decoding accuracy by increasing the electrode set in order from the EEG electrode having the high correlation value.

Referring to another embodiment of the step (S140) of determining the final electrode set with reference to FIG. 4, first, it is composed of the first rank electrode having the first low covariance value and the first rank electrode having only the first low covariance value In addition to the first electrode set, a second electrode set is generated (S141).

By comparing the BMI decoding accuracy of the second electrode set and the first electrode set, if the BMI decoding accuracy of the first electrode set is better than the BMI decoding accuracy of the second electrode set (Yes in S142), the first electrode set is the existing electrode set. If the first electrode set is maintained (S143) (that is, the existing electrode set in which the i-rank electrode is excluded) and the above-described accuracy comparison for all the EEG electrodes is completed (Yes in S144), the first electrode set is maintained. The final set of electrodes is determined.

However, if the above-described accuracy comparison for all the EEG electrodes has not yet been completed (in the case of No in S144), until the accuracy comparison is completed for all the EEG electrodes (in the case of Yes in S144 or in the case of Yes in S146). The remaining EEG electrodes are sequentially added to the first electrode set according to the ranking (S145), and the accuracy comparison step (S142) is executed again.

However, if the BMI decoding accuracy of the second electrode set is better than the BMI decoding accuracy of the first electrode set (No of S142), and yet, the above-described accuracy comparison for all the EEG electrodes has not been completed (S146 of In case of No), the i-rank electrode is kept as it is and the i + 1-th low covariance value is maintained until the accuracy comparison is completed for all EEG electrodes (Yes in S144 or Yes in S146). The third electrode set is added to the second electrode set to generate the third electrode set (S147), and the accuracy comparison step (S142) is executed again.

That is, the step S140 of determining the final electrode set having the above steps corresponds to a method of comparatively analyzing the BMI decoding accuracy while increasing the electrode set sequentially from the EEG electrode having the low covariance value.

5 is a flowchart illustrating a method of selecting an EEG minimum electrode for BMI according to another embodiment of the present invention.

Referring to FIG. 5, in the method of selecting an EEG minimum electrode for BMI according to another embodiment of the present invention, after determining the final electrode set of FIG. 2 (S140), in operation S140, the final electrode set is determined. Among the EEG electrodes constituting the determined final electrode set, a first electrode set including electrodes for which similar feature attribute data is obtained is clustered (S150).

In one embodiment, in step S150 described above, each EEG electrode having similar discrimination power, using discriminability information of EEG electrodes for each class in all trials calculated in step S120 described above in FIG. 2. The first electrode set may be generated by clustering the two electrodes.

In one embodiment, clustering step S150 may include at least one of a K-means clustering technique and a similar technique of an unsupervised learning sequence technique. By using one technique, identification power information of all EEG electrodes may be clustered into a plurality of groups to generate a first electrode set including electrodes in which similar characteristic attribute data are obtained.

Here, among the plurality of electrode sets generated through K-average clustering, the average value of discrimination power or a separate cross-validation loss, the validity of the study results, are evaluated by other samples not used in the study. Validation Method) Since the low electrode set will be the minimum electrode set that contributes optimally to the BMI performance, the minimum electrode selection can be automated.

In this case, the unsupervised learning is possible, that is, without the hassle of training in the BMI demonstration (training) has the advantage that can be used directly in real time.

Therefore, according to the present invention, it is also possible to apply other techniques belonging to the unsupervised learning type, such as K mean clustering, and selectively apply a technique for performing the best minimum electrode set configuration according to the paradigm.

In operation S160, the second electrode set is generated by removing a randomly selected electrode from the clustered first electrode set.

The BMI decoding accuracy of the first electrode set generated in S140 and the second electrode set generated in S160 are compared to determine a final electrode set composed of at least one EEG electrode essential for brain signal measurement in all trials (S170).

Here, steps S150, S160, and S170 will be described later with reference to FIG.

In the EEG minimum electrode selection method for BMI according to another embodiment having the above-described steps, after comparing the BMI decoding accuracy of each final electrode set determined in step S140 or step S170 described above with reference to FIG. The electrode set having the decoding accuracy may be determined as the final electrode set.

6 is a flowchart illustrating a method of selecting an EEG minimum electrode for BMI according to another embodiment of the present invention.

The EEG minimum electrode selection method for BMI according to another embodiment of the present invention is effective for BMI decoding performance by determining the increase or decrease of BMI decoding accuracy while excluding EEG electrodes one by one from BMI decoding accuracy achieved by all EEG electrodes. This method removes the failed EEG electrodes in order.

Referring to FIG. 6, first, in performing a mental task for driving a Brain-Machine® Interface (BMI), a plurality of feature data obtained by obtaining similar feature data among all trials of brain signal measurement are obtained. In operation S210, clustering of the first electrode set including the EEG (electroencephalogram) electrodes is performed.

The second electrode set is generated by removing the EEG electrode randomly selected from the clustered first electrode set in step S210 described above (S220).

In one embodiment, the K-means clustering technique may be used as the method for removing the EEG electrode in step S220 described above.

That is, after generating a cluster of electrode sets having similar characteristics among all the EEG electrodes, the electrode having the lowest cross-validation accuracy among the EEG electrodes belonging to the electrode set may be excluded from the first electrode set. have.

By comparing the BMI decoding accuracy of the first electrode set generated in the above-described step S210 and the second electrode set generated in the above-mentioned step S220 to determine a final electrode set composed of at least one EEG electrode essential for brain signal measurement in all trials. (S230). The minimum EEG electrode constituting the final electrode set can be selected automatically.

Here, the method of comparing the BMI decoding accuracy of the first electrode set and the second electrode set in step S230 described above will be described later with reference to FIG. 8.

EEG minimum electrode selection method for BMI according to another embodiment of the present invention having the steps as described above, by determining the increase or decrease of the BMI decoding accuracy when the EEG electrode is excluded, by excluding the electrode ineffective The electrode can be selected.

In addition, the EEG electrodes having similar characteristics are collected by various clustering methods, and after excluding the lowest performing electrodes, the method of performing interpolation with the remaining electrodes is repeated, and as shown in FIG. That is, O1, Oz and O2) of FIG. 7 can be searched.

FIG. 8 is a flowchart for describing a step of determining the final electrode set of FIG. 6.

Referring to FIG. 8, in the determining of the final electrode set (S230), the BMI decoding accuracy is calculated using only the EEG electrodes of the second electrode set generated in the above-described step S220 (S231).

If the BMI decoding accuracy of the second electrode set is better than the BMI decoding accuracy of the first electrode set (No of S232), and the above-described accuracy comparison for all the EEG electrodes has not yet been completed (No of S236). ), Until the accuracy comparison is completed for all EEG electrodes (Yes in S234 or Yes in S236), the third order of EEG electrodes is further removed from the first electrode set from which the corresponding EEG electrodes are removed. While generating the electrode set (S237), the accuracy comparison step (S232) is executed again.

However, if the BMI decoding accuracy of the first electrode set is better than the BMI decoding accuracy of the second electrode set (Yes of S232), the corresponding EEG electrode is maintained in the first electrode set (S233), and all the EEG electrodes are If the above-described accuracy comparison is completed (Yes in S234), the first electrode set is immediately determined as the final electrode set.

However, if the above-described accuracy comparison for all the EEG electrodes has not yet been completed (in the case of No in S234) until the accuracy comparison is completed for all the EEG electrodes (in the case of Yes in S144 or in the case of Yes in S146). The EEG electrode of the next rank (ie, i + 1) is further removed from the first electrode set in which the EEG electrode is maintained to generate the third electrode set (S235), and the accuracy comparison step (S232) is executed again.

9 is a flowchart illustrating a method of selecting an EEG minimum electrode for BMI according to still another embodiment of the present invention.

Referring to FIG. 9, in the method for selecting an EEG minimum electrode for BMI according to still another embodiment of the present invention, each EEG constituting the electrode set determined in operation S230 of determining the final electrode set described above with reference to FIG. 6 is described. Feature data of the brain signal is obtained from the electrodes (S240).

Discriminability information of the EEG electrodes for each class in all trials is calculated using the characteristic attribute data of the brain signal acquired in the above-described step S240 and the target BMI decoding performance (S250).

The priority of the EEG electrodes is selected based on the degree of influence on the target BMI decoding accuracy using the discrimination power information calculated in step S250 (S260).

Based on the priorities of the EEG electrodes selected in step S260 described above, a final electrode set composed of at least one EEG electrode essential for brain signal measurement in all trials is determined (S270).

Herein, the contents of the above-described steps S240, S250, S260, and S270 may be equally applied to the contents of S110, S120, S130, and S140 of FIG. 2.

The method for selecting an EEG minimum electrode for BMI according to still another embodiment having the above-described steps may be performed after comparing the BMI decoding accuracy of each final electrode set determined in step S230 or step S270 described above with reference to FIG. 6. An electrode set having a high BMI decoding accuracy may be determined as the final electrode set.

The EEG minimum electrode selection method for BMI as described above may be implemented in the form of program instructions that may be implemented as an application or executed through various computer components, and recorded in a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc. alone or in combination.

The program instructions recorded on the computer-readable recording medium are those specially designed and configured for the present invention, and may be known and available to those skilled in the computer software arts.

Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tape, optical recording media such as CDROMs, DVDs, and magneto-optical media such as floptical disks. And hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like.

Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device may be configured to operate as one or more software modules to perform the process according to the invention, and vice versa.

Although described above with reference to the embodiments, those skilled in the art can be variously modified and changed within the scope of the present invention without departing from the spirit and scope of the invention described in the claims below. I can understand.

According to the present invention, it is applicable to various experimental paradigms, it is easy to select and adopt various known analysis techniques, thereby realizing versatility, easily combining existing developed techniques, and easily customizing with a templated structure. It can be easily implemented and can be developed at low cost because it can accommodate open technologies, and can be immediately applied to the currently used EEG and BMI devices, ensuring practicality, and automatically searching for the optimal electrode set. The results can be derived, and it is possible to provide convenience by actively adapting and driving each user, paradigm, and electrode arrangement environment.

Claims (10)

In performing a mental task for driving a brain-machine interface (BMI), feature data of a brain signal is obtained from a plurality of EEG (electroencephalogram) electrodes during every trial of brain signal measurement. Making;
Calculating discriminability information of the EEG electrodes for each class in all the trials by using the acquired characteristic data of the brain signal and a target BMI decoding performance;
Prioritizing the EEG electrodes according to the degree of influence on the BMI decoding accuracy using the calculated discrimination power information; And
Determining a final electrode set consisting of at least one EEG electrode essential for brain signal measurement in all said attempts based on priorities of the EEG electrodes.
The method of claim 1, wherein the calculating of the discrimination power information comprises:
In all the above trials, the highest correlation value or the lowest covariance value between the feature attribute data of the brain signal obtained from each EEG electrode and the BMI decoding performance was calculated to calculate the identification information for each electrode of each class of BMI. EEG minimum electrode selection method for BMI.
The method of claim 1, wherein the determining of the final electrode set comprises:
a set expansion step of generating a second electrode set by adding an i-th rank electrode having a high correlation value to a first electrode set including only a first rank electrode having a high correlation value;
Comparing the BMI decoding accuracy of the second electrode set and the first electrode set, if the BMI decoding accuracy of the first electrode set is better than the BMI decoding accuracy of the second electrode set, the first electrode set is replaced with the final electrode set. If the BMI decoding accuracy of the second electrode set is better than the BMI decoding accuracy of the first electrode set, an i + 1 rank electrode having the highest correlation value i + 1 is added to the second electrode set. An accuracy comparison step of generating a third set of electrodes; And
And repeating the execution of the accuracy comparison step again while sequentially adding the remaining EEG electrodes to the third electrode set in order until a final electrode set is determined.
The method of claim 1, wherein the determining of the final electrode set comprises:
a set expansion step of generating a second electrode set by adding an i-th rank electrode having the lowest covariance value to a first electrode set including only the first rank electrode having the lowest covariance value;
Comparing the BMI decoding accuracy between the second electrode set and the first electrode set, if the BMI decoding accuracy of the first electrode set is better than the BMI decoding accuracy of the second electrode set, the first electrode set is replaced with the final electrode set. If the BMI decoding accuracy of the second electrode set is better than the BMI decoding accuracy of the first electrode set, an i + 1 rank electrode having the i + 1th lower covariance value is added to the second electrode set. An accuracy comparison step of generating a third electrode set; And
And repeating the execution of the accuracy comparison step again while sequentially adding the remaining electrodes to the third electrode set in order until a final electrode set is determined.
The method of claim 1,
Clustering a first electrode set including electrodes having similar feature attribute data among the EEG electrodes constituting the final electrode set determined in the determining of the final electrode set;
Generating a second electrode set by removing a randomly selected electrode from the first electrode set; And
Comparing the BMI decoding accuracy of the first electrode set and the second electrode set to determine a final electrode set consisting of at least one EEG electrode essential for brain signal measurement in all the attempts; How to choose.
The method of claim 5, wherein the clustering step,
By using at least one of a K-means clustering technique and a similar technique of the non-supervised learning series, clusters of identification information of all the EEG electrodes are clustered into a plurality of groups to obtain similar characteristic attribute data. EEG minimum electrode selection method for BMI, which generates a configured first electrode set.
In performing a mental task for driving a brain-machine interface (BMI), a plurality of EEG (electroencephalogram) electrodes in which feature data similar to each other are obtained during every trial of brain signal measurement. Clustering the configured first electrode set;
Removing a randomly selected EEG electrode from the first electrode set to generate a second electrode set; And
Comparing the BMI decoding accuracy of the first electrode set and the second electrode set to determine a final electrode set consisting of at least one EEG electrode essential for brain signal measurement in all attempts. Way.
The method of claim 7, wherein determining the final electrode set,
A comparison accuracy calculation step of calculating a BMI decoding accuracy using only the EEG electrodes of the second electrode set;
If the BMI decoding accuracy of the second electrode set is better than the BMI decoding accuracy of the first electrode set, the corresponding EEG electrode is removed from the first electrode set, and the BMI decoding accuracy of the first electrode set is the second electrode. Determining an electrode set which maintains the corresponding EEG electrode in the first electrode set if it is better than the BMI decoding accuracy of the set; And
EEG minimum electrode for BMI, comprising repeating the step of determining the electrode set again while removing or maintaining the remaining EEG electrodes in any order from the electrode set determined in the electrode set determination step until the final electrode set is determined. How to choose.
The method of claim 7, wherein
Acquiring feature data of a brain signal from respective EEG electrodes constituting the electrode set determined in the determining of the final electrode set;
Calculating discriminability information of EEG electrodes for each class in all the trials by using the acquired characteristic data of the brain signal and the target BMI decoding performance;
Selecting priorities of the EEG electrodes according to the degree of influence on the target BMI decoding accuracy using the calculated discrimination power information; And
Determining a final electrode set consisting of at least one EEG electrode essential for brain signal measurement in all said attempts based on priorities of the EEG electrodes.
A computer-readable recording medium having a computer program recorded thereon for performing the EEG minimum electrode selection method for BMI according to any one of claims 1 to 9.

Images