KR20050011396A - Method and apparatus for removal of artifact from contaminated electroencephalogram with unexpected base line drift artifact - Google Patents

Abstract

PURPOSE: A method and an apparatus for removing artifacts from a contaminated electroencephalogram having unexpected baseline shifts are provided to remove a band-shaped portion, made by a movement of pupils, on the overall frequency range of detected electroencephalogram by compensating contaminated electroencephalogram. CONSTITUTION: An electroencephalogram measurer includes an electroencephalogram detector(110), a filter(120), a normalizer(130), a slope calculator(140), a threshold detector(150), a smoothing processor(160), a compensator(170), and an output portion(180). The filter is comprised of a wavelet filter for bandpass-filtering electroencephalogram data detected by the electroencephalogram detector at 16 to 100Hz band. The slope calculator calculates a slope of the normalized electroencephalogram data in a time domain. The threshold detector detects an abrupt change in the slope. The smoothing processor reduces the abrupt change. The compensator compensates the electroencephalogram by removing the smoothed base line shifts.

Description

Method and apparatus for removal of artifacts with sudden baseline fluctuations {Method and apparatus for removal of artifact from contaminated electroencephalogram with unexpected base line drift artifact}

The present invention relates to an EEG measuring apparatus and method, and more particularly to an EEG measuring apparatus and method for measuring the pure EEG by filtering out the harmonics.

EEG, which captures brain activity spatio-temporal, is a representative biosignal and has been widely used in clinical and brain function research. Recently, it has been used for bio-feedback, which improves the mental state of the user through the modulation of EEG by external stimulation, and with human bio signals such as electrocardiogram or galvanic skin resistance (GSR) It is also used in emotional engineering to evaluate emotion and apply it to product development. In addition, the scope of use has been extended to the field of brain-computer interface (BCI) for direct interface between humans and machines through brain waves without language or body movement.

When measuring EEG for this use, electromyogram (EGM) by human motion, electrooculogram (EOG) by electromotor movement, electrocardiogram (ECG), slow wave by respiration and sweating ) And artifacts such as signals generated by deformation of the skin by electrodes attached during EEG measurement are incorporated into the EEG. When such a wave is mixed, it becomes a problem because the brain wave of a subject cannot be measured correctly.

In the case of harmonics caused by safety, techniques have been developed to obtain safety by attaching electrodes to the top, bottom, left and right sides of the eye, and to detect and remove brain waves in which safety is mixed. For example, L. Vigon et al., In their article "Quantitative evaluation of techniques for ocular artifact filtering of EEG waveforms," extended ICA (Extended Independent Component Analysis) to eliminate eye movement and eye blink caused by eye movement, The performance of four methods, such as joint approximation diagonalisation ofeigenmatrices (JADE), principal component analysis (PCA), and EOG subtraction, were evaluated (IEE Proc.-Sci.Meas.Technol., Vol. 147, No. 5). , pp. 219-228, 2000). In order to use each of these methods, there are problems that a number of basic conditions must be met, such as a condition in which the signal source is statistically independent and Gaussian is required. Also, none of the four methods is significantly superior to the others. In addition, in the case of ICA or PCA, there is a problem that the information of the prefrontal cortex related to the higher cognitive function of humans may be missing.

A method for removing undues related to motion and eye movement can be found in US Pat. No. 5,513,649 to A. Gevins et al. In the method disclosed herein, the head movement is detected from an accelerometer attached to the head, and eye movement measurement is a safety tracking method measured from an electrode attached to the upper, lower, left, and right sides of the eye, and a gaze tracking method using a light emitting device (LED) or a camera, etc. Use them. When an EEG signal containing mixed waves is input, an adaptive filter having a coefficient adjusted according to the least square method is used to evaluate the degree of contamination caused by the interference using a reference signal and remove the noise from the EEG to correct the EEG signal. Get In order to minimize the incorporation of head and body movements by the movement of the brain, the subject is required to be careful when measuring EEG.However, in experiments in which EEG is measured while performing a certain task, eye movements are inevitable due to response to stimulation. That's how it is.

In addition, the method of removing the mixed wave of electrocardiogram is shown in US Patent No. 5,792,069 to S.D. Greenwald et al., And the method of removing the mixed wave by operation is disclosed in US Patent No. 4,550,736 to Broughton et al.

Most of the above-described method for removing blemishes has a problem in that the amplitude of the EEG varies according to the subject and does not consider that the frequency components of the EEG vary according to physiological and cognitive information processing. In addition, the sensor for detecting the motion to remove the miscellaneous, there is a disadvantage that requires an additional cost and inconvenience when measuring the EEG by attaching an electrode around the eye to measure the safety.

Therefore, the technical problem to be achieved by the present invention is to provide a brain wave measuring method that can remove the brain waves contaminated by the wave, in particular safety, etc. in consideration of individual differences, and can output only the brain wave is not mixed.

In addition, another technical problem to be achieved by the present invention is to provide an EEG measuring apparatus that can implement the above EEG measuring method.

1 is a flowchart illustrating a method for measuring brain waves according to the present invention.

2 is a block diagram schematically showing an EEG measuring apparatus according to the present invention.

3A to 3E are graphs of the EEGs according to the EEG measurement method according to the EEG method of FIG.

Figure 4 shows the spectrogram of the wavelet filtered EEG signal in accordance with the EEG measuring method of the present invention.

5 is a spectrogram showing brain waves corrected using the EEG measuring method of the present invention.

In the EEG measuring method according to the present invention for achieving the above technical problem, after detecting the EEG data of the subject, using the wavelet filter (wavelet multiresolution filter) to perform the band pass filtering (band pass filtering). The filtered EEG data is normalized so that there is no dependency of amplitude according to the subject, and then, the slope on the time axis of the normalized EEG data is obtained to find a baseline variation portion where the gradient change is greater than a threshold value. After smoothing (in other words, smoothing) the baseline variation portion, the smoothed baseline variation portion is removed from the filtered EEG data to output the corrected EEG.

In the method according to the invention, the band pass filtering step may be to pass a band of 16 ~ 100Hz. In addition, the normalizing may be performed by subtracting the mean from each data and dividing it by the standard deviation. In the finding of the baseline variation, the slope may be calculated by obtaining a difference from the current point at the next point after the current point. In addition, the smoothing may selectively smooth only a region of interest (ROI).

The EEG measuring apparatus according to the present invention for achieving the above another technical problem, the EEG detection unit for detecting the EEG data, the filter unit consisting of a wavelet filter for filtering the EEG data detected by the EEG detection unit 16 ~ 100Hz band pass, Normalizer to normalize the filtered EEG data so that the subject is not dependent on the amplitude, Slope calculator for obtaining the slope on the time axis of the normalized EEG data, Partial baseline where the gradient change is greater than the threshold A detector for finding a variation, a smoothing processor for smoothing the baseline variation, a correction unit for correcting the brain wave by removing the smoothed baseline variation from the filtered EEG data, and outputting the corrected brain wave It includes an output unit.

As described above, the EEG measuring method and apparatus according to the present invention corrects the EEG having a sudden baseline fluctuation by using the gradient change on the time axis. In addition to abrupt baseline fluctuations caused by eye movements, we can also eliminate baseline fluctuations caused by other similar causes. In particular, there is an advantage that the information of the frontal lobe does not fall out.

Hereinafter, the present invention will be described in more detail with reference to the accompanying drawings. The embodiments described below may be modified in many different forms, and the scope of the present invention is not limited to the embodiments described below. The embodiments of the present invention are provided to more completely explain the present invention to those skilled in the art.

1 is a flowchart illustrating a method for measuring brain waves according to the present invention.

First, in step 10, the EEG data of the subject is detected. Electroencephalogram data usually contains blemishes (motion, blinking eyes, sudden baseline fluctuations, etc.).

In a next step 20, the wavelet filter is used to band pass filter the EEG data. Between steps 10 and 20, a 60 Hz notch filter may be used in advance in the hardware filter to remove 60 Hz noise. The band pass filtering step is performed to remove harmonics of 100 Hz or more generated by the operation and those of 16 Hz or less generated by eye blinking or breathing. Therefore, the wavelet filter uses a band pass filter of 16 to 100 Hz.

Step 30 is a step of normalizing the signal size (electroencephalogram amplitude) is different for each subject. That is, the filtered EEG data is normalized so that there is no dependency of amplitude according to the subject. As shown in Equation 1 below, the normalized value (X _normalize ) is calculated by subtracting the mean (μ) from each of the N data (X _n ) and dividing it by the standard deviation (ρ).

In step 40, the slope on the time axis of the normalized EEG data is found to find the portion where the baseline variation is large, and then the portion where the gradient change is larger than the threshold is found. Assuming that there is a sudden baseline change of more than 3 points within 10ms, the slope can be calculated by calculating the difference from the current point at the next point after the current point, as shown in Equation 2 below.

Here, the case where the threshold value is 3 or more is a point where the degree of safety with baseline variation occurs.

In step 50, smoothing is performed at the position of safety with the baseline variation found in step 40. For example, do 5 point smoothing. In particular, only a region of interest (ROI) may be selectively smoothed.

In step 60, the smoothed baseline variation is removed from the filtered EEG data to correct the EEG and output. That is, the filtered signal of step 20 and the smoothed signal of step 50 are added or subtracted to generate a corrected brain wave.

According to the method of the present invention, there is no need to attach an electrode around the eye to measure the safety or the sensor for detecting the motion for the removal of the miscellaneous, it does not require additional cost and inconvenience when measuring EEG. Since only the part with the miscellaneous wave on the time axis is corrected and the other part is maintained, the information of the frontal lobe does not deform or fall out.

2 is a block diagram schematically showing an EEG measuring apparatus according to the present invention. The EEG measuring apparatus 100 as shown in FIG. 2 may be used to easily implement the EEG measuring method as shown in FIG. 1.

As shown in FIG. 2, the EEG measuring apparatus 100 includes an EEG detector 110 for detecting EEG data, and a wavelet filter for bandpass filtering the EEG data detected by the EEG detector 110 from 16 to 100 Hz. The filter unit 120 is included to handle steps 10 and 20 of FIG. 1. The EEG detector 110 is an electrode for measuring EEG data, which is attached to the scalp of the subject to detect EEG data. At this time, the electrode attached to the scalp of the subject may be disposed according to the 10-20 international nomenclature, or may be arranged by various other methods. The filter unit 120 may be implemented as a filter bank. That is, a high pass filter and a low pass filter may be used at the same time to separate the respective bands. If necessary, it may further include an amplifier for amplifying the EEG data before separation.

In addition, the EEG measuring apparatus 100 includes a normalizer 130 that normalizes the filtered EEG data so that the amplitude is not dependent on the subject, and is responsible for step 30 of FIG. 1. Step 40 of FIG. 1 can be performed by including a slow calculator 140 for obtaining the slope on the time axis of the normalized EEG data and a detector 150 for finding the portion of the baseline variation where the gradient change is greater than the threshold. . In addition, the smoothing processing unit 160 for smoothing the baseline fluctuation portion, the correction unit 170 for correcting the brain wave by removing the smoothed baseline fluctuation portion from the filtered EEG data and the output unit 180 for outputting the corrected brain wave It is configured to perform step 60 of Figure 1 by including. The smoothing processor 160 may selectively smooth only the portion ROI of particular interest.

3A to 3E are graphs of the EEGs according to the EEG measurement method according to the EEG measuring safety according to the sudden baseline fluctuations.

First, FIG. 3A is raw EEG data obtained in step 10 of FIG. 1. Reference numeral 220 denotes the safety portion with sudden baseline variations.

3B is a graph after performing steps 20 and 30 of FIG. 1. That is, a signal obtained through normalization is represented by using Equation 1 to extract 16 to 100 Hz using a wavelet filter and to correct a difference in power between subjects. Sudden baseline fluctuations are large, as indicated by reference numeral 230. When the signal is used for frequency analysis as it is, a band appears in the entire frequency band as shown by reference numeral 280 of FIG. 4 to generate a frequency spectrogram such as a timecourse or a topographic map. It will have a big impact on the analysis. Therefore, it is desirable to remove it.

FIG. 3C shows the threshold 240 applied to the signal using the equation (2) described in step 40 of FIG. 1 in the graph of FIG. 3B.

FIG. 3D shows the baseline 250 obtained by five point smoothing only at the abrupt baseline variation position obtained by the threshold, as in step 50 of FIG. 1.

3E is a signal obtained by subtracting the baseline 250 of FIG. 3D from the signal of FIG. 3B. Only portions corresponding to abrupt baseline fluctuations are removed and a corrected EEG is obtained.

FIG. 5 is a spectrogram displayed by using the EEG corrected using the method of the present invention, in which the reference numeral “290” is a portion in which safety of sudden baseline variation is removed. It can be seen that the portion “280” of FIG. 4 is removed.

Although the present invention has been described in detail with reference to preferred embodiments, the present invention is not limited to the above embodiments, and various modifications may be made by those skilled in the art within the scope of the technical idea of the present invention. Do. For example, the present invention applies not only to safety with sudden baseline fluctuations due to eye movement, but also to sudden baseline fluctuations caused by other causes.

As described in detail above, according to the present invention, the EEG contaminated by the safety of the abrupt baseline fluctuation caused by the movement of the eye, among the EEG into which the wave is mixed, affects the time-frequency analysis method. The advantage is the elimination of bands over the entire band.

In addition, the method of the present invention has the advantage that the information of the frontal lobe is not deformed or missed because only the part with the miscellaneous wave on the time axis is corrected and the other part is kept intact.

In addition, there is an advantage that the existing EEG can be used as it is without motion or eye movement sensor. There is no need for additional cost and inconvenience in measuring EEG because no sensor is attached around the eye to measure safety or sensor to detect motion for elimination of harmonics.

For example, when measuring EEG while performing a task with visual stimulation, it is possible to correct an EEG in which a hybrid wave with a large baseline variation due to eye movements such as nystagmus is incorporated. In frequency analysis, the nystagmus band (impulse form on the time axis) that appears over the entire band of the spectrum can be removed.

Claims (8)

Detecting EEG data of the subject; Band pass filtering the EEG data using a wavelet multiresolution filter; Normalizing the filtered EEG data such that there is no dependency of amplitude on a subject; Finding a baseline variation portion whose slope change is greater than a threshold by obtaining a slope on the time axis of the normalized EEG data; Smoothing the baseline variation portion; And And removing the smoothed baseline variation from the filtered EEG data to output a corrected EEG. The method of claim 1, wherein the band pass filtering step passes a band of 16 to 100 Hz. The method of claim 1, wherein the normalizing is performed by subtracting an average from each data and dividing it by a standard deviation. The EEG measuring method of claim 1, wherein the finding of the baseline variation comprises calculating a slope by obtaining a difference from a current point at a next point after a current point. The method of claim 1, wherein the smoothing comprises selectively smoothing only a region of interest (ROI). An EEG detector for detecting EEG data; A filter unit comprising a wavelet filter for performing 16 to 100 Hz band pass filtering on the EEG data detected by the EEG detector; A normalizer for normalizing the filtered EEG data so that there is no dependency of amplitude on a subject; A slope calculator for obtaining a slope on a time axis of the normalized EEG data; A threshold detector for finding a portion of the baseline variation where the gradient change is greater than the threshold; A smoothing processor for smoothing the baseline variation portion; A correction unit correcting the EEG by removing the smoothed baseline variation from the filtered EEG data; And EEG measuring apparatus comprising an output unit for outputting the corrected brain waves. The EEG measuring apparatus of claim 6, wherein the slow calculator calculates a slope by obtaining a difference from a current point at a next point after the current point. The EEG measuring apparatus of claim 6, wherein the smoothing processor selectively smoothes only a portion of interest.