KR20150029969A - Sensibility classification method using brain wave - Google Patents

Abstract

The purpose of the present invention is to provide an emotion evaluation algorithm, which classifies four emotional states, such as relaxation, joy, sadness, and anger, to evaluate human emotions more accurately using brainwaves. A method for classifying emotional states using brainwave measurement is disclosed. The method for classifying emotional states comprises the steps of: extracting feature parameters indicating emotion from a user′s brainwave data measured from an electroencephalogram (EEG) measurement device; and classifying the user′s emotional states by classifying patterns of the brainwave data using the extracted feature parameters.

Description

[0001] SENSITIVITY CLASSIFICATION METHOD USING BRAIN WAVE [0002]

Embodiments of the present invention relate to techniques for classifying human emotions using brain waves.

Emotional engineering refers to the measurement, evaluation, and calculation of human emotions in a quantitative manner, and it is necessary to quantitatively / qualitatively analyze emotions in addition to the understanding of the essence of the biological system, since the subject is human emotions. It is also important to accurately analyze the many emotional stimuli felt by humans and the physiological signals that emerge from them. However, many researchers at home and abroad are presenting the results of this research, but they have not been able to present a clear physiological index that can accurately express human emotion. This is not a problem that can be easily solved in consideration of the complexity and nonlinearity of the biological system.

The evaluation method of human emotion can be roughly divided into an objective emotional evaluation method of collecting bio-signals and analyzing and evaluating them, and a subjective evaluation method of evaluating with a questionnaire. In order to obtain a more accurate evaluation result, There are also cases where the method is used together.

In studying human emotions, we use brain waves more than signals by autonomic nervous system, because brain waves are known to contain the most information about brain activity.

For example, Korean Patent Laid-Open No. 10-2007-0087787 discloses a technique of classifying human emotional states in real time based on biological signals such as human brain waves, electrocardiogram, blood pressure, skin conductivity, respiration, and the like.

However, brain waves are very complicated due to various brain activities, and the characteristics of signals can be drastically changed depending on various conditions such as health, mood, and environment even in the same person as well as individual differences according to personality and sex. Therefore, it is difficult to accurately obtain the desired information from such a complex EEG signal, and this is a great obstacle in research on emotion engineering using EEG.

Disclosure of Invention Technical Problem [8] The present invention has been made in order to solve the above problems, and it is an object of the present invention to provide a method for more accurately evaluating emotions of a human using EEG, including relaxation, joy, sadness, And anger are classified into four emotion states.

According to an embodiment of the present invention, there is provided a method of classifying emotion states, comprising: extracting a feature parameter representing emotion from user's brain wave data measured from an EEG (electroencephalogram) measuring apparatus; And classifying the emotional state of the user by classifying the pattern of the brain wave data using the feature parameter.

According to one aspect, the EEG data can be measured using 16 channels of electrodes.

According to another aspect, the electrode may be attached to 16 portions of the scalp according to the ten-twenty electrode system of the international standard.

According to another aspect, the emotion state classification method may further include a preprocessing step of removing noise included in the brain wave data before extracting the feature parameter.

According to another aspect, the emotion state classification method may further include a preprocessing step of removing a noise signal due to blinking in the EEG data before extracting the feature parameter.

According to another aspect of the present invention, the pre-processing includes: setting a threshold for distinguishing the noise signal; Detecting a peak value of an eye blinking portion by using an inclination at each point after changing the brain wave data to an absolute value; And determining the noise signal as the noise signal if the peak value is greater than the threshold value.

According to another aspect of the present invention, the step of setting the threshold value may set a value obtained by calculating an average value of the total length of the brain wave data and then multiplying the average value by a certain multiple.

According to another aspect, the emotional state classification method may further include a preprocessing step of filtering a frequency component including information related to emotion in the EEG data before extracting the feature parameter.

According to another aspect of the present invention, the preprocessing step includes the steps of removing a DC offset from the EEG data using a high pass filter, generating a high frequency signal from the EEG data using a low pass filter, Noise can be removed.

According to another aspect of the present invention, the extracting of the feature parameter may extract a linear predictor coefficient of the brain wave data as the feature parameter through a linear predictive analysis.

According to another aspect, the number of linear prediction machines can be obtained by applying a prediction error between samples predicted by linear combination of previous signals to a current signal to a least squares criterion.

According to another aspect, the prediction error can be calculated using Equation (1).

Equation 1:

은 예측된 샘플로서

(이때, -a_i(i=1, 2, …, M)가 선형예측기계수를 나타낸다.)와 같다.)(Where n (n) represents a prediction error, s (n) represents a current signal sample,

Lt; RTI ID = 0.0 >

(Here, -a _i (i = 1, 2, ..., M) represents the number of linear prediction machines.)

According to another aspect of the present invention, the step of extracting the feature parameter may extract a cross-correlation coefficient for each band of the brain wave data as the feature parameter.

According to another aspect, the cross-correlation coefficient may be obtained by Equation (2).

Equation 2:

Where C (a: jk) represents the cross-correlation coefficient between j and k electrodes in a frequency band, and X _j (f _n ) represents the frequency spectrum at frequency n at j th electrode.

According to another aspect of the present invention, the step of extracting the characteristic parameter comprises the steps of extracting the feature parameter from the frequency spectrum of an FFT (Fast Fourier Transform) spectrum or an AR (Auto-Regressive) Can be extracted for each band.

According to another aspect of the present invention, the classifying the emotional state of the user may classify the pattern of the brain wave data using a multilayer perceptron composed of the feature parameters.

According to another aspect, the step of classifying the emotional state of the user includes: selecting a template of an EEG wave closest to the user's brain wave data among weight templates configured by learning EEG on emotions of a plurality of users; ; And classifying the pattern of the EEG data by analyzing the outputs of the multi-layer neural network after reading the weighted values of the selected template into the multi-layer neural network.

According to another aspect of the present invention, the step of classifying the patterns of the EEG data by analyzing the outputs of the multi-layer neural network may include calculating the emotional state of the user with the largest value and the second largest value among the outputs of the multi- It can be judged.

According to the embodiment of the present invention, the emotional state classifying apparatus comprises: an extracting unit for extracting a feature parameter indicating emotion from user's brain wave data measured from an EEG (electroencephalogram) measuring apparatus; And a classifier for classifying the emotional state of the user according to classifying the pattern of the brain wave data using the feature parameter.

According to the embodiment of the present invention, it is easy to classify and analyze brain waves by extracting characteristic parameters of brain waves, and the output of the neural network is regarded as an indicator for each emotion, and the results are classified into negative emotion and positive emotion, It is possible to perform the most appropriate sensitivity evaluation.

According to the embodiment of the present invention, there is an advantage that it can be applied to the sensitivity evaluation for an unspecified number of people.

1 is a flowchart illustrating an emotional state classification method using EEG measurement according to an exemplary embodiment of the present invention.
2 illustrates an electrode attachment position for EEG measurement according to an embodiment of the present invention.
3 is a block diagram illustrating a preprocessing process for an EEG signal according to an embodiment of the present invention.
FIG. 4 shows an EEG waveform before (upper) and after (lower) eye blink removal in an embodiment of the present invention. FIG.
5 is a diagram for explaining a process of evaluating emotion using a multi-layer neural network in an embodiment of the present invention.
6 is a block diagram illustrating an internal configuration of an emotion state classifying apparatus using EEG measurement according to an embodiment of the present invention.

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

The present invention relates to a method of classifying emotional states using EEG, and more particularly, it is possible to classify emotions using a multi-layer neural network through acquiring brain waves, pre-processing for blinking and extracting feature parameters.

1 is a flowchart illustrating an emotional state classification method using EEG measurement according to an exemplary embodiment of the present invention.

The emotional state classification method according to an exemplary embodiment is a method for effectively acquiring an emotion state of a user through a preprocessing process and a feature extraction pattern classification by acquiring an EEG signal, May be performed.

Referring to FIG. 1, the emotional state classification method according to an exemplary embodiment includes an EEG signal acquisition step S100, a preprocessing step S200, a feature extraction step S300, a selection of well-matched template step S400, a weight templates step S500, and a sensibility evaluation step S600.

First, in the brain wave acquisition step (S100), the emotion state classification apparatus can acquire an EEG signal, which is an EEG signal measured from the EEG measurement apparatus. At this time, the emotion state classification apparatus can acquire the user's brain wave signal measured through the 16-channel electrode.

2 illustrates an electrode attachment position for EEG measurement according to an embodiment of the present invention.

The emotional state classifier attaches an electrode to the brain of the user to acquire EEG data measured through the electrode. At this time, the electrode attaching position for measuring EEG is 10-20 electrode system (Fp1, Fp2, F3, F4, C3, C4, P3, P4, O1, O2, F7, F8, T3, T4, T5 and T6) of the scalp.

Referring again to FIG. 1, in the pre-processing step S200, the emotional state classifier may remove noise included in the brain wave signal for purifying the brain wave signal acquired in the brain wave acquisition step S100.

The EEG signal is a weak signal of a few tens of microvolts, very sensitive to external noise and artifacts, and usually contains a lot of noise. Therefore, the EEG signal can be analyzed only after it is purified through a predetermined pretreatment process.

Table 1 below shows frequency values and conscious state according to each EEG.

Electroencephalogram Frequency range Characteristic delta wave 0 to 4 Hz Sleep state θ (Theta) wave 4 to 8 Hz Sleepy or meditative alpha (alpha) wave 8-12Hz Relaxed state SMR wave 12 to 15 Hz Attention state Mid-β wave 15-20 Hz Activity status High-β wave 20-30Hz Stress excited or stressed γ (Gamma) wave 30 to 50 Hz Anxiety, excitement, stressful state

3 is a block diagram illustrating a preprocessing process for an EEG signal according to an embodiment of the present invention.

Prior to the parameter extracting step S300, the preprocessing step S200 includes a blinking removal process S201 and a digital filtering process S202 as a preprocessing procedure, as shown in FIG.

Eye Flicker Removal Process (S201):

(1) In order to remove eye flicker, which is the largest artifact in an EEG signal, a threshold is set to discriminate a blink signal from a pure EEG. At this time, the threshold value can be determined by calculating its average over the entire length of the read data, and then multiplying the average value by 4 times.

(2) First, to find the blinking part of the eye, the peak value of the blinking part of the eye is detected. To do this, change the read data to an absolute value, and then search the slope code at each point to find a slope at two points backward with respect to the current point +, and a slope at two points ahead of the current point.

(3) The peak value is compared with the threshold value to distinguish whether the peak of the point is pure EEG or a peak of the eye flicker signal. If the peak value at that point is greater than the threshold value, it is regarded as an eye flicker signal. At this time, to find the point where the blink signal ends, find the point where the sign changes twice from the blink peak point. This position can be seen as the end of the blinking of the eye, but considering the influence of the remaining components on the back, we can set the point where we skipped a certain section as the starting point for new EEG processing.

FIG. 4 shows the waveforms before and after the blinking elimination with respect to the EEG signal. As shown in FIG. 4, a pure EEG waveform obtained by removing an eye blink signal from an EEG signal can be obtained through the eye flicker removal process (S201).

Digital filtering process (S202):

The emotion state classification apparatus filters the frequency components below 30 Hz to extract specific parameters from the EEG signal since the information related to emotion in the EEG is included in the components below 30 Hz. To remove the DC offset, a high pass filter (HPF) with a cutoff frequency of 4 Hz was constructed in the form of a quadratic IIR filter (infinite impulse response filter). In order to remove high frequency noise, Pass filter (LPF) of the 4th-order IIR filter, it is possible to filter only the frequency components of 4 to 30 Hz.

As described above, the process for extracting the feature parameters (S300) is performed using the pre-processed EEG signals of the eye flicker removing process S201 and the digital filtering process S202.

Referring again to FIG. 1, in the parameter extracting step (S300), the emotional state classifier can extract a feature parameter indicating emotion from a user's brain wave signal.

For example, the emotion state classifier can extract a linear predictor coefficient of an EEG signal as a parameter of emotion.

FFT (Fast Fourier Transform) is used in many fields of signal processing, and it is also used as a main analysis means in the case of EEG signals, and also in research on emotion or emotion. However, the influence of side lobe and spectral leakage is large in FFT, which may be a disadvantage in fine analysis such as emotional research.

In this embodiment, a method of analyzing a signal for expressing a signal characteristic also includes a method in which a signal transmission process such as a linear predictive analysis is regarded as one system, and a linear prediction The number of machines is extracted and used as characteristics.

이라 하면, 예측 오차(prediction error) e(n)은 수학식 1과 같이 정의할 수 있다.The number of linear prediction machines is obtained by linear prediction analysis applying the least squares criterion to the error between the predicted samples, in which the current signal is represented by a linear combination of previous signals. The current signal sample is denoted by s (n), the predicted sample is denoted by s

, The prediction error e (n) can be defined as Equation (1).

은 수학식 2와 같다.Here,

(2) "

Here, -a _i (i = 1, 2, ..., M) is the number of linear prediction machines.

The 6th order Burg algorithm can be used for linear prediction analysis. The linear predictor divisor can be directly used as a specific parameter or, as another example, the cross-correlation coefficient of the band of the AR (Auto-Regressive) spectrum can be extracted and used as a characteristic parameter of the emotion. At this time, the AR spectrum S _AR (?) Is obtained by the equation (3).

The cross-correlation coefficient of each band of EEG is expressed by Equation (4) as an example of the cross-correlation coefficient between j and k electrodes in a band.

Here, X _j (f _n ) represents the frequency spectrum at the frequency n at the j-th electrode. The cross-correlation coefficients to be used as the feature parameters are obtained by extracting the cross-correlation coefficients of the θ-wave (5-8 Hz), α-wave (8-13 Hz) and β-wave (13-20 Hz) among the frequency bands of the FFT spectrum and the AR spectrum can do.

Referring again to FIG. 1, in the pattern classification step S400, the emotional state classification apparatus can classify the emotion of an EEG signal using the number of linear prediction machines or the cross correlation coefficient extracted in the parameter extraction step S300.

In this embodiment, brain wave characteristic analysis and parameter extraction method are implemented for the emotion evaluation using four emotions such as relaxation, joy, sadness, and anger. And then compare the discrimination power with the pattern classification, and then use the emotion evaluation technique to classify it as positive emotion and negative emotion.

The approach for pattern classification may be a similarity measure, a distance measure, a perceptron, a fuzzy theory, etc. In this embodiment, the multi- A multilayer perceptron can be used.

5 is a diagram for explaining a process of evaluating emotion using a multi-layer neural network in an embodiment of the present invention.

The number of nodes in the input layer of the neural network depends on the feature parameter extraction method. When using the number of linear prediction machines, 6 (difference) × 10 (channel) = 60, and when using the cross-correlation coefficient for each frequency band, three EEG bands (θ, α , β), so 10C2 × 3 (bands) = 135. The number of nodes in the hidden layer is 60 in the case of the number of linear prediction machines and 100 in the case of the cross correlation coefficient, and the output layer has four nodes. The network thus constructed can be trained by a general error backpropagation learning algorithm.

The pattern of the EEG is almost impossible to be recognized by only one reference template because the characteristics of the EEG varies according to various environments and psychological states such as the viewpoint, mood, and temperature, as well as individual differences. Therefore, we propose an emotional recognition method that constructs EEG templates of many people and classifies the patterns by reading the template that is closest to the input test data.

For this, in the weighting step (S500), the emotional state classifier may construct a personality group template according to weight templates for several human brain waves. For example, the emotional state classifier may store the weights of neural networks generated by learning EEG on four emotions of a plurality of subjects in a template database.

In the pattern classification step (S400), the most similar EEG template is selected from the template database using the stable frontal EEG, and is read as a weight to the neural network, and the output of the neural network is analyzed Patterns can be classified. However, this method selects a weight template of the nearest EEG based on the earliest steady state EEG, so that a bias phenomenon may occur due to a stable EEG.

In order to compensate for this, the emotion state classifier in the final stage of sensibility evaluation (S600) can determine the final emotion classification result with the largest value and the second largest value of the neural network output. In other words, when testing a stable EEG, it is judged that the largest value appears at the corresponding node, and other emotions (joy, sadness, anger) are recognized only when the corresponding node exhibits the largest value However, it can be judged that the second highest value appears in the node of the emotion, and the node with the highest value at this time is recognized as stable node.

Therefore, the emotional state classification method according to the present invention extracts the number of linear prediction machines of EEG signals and uses them as feature parameters of the emotion, and emotion is evaluated through the emotion evaluation algorithm using the multi-layer neural network and the multi- Can be classified.

The emotional state classification method according to the present invention may include at least two operations based on the details described with reference to Figs. At this time, two or more operations can be combined, and the order or position of operations can be changed.

The methods according to embodiments of the present invention may be implemented in the form of a program instruction that can be executed through various computer systems and recorded in a computer-readable medium. In particular, in this embodiment, extracting a feature parameter indicating emotion from user's brain wave data measured from an EEG (electroencephalogram) measuring device; And classifying the emotional state of the user according to classifying the pattern of the brain wave data using the feature parameter.

6 is a block diagram illustrating an internal configuration of an emotion state classifying apparatus using EEG measurement according to an embodiment of the present invention. As shown in FIG. 6, the emotion state classifying apparatus according to one embodiment may include a preprocessing unit 610, an extracting unit 620, and a classifying unit 630. The emotion state classifying apparatus according to one embodiment may further include components or may further include additional components based on the details of the emotion state classifying method described with reference to Figs. 1 to 5.

The preprocessor 610 may remove the noise included in the EEG signal to refine the EEG signal, which is the EEG signal measured from the EEG measurement device. At this time, the EEG signal can be acquired through the electrodes of 16 channels (Fp1, Fp2, F3, F4, C3, C4, P3, P4, O1, O2, F7, F8, T3, T4, T5 and T6).

For example, the preprocessor 610 may remove the blinking signal, which is the largest artifact in the EEG signal. At this time, the preprocessing unit 610 changes the read EEG data to an absolute value, and then searches for the slope code at each point, and calculates a slope at two points backward from the current point, The peak value of the eye blinking portion can be detected by comparing the peak value of the eye blinking portion with the predetermined threshold value to distinguish whether the peak of the blinking portion is a pure EEG peak or a blinking eye peak, If the peak value at that point is greater than the threshold value, it is regarded as an eye flicker signal.

As another example, the preprocessing unit 610 may filter the frequency components of 30 Hz or less through digital filtering to extract a specific parameter from an EEG signal. For this purpose, a high pass filter (HPF) with a cutoff frequency of 4 Hz is constructed in the form of a quadratic IIR filter (infinite impulse response filter) to remove the DC offset, By configuring a low pass filter (LPF) with a cut-off frequency of 30 Hz as a fourth-order IIR filter, it is possible to filter only frequency components in the 4 to 30 Hz range.

The extracting unit 620 may extract a feature parameter indicating emotion from the refined EEG signal. For example, the extractor 620 can extract the number of linear prediction machines of EEG as a feature parameter of emotion through linear prediction analysis. The number of linear prediction machines is obtained by linear prediction analysis applying the least squares criterion to the error between the predicted samples, in which the current signal is represented by a linear combination of previous signals. As another example, the extractor 620 may extract the cross-correlation coefficient of each EEG band as a feature parameter of emotion. At this time, the extracting unit 620 may extract the cross-correlation coefficients of the bands corresponding to the? Wave and the? And? Waves among the frequency bands of the FFT spectrum or the AR spectrum, respectively, and use the extracted cross-correlation coefficients as the characteristic parameters.

The classifier 630 can classify the emotion of an EEG signal based on a multilayer neural network using a linear prediction machine number or a cross correlation coefficient. In this case, the input layer of the neural network becomes 6 (difference) × 10 (channel) = 60 when the number of linear prediction machines is used, and when using the cross-correlation coefficient by frequency band, (Bands) = 135 because of the use of the EEG bands ([theta], [alpha], and [beta]). The nodes of the hidden layer can use 60 nodes in the case of the number of linear prediction machines and 100 nodes in the case of the cross correlation coefficients, and the output layer has four nodes according to the sensitivity to be classified.

The classification unit 630 may store the weights of neural networks generated by learning brain waves of four emotions of several subjects to construct a personality group template for the brain waves of various persons in the template database. Then, the classification unit 630 selects the most similar EEG template in the template database using the stable frontal EEG when the emotion evaluation test is performed, and reads the weighted value as a weight to the neural network, Can be classified.

In order to compensate the bias phenomenon of the stable state EEG when selecting the weight template of the closest EEG based on the earliest stable EEG based on the earliest stable EEG, the classifier 630 calculates the maximum value and the second largest value of the neural network output, The sensitivity classification result can be determined. In other words, when testing a stable EEG, it is judged that the largest value appears at the corresponding node, and other emotions (joy, sadness, anger) are recognized only when the corresponding node exhibits the largest value However, it can be judged that the second highest value appears in the node of the emotion, and the node with the highest value at this time is recognized as stable node.

According to the above configuration, the emotional state classifying apparatus using EEG can acquire EEG signals and can effectively grasp the emotional state of the user based on the multi-layer neural network and the multiple templates through the preprocessing process and feature extraction pattern classification.

As described above, according to the embodiment of the present invention, it is easy to classify and analyze brain waves by extracting the characteristic parameters of the EEG, and the output of the neural network is regarded as an indicator for each emotion, and the results thereof are classified into the negative emotion and the positive emotion It has the advantage of being able to perform emotional evaluation which is best suited to the characteristics of EEG. Further, according to the embodiment of the present invention, there is an advantage that it can be applied to the sensitivity evaluation for an unspecified number of people.

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

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

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

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

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

610:
620:
630:

Claims (29)

Extracting a feature parameter indicating emotion from user's brain wave data measured from an EEG (electroencephalogram) measuring device; And
Classifying the emotional state of the user according to classifying the pattern of the brain wave data using the feature parameter
Wherein the emotional state classification method comprises: The method according to claim 1,
The EEG data are measured using electrodes of 16 channels
Wherein the emotional state classification method comprises the steps of: 3. The method of claim 2,
The electrode is attached to 16 portions of the scalp according to the international standard 10-20 method (ten-twenty electrode system)
Wherein the emotional state classification method comprises the steps of: The method according to claim 1,
A pre-processing step of removing noise included in the brain wave data before extracting the feature parameter
Further comprising the steps of: The method according to claim 1,
A pre-processing step of removing a noise signal due to blinking in the brain wave data before extracting the feature parameter
Further comprising the steps of: 6. The method of claim 5,
The pre-
Setting a threshold for distinguishing the noise signal;
Detecting a peak value of an eye blinking portion by using an inclination at each point after changing the brain wave data to an absolute value;
Determining the noise signal as the noise signal if the peak value is greater than the threshold value
Wherein the emotional state classification method comprises: The method according to claim 6,
Wherein the step of setting the threshold value comprises:
Calculating a mean value of the total length of the EEG data, and then multiplying the mean value by a certain multiple to set the threshold value
Wherein the emotional state classification method comprises the steps of: The method according to claim 1,
A pre-processing step of filtering a frequency component including information related to emotion in the EEG data before extracting the feature parameter,
Further comprising the steps of: 9. The method of claim 8,
The pre-
Removing high frequency noise from the brain wave data by removing a DC offset from the EEG data using a high pass filter and using a low pass filter
Wherein the emotional state classification method comprises the steps of: The method according to claim 1,
Wherein the extracting of the feature parameter comprises:
Extracting a linear predictor coefficient of the brain wave data as the feature parameter through a linear predictive analysis
Wherein the emotional state classification method comprises the steps of: 11. The method of claim 10,
The number of linear prediction machines is obtained by applying a prediction error between samples predicted by the linear combination of previous signals to a current signal to a least squares criterion
Wherein the emotional state classification method comprises the steps of: 12. The method of claim 11,
The prediction error is obtained by using Equation (1)
Wherein the emotional state classification method comprises the steps of:
Equation 1:

(Where n (n) represents a prediction error, s (n) represents a current signal sample,

Lt; RTI ID = 0.0 >

(Here, -a _i (i = 1, 2, ..., M) represents the number of linear prediction machines.) The method according to claim 1,
Wherein the extracting of the feature parameter comprises:
Extracting a cross-correlation coefficient for each band of the EEG data as the feature parameter
Wherein the emotional state classification method comprises the steps of: 14. The method of claim 13,
The cross-correlation coefficient is calculated by Equation (2)
Wherein the emotional state classification method comprises the steps of:
Equation 2:

Where C (a: jk) represents the cross-correlation coefficient between j and k electrodes in a frequency band, and X _j (f _n ) represents the frequency spectrum at frequency n at j th electrode. 14. The method of claim 13,
Wherein the extracting of the feature parameter comprises:
(Theta) wave and the alpha (alpha) and beta (beta) waves among the frequency bands of the Fast Fourier Transform (FFT) spectrum or the AR (Auto-Regressive) spectrum
Wherein the emotional state classification method comprises the steps of: The method according to claim 1,
Wherein classifying the emotional state of the user comprises:
Classifying the pattern of the EEG data using a multilayer perceptron composed of the feature parameters
Wherein the emotional state classification method comprises the steps of: 17. The method of claim 16,
Wherein classifying the emotional state of the user comprises:
Selecting a template of an EEG wave closest to the user's brain wave data among weight templates configured by learning an EEG wave for sensitivity of several persons; And
Classifying the pattern of the EEG data by analyzing the output of the multi-layer neural network after reading the weight of the selected template into the multi-layer neural network
Wherein the emotional state classification method comprises: 18. The method of claim 17,
Wherein the classifying the pattern of the EEG data by analyzing the outputs of the multi-
Determining the emotional state of the user with the largest value and the second largest value among the outputs of the multi-layer neural network
Wherein the emotional state classification method comprises the steps of: 15. A computer readable medium comprising instructions for controlling a computer system to classify emotional states through brain wave measurements,
The command includes:
Extracting a feature parameter indicating emotion from user's brain wave data measured from an EEG (electroencephalogram) measuring device; And
Classifying the emotional state of the user according to classifying the pattern of the brain wave data using the feature parameter
≪ / RTI > wherein said computer system is controllable by said computer system. An extracting unit for extracting a feature parameter indicating emotion from user's brain wave data measured from an EEG (electroencephalogram) measuring apparatus; And
And a classifying unit for classifying the emotional state of the user according to classifying the pattern of the brain wave data using the feature parameter,
And an emotional state classifying device. 21. The method of claim 20,
Processing unit for removing noise included in the brain wave data before extracting the feature parameter,
Further comprising: 21. The method of claim 20,
Further comprising a preprocessing unit for removing a noise signal due to blinking in the brain wave data before extracting the feature parameter,
The pre-
A threshold value for discriminating the noise signal is set, a peak value of an eye blinking part is detected using an inclination at each point after changing the brain wave data to an absolute value, and the peak value If the noise signal is larger than the threshold value, it is determined to be the noise signal
Wherein the emotional state classifying device comprises: 21. The method of claim 20,
Further comprising a preprocessor for filtering a frequency component including information related to emotion in the brain wave data before extracting the feature parameter,
The pre-
Removing high frequency noise from the brain wave data by removing a DC offset from the EEG data using a high pass filter and using a low pass filter
Wherein the emotional state classifying device comprises: 21. The method of claim 20,
The extracting unit extracts,
Extracting a linear predictor coefficient of the brain wave data as the feature parameter through a linear predictive analysis
Wherein the emotional state classifying device comprises: 25. The method of claim 24,
The number of linear prediction machines is obtained by applying a prediction error between samples predicted by the linear combination of previous signals to a current signal to a least squares criterion
Wherein the emotional state classifying device comprises: 21. The method of claim 20,
The extracting unit extracts,
Extracting a cross-correlation coefficient for each band of the EEG data as the feature parameter
Wherein the emotional state classifying device comprises: 27. The method of claim 26,
The extracting unit extracts,
(Theta) wave and the alpha (alpha) and beta (beta) waves among the frequency bands of the Fast Fourier Transform (FFT) spectrum or the AR (Auto-Regressive) spectrum
Wherein the emotional state classifying device comprises: 21. The method of claim 20,
Wherein,
Classifying the pattern of the EEG data using a multilayer perceptron composed of the feature parameters
Wherein the emotional state classifying device comprises: 29. The method of claim 28,
Wherein,
Selecting a template of an EEG wave closest to the user's brain wave data among weight templates configured by learning brain waves of emotions of various people, reading the brain wave data with weight of the selected template into the multi-layer neural network, Classifying the pattern of the EEG data by analyzing the output of the network
Wherein the emotional state classifying device comprises:

Images