CN106236080A - Based on the removing method of myoelectricity noise in multichannel EEG signals - Google Patents
Based on the removing method of myoelectricity noise in multichannel EEG signals Download PDFInfo
- Publication number
- CN106236080A CN106236080A CN201610692634.7A CN201610692634A CN106236080A CN 106236080 A CN106236080 A CN 106236080A CN 201610692634 A CN201610692634 A CN 201610692634A CN 106236080 A CN106236080 A CN 106236080A
- Authority
- CN
- China
- Prior art keywords
- eeg signals
- intrinsic
- myoelectricity
- noise
- modal components
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/40—Detecting, measuring or recording for evaluating the nervous system
- A61B5/4076—Diagnosing or monitoring particular conditions of the nervous system
- A61B5/4094—Diagnosing or monitoring seizure diseases, e.g. epilepsy
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Animal Behavior & Ethology (AREA)
- Veterinary Medicine (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Surgery (AREA)
- Physics & Mathematics (AREA)
- Neurology (AREA)
- Neurosurgery (AREA)
- Physiology (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a kind of based on the removing method of myoelectricity noise in multichannel EEG signals, its feature includes: 1, first decomposes each passage EEG signals with population mean empirical modal, obtains the intrinsic modal components of each passage;2, judged the intrinsic modal components of Noise by autocorrelation coefficient, be made up of noisy intrinsic mode Component Matrices the intrinsic modal components of Noise;3, noisy intrinsic mode Component Matrices is carried out Blind Signal Separation;4, the exemplary component of Noise, zero setting noise component(s) reconstruction signal is judged with autocorrelation coefficient.The present invention not only achieves the purpose removing myoelectricity, remains the composition of doubtful brain electricity in each intrinsic modal components, it is ensured that brain electric information is not lost in processing procedure simultaneously.
Description
Technical field
The invention belongs to EEG Processing technical field, be specifically related to a kind of based on population mean empirical mode decomposition and
Canonical correlation analysis, automatically identifies myoelectricity noise the method eliminated from the multichannel brain signal of telecommunication, is mainly used in human brain phase
Related disorders and the research of human brain function.
Background technology
EEG signals is that the bioelectrical activity of cranial nerve cell is in cerebral cortex or the overall reflection of scalp surface.Electroencephalogram
As the equipment of record EEG signals, the same with electrocardiogram, X-ray examination, become clinical important detection methods.But
EEG signals, as faint electricity physiological signal, is often subject to such as the interference of the multiple noises such as electrocardio, eye electricity and myoelectricity, after impact
The continuous accuracy to brain electricity analytical.The myoelectricity interference caused due to musculation (such as sting, chew and frown), at eeg signal acquisition
Process is difficult to avoid that, and its amplitude is big, frequency domain distribution wide, and causing myoelectricity noise is the interference source being most difficult to eliminate.
Myoelectricity interference removed by past usual low pass filter.But, if myoelectricity interference and EEG signals interested
Spectrum overlapping, frequency filter not only can suppress myoelectricity to disturb, and may filter valuable brain electric information.
After serve scholar's proposition independent component analysis and eliminate problem to solve in brain electricity myoelectricity noise.Independent element divides
Analysis (ICA) utilizes independence that EEG signals resolves into the component of mutual statistical independence.Independent component analysis is typically by artificial
Observe and judge whether isolated component comprises myoelectricity, rebuild after these components containing myoelectricity of zero setting and obtain clean brain telecommunications
Number.Experiment proves that independent component analysis has good effect at removal eye electricity and electrocardio noise, but is removing myoelectricity noise
Effect be not highly desirable, this is because the most of isolated component obtained by independent component analysis had both comprised brain electricity also
Comprise myoelectricity.
To this end, some scholars propose with canonical correlation analysis (Clercq WD, VergultA, Vanrumste B, Van
Paesschen W,Van Huffel S,Canonical correlation analysis applied to remove
muscle artifacts from the electroencephalogram.IEEE Transactions on
Biomedical Engineering, 2006,53 (12): 2583-2587.) solve myoelectricity noise elimination problem in brain electricity.Allusion quotation
Type correlation analysis utilizes autocorrelation that EEG signals resolves into orthogonal component, by asking for the auto-correlation system of component
Number, if less than setting threshold value, then these components are determined is myoelectricity noise, rebuilds and obtain clean brain electricity after these components of zero setting
Signal.Owing to the characteristic of myoelectricity noise is similar with white noise, therefore myoelectricity noise is compared EEG signals and is had relatively low from phase
Guan Xing, canonical correlation analysis can concentrate in minority canonical variable myoelectricity noise, by arranging autocorrelation coefficient threshold value, permissible
Realize removing the purpose of myoelectricity interference.Experiment proves that canonical correlation analysis has than independent component analysis and preferably removes myoelectricity noise
Effect.But, after EEG signals is resolved into multiple sources component by the method such as independent component analysis and canonical correlation analysis, it is determined
For still comprising brain electricity composition in the source of myoelectricity component, although removing these source components can reach the purpose of denoising, but not
Part brain electric information can be lost with avoiding.
In the recent period, scholar is had to propose a kind of based on population mean empirical mode decomposition with the novel manifold of independent component analysis
Road denoising method (KeZeng, DanChen, GaoxiangOuyang, LizheWang, XianzengLiu, XiaoliLi,
AnEEMD-ICA approach to enhancing artifact rejection for noisy multivariate
neural data.IEEE transactions on NeuralSystemsandRehabilitationEngineering,
2016,24(6):630-638.).Experiment proof the method is in terms of the multichannel brain signal of telecommunication removes myoelectricity noise, than independent one-tenth
Analysis is divided to have more preferable denoising effect.But, the method carries out blind source separating owing to have employed independent component analysis, and it obtains
Isolated component in usually contain both myoelectricity and brain electricity information, by these component zero setting will inevitably lost part brain electricity
Information.
Summary of the invention
In place of the present invention is in order to overcome the deficiencies in the prior art, proposes one and make an uproar based on myoelectricity in multichannel EEG signals
The removing method of sound, to removing the impact on EEG signals of the myoelectricity noise, and retains doubtful brain in each intrinsic modal components
The information of electricity composition is not lost, thus improves the accuracy of electroencephalogramsignal signal analyzing.
The present invention solves technical problem, adopt the following technical scheme that
The present invention is a kind of is to include walking as follows based on the feature of the removing method of myoelectricity noise in multichannel EEG signals
Rapid:
Step one: gathered and recorded the EEG signals of t N channel by EEG measuring equipment, be designated as: X (t)=[x1
(t),x2(t),…,xn(t),…,xN(t)]T, xnT () is the EEG signals of t the n-th passage, T is the transposition of matrix;1≤n
≤N;
Step 2: application population mean empirical mode decomposition is by EEG signals x of described n-th passagenT () is decomposed into P
Intrinsic modal components, is designated as: In(t)=[i1(n)(t),i2(n)(t),…,ip(n)(t),…,iP(n)(t)]T;ip(n)When () is t t
Carve EEG signals x of the n-th passagenPth the intrinsic modal components of (t);1≤p≤P;Thus obtain the brain electricity of t N channel
The intrinsic mode Component Matrices of signal X (t), is designated as: I (t)=[I1(t),I2(t),…,In(t),…,IN(t)]T;
Step 3: ask for EEG signals x of described n-th passagenPth intrinsic modal components i of (t)p(n)(t) from phase
Coefficient values Rp(n), as described autocorrelation coefficient Rp(n)During less than threshold θ, it is determined that described pth intrinsic modal components ip(n)(t)
For the intrinsic modal components containing myoelectricity noise;Thus pick out all containing flesh from described intrinsic mode Component Matrices I (t)
Electrical noise intrinsic modal components, and form containing the intrinsic mode Component Matrices of myoelectricity noise, be designated as M (t)=[m1(t),
m2(t),…,mB(t)]T;B represent containing myoelectricity noise the sum of intrinsic modal components;
Step 4: described intrinsic mode Component Matrices M (t) containing myoelectricity noise is carried out blind source with canonical correlation analysis
The separation of signal, obtains hybrid matrix A, solves mixed matrix W and source signal matrix Y (t)=[y1(t),y2(t),…,yb(t),…,
yB(t)]T;ybT () represents the b canonical variable, and have: M (t)=AY (t) or Y (t)=WM (t);1≤b≤B;
Step 5: ask for the b exemplary component y in described source signal matrix Y (t)bThe autocorrelation coefficient values r of (t)b, when
Described auto-correlation system rbDuring less than set threshold value e, it is determined that described the b exemplary component ybT () is the allusion quotation containing myoelectricity noise
Type component;And zero will be set to containing the exemplary component of myoelectricity noise;Thus by all containing flesh in described source signal matrix Y (t)
The exemplary component of electrical noise is all set to zero, obtains not containing the source signal matrix of myoelectricity noise
Step 6: utilize formula (1) to obtain not containing the intrinsic mode Component Matrices of myoelectricity noise
Step 7: by the described intrinsic mode Component Matrices not containing myoelectricity noiseIn each intrinsic modal components press
According to the position before selecting in described intrinsic mode Component Matrices I (t) of each leisure, replace in described intrinsic mode Component Matrices I (t)
Corresponding intrinsic modal components;Thus obtain removing intrinsic mode Component Matrices I ' (t) after making an uproar=[I '1(t),I′2(t),…,
I′n(t),…,I′N(t)]T;
Step 8: utilize formula (2) to obtain removing the clean EEG signals of the n-th passage after making an uproarThus obtain removal
N channel after making an uproar EEG signals
In formula (2), i 'p(n)T () represents EEG signals x of the n-th passagenPth intrinsic the mode t removal of () is made an uproar after is divided
Amount.
The present invention, compared with tradition multichannel denoising method, can not only remove the myoelectricity impact on brain electricity, can reduce simultaneously
Brain electric information loss in processing procedure;Concrete has the beneficial effect that:
1, step 2 of the present invention and step 3, carries out the average mode decomposition of overall experience to each passage EEG signals,
The intrinsic modal components of the multiple passages arrived, by asking for the autocorrelation coefficient of each intrinsic modal components, autocorrelation coefficient
Constitute noisy intrinsic mode Component Matrices less than the intrinsic modal components setting threshold value, be for further processing.This mode is protected
Stay the intrinsic modal components containing only brain electricity, i.e. remain brain electric information therein.Divide with canonical correlation analysis and independent element
The mode to whole brain electric treatment of analysing is compared, it is possible to preferably retain original brain electricity, decreases the brain electricity brought because of denoising process
Information dropout.
2, step 4 of the present invention, step 5 and step 6, uses canonical correlation analysis to carry out blind source signal separation, it is possible to will
Myoelectricity concentrates in the exemplary component of minority.And it is existing based on population mean empirical mode decomposition with independent component analysis (EEMD-
ICA) multichannel denoising method then carries out blind source signal separation by independent component analysis, because what independent component analysis obtained
The most both having comprised myoelectricity in isolated component and also comprised brain electricity, the noisy isolated component of zero setting can lose brain electricity, the isolated component of reservation
Middle meeting comprises myoelectricity.Therefore the present invention is compared to EEMD-ICA multichannel denoising method, can not only preferably remove myoelectricity, and
Can preferably retain brain electric information.
Accompanying drawing explanation
Fig. 1 is the main flow chart of the inventive method;
Fig. 2 a is the clean simulation EEG signals schematic diagram of 19 passages;
Fig. 2 b is by simulation EEG signals (λ=1) schematic diagram of 19 passages of myoelectricity noise jamming;
Fig. 2 c is the intrinsic modal components schematic diagram that the part obtained by the inventive method is noisy;
Fig. 2 d is the part exemplary component schematic diagram obtained by the inventive method;
Fig. 2 e is to eliminate, by what the inventive method obtained, the EEG signals schematic diagram rebuild after myoelectricity noise;
Fig. 3 a is the simulation EEG signals schematic diagram of 19 passages;
Fig. 3 b is by simulation EEG signals (λ=3) schematic diagram of 19 passages of myoelectricity noise jamming;
Fig. 3 c is to process simulation brain electricity, the EEG signals rebuild after obtained removal myoelectricity noise by the inventive method
Schematic diagram;
Fig. 3 d is to process simulation brain electricity, weight after obtained removal myoelectricity noise by EEMD-ICA band-wise processing method
Build EEG signals schematic diagram;
Fig. 3 e is that the inventive method processes simulation brain electricity with EEMD-ICA method, and its denoising performance index is the most root mean square
Comparison diagram;
Fig. 3 f is that the inventive method processes simulation brain electricity, the ratio of its denoising performance index correlation coefficient with EEMD-ICA method
Relatively scheme;
Fig. 4 a is the clean true EEG signals schematic diagram of 19 passages;
Fig. 4 b is by the true EEG signals schematic diagram of 19 passage mixing of myoelectricity noise jamming;
Fig. 4 c is for processing true brain electricity, the EEG signals rebuild after obtained removal myoelectricity noise by the inventive method
Schematic diagram;
Fig. 4 d is to process true brain electricity, after obtained removal myoelectricity noise by the band-wise processing method of EEMD-ICA
Rebuild EEG signals schematic diagram;
Fig. 4 e is that the inventive method processes true brain electricity with EEMD-ICA method, and its denoising performance index is relative to root-mean-square by mistake
The comparison diagram of difference;
Fig. 4 f is that the inventive method processes true brain electricity, the ratio of its denoising performance index correlation coefficient with EEMD-ICA method
Relatively scheme;
Fig. 5 a is the epileptic EEG Signal schematic diagram that 21 passages comprise myoelectricity and eye electrical noise;
Fig. 5 b is the EEG signals schematic diagram by rebuilding after the removal myoelectricity noise obtained by the inventive method.
Detailed description of the invention
As it is shown in figure 1, based on the removing method of myoelectricity noise in multichannel EEG signals be: first use population mean warp
Test mode each passage EEG signals is decomposed, obtain the intrinsic modal components of each passage;Pass through autocorrelation coefficient again
Judge the intrinsic modal components of Noise, be made up of noisy intrinsic mode Component Matrices the intrinsic modal components of Noise;Then
Noisy intrinsic mode Component Matrices is carried out Blind Signal Separation;Finally judge the exemplary component of Noise with autocorrelation coefficient, put
Zero noise component(s) reconstruction signal.
As a example by separately below by pure simulation EEG signals, half simulation EEG signals and actual measurement EEG signals, in conjunction with accompanying drawing
Specific embodiment is described.
The purest simulation EEG signals
In this part, will illustrate that two examples, first example are to introduce the detailed description of the invention of the present invention, second
Individual example be by the present invention compared with EEMD-ICA band-wise processing method.
(1) example one
Step one: gathered and record t N=19 the passage brain telecommunications by myoelectricity noise jamming by EEG measuring equipment
Number X (t)=[x1(t),x2(t),…,xn(t),…,x19(t)]T, 1≤n≤N;Wherein, xnT () is the brain of t the n-th passage
The signal of telecommunication, T is the transposition of matrix;X (t) as shown in Figure 2 b=XEEG(t)+λXEMG(t), λ=1;And X as shown in Figure 2 aEEG
(t)=[xEEG1(t),xEEG2(t),…,xEEG19(t)]TRepresent the pure simulation EEG signals of 19 passages;XEMG(t)=[xEMG1(t),
xEMG2(t),…,xEMG19(t)]TRepresent the simulation electromyographic signal of 19 passages.Simulation brain electricity XEEGT () each passage is by 5
The EEG fragment of individual 2s length is formed by connecting, and each EEG fragment is by 4 frequencies sinusoidal signal in the range of 4-30Hz
It is formed by stacking, and these 4 frequencies randomly generate.EMG signal is then by one section of 19 in 20-60Hz frequency range
Passage white noise generates.Signal sampling frequency is 250Hz, takes the analogue signal of totally 10 seconds, the most a total of 2500 sampled points;
Step 2: application population mean empirical mode decomposition will mix each passage x in EEG signals X (t)nT () decomposes
Become p=11 intrinsic modal components In(t)=[i1(n)(t),i2(n)(t),…,ip(n)(t),…,i11(n)(t)]T;ip(n)(t)
EEG signals x for t the n-th passagenPth the intrinsic modal components of (t);1≤p≤P;EEG signals to 19 passages
Intrinsic mode Component Matrices I (t)=[I of 19 passages is obtained respectively after application population mean empirical mode decomposition1(t),I2
(t),…,In(t),…,I19(t)]T;
Step 3: ask for EEG signals x of the n-th passagenPth intrinsic modal components i of (t)p(n)The auto-correlation system of (t)
Numerical value Rp(n), set threshold θ=0.995, as autocorrelation coefficient Rp(n)During less than threshold θ, it is determined that pth intrinsic modal components
ip(n)T () is the intrinsic modal components containing myoelectricity noise;Thus from intrinsic mode Component Matrices I (t), pick out all containing
Have myoelectricity noise intrinsic modal components, and form containing the intrinsic mode Component Matrices of myoelectricity noise, be designated as M (t)=[m1
(t),m2(t),…mb(t),…,m127(t)]T;B=127 represent containing myoelectricity noise the sum of intrinsic modal components;Its
Middle part noisy intrinsic modal components such as Fig. 2 c;
Step 4: intrinsic mode Component Matrices M (t) containing myoelectricity noise is carried out blind source signal with canonical correlation analysis
Separation, obtain hybrid matrix A, solve mixed matrix W and source signal matrix Y (t)=[y1(t),y2(t),…,yb(t),…,yB
(t)]T;ybT () represents the b canonical variable, and have: M (t)=AY (t) or Y (t)=WM (t);1≤b≤B;
In the present embodiment, noisy intrinsic mode Component Matrices M (t) is carried out the process that time delay is 1, obtains 2 data sets:
M (t) and J (t)=M (t-1), then carries out the separation of blind source signal, thus obtains with canonical correlation analysis to the two data set
Source signal matrix Y (t)=[y constituted to hybrid matrix A, the mixed matrix W of solution and 127 canonical variables of M (t)1(t),y2
(t),…,yb(t),…,y127(t)]T, wherein part canonical variable such as Fig. 2 d;
Step 5: ask for the b exemplary component y in source signal matrix Y (t)bThe autocorrelation coefficient values r of (t)b, when from phase
Relation rbDuring less than set threshold value e, it is determined that the b exemplary component ybT () is the exemplary component containing myoelectricity noise, herein e
=0.95;And zero will be set to containing the exemplary component of myoelectricity noise;Thus make an uproar all in source signal matrix Y (t) containing myoelectricity
The exemplary component of sound is all set to zero, obtains not containing the source signal matrix of myoelectricity noise
Step 6: utilize formula (1) to obtain not containing the intrinsic mode Component Matrices of myoelectricity noise
Step 7: the intrinsic mode Component Matrices of myoelectricity noise will not containedIn each intrinsic modal components according to respectively
Position before selecting in comfortable intrinsic mode Component Matrices I (t), replaces eigen mode corresponding in intrinsic mode Component Matrices I (t)
State component;Thus obtain removing intrinsic mode Component Matrices I ' (t) after making an uproar=[I '1(t),I′2(t),…,I′n(t),…,
I′19(t)]T;
Step 8: utilize formula (2) to obtain removing the clean EEG signals of the n-th passage after making an uproarThus obtain removal
N channel after making an uproar EEG signals
In formula (2), i 'p(n)T () represents EEG signals x of the n-th passagenThe removal of (t) make an uproar after pth intrinsic
Modal components.After 19 passages are processed successively, 19 passage EEG signals of the noise that is eliminatedAs shown in Figure 2 e;
By the EEG signals after denoisingClean true EEG signals X with simulationEEGT () contrasts, from Fig. 2 a and
Fig. 2 e can be clearly observed myoelectricity noise be substantially completely eliminated, and remain former clean true brain telecommunications well
Number detailed information, illustrate the present invention in the multichannel brain signal of telecommunication myoelectricity noise eliminate effectiveness.
(2) example two
For the effect of the quantitative evaluation present invention, for this by inventive method compared with EEMD-ICA band-wise processing method
Relatively.
First pure simulation EEG signals X of the multichannel by myoelectricity noise jamming of N=19 is simulated1(t)=[x1(t),x2
(t),…,x19(t)]T, 1≤n≤N;The wherein X shown in Fig. 3 b1T () is XEEG(t) and XEMGThe mixed signal of (t), i.e. X1(t)=
XEEG(t)+λXEMG(t), the X shown in Fig. 3 aEEG(t)=[xEEG1(t),xEEG2(t),…,xEEG19(t)]TAnd XEMG(t)=[xEMG1
(t),xEMG2(t),…,xEMG19(t)]TRepresent the 19 clean EEG signals of passage and myoelectricity noises of simulation respectively.
Signal sampling frequency is 250Hz, takes the analogue signal of totally 10 seconds, the most a total of S1=2500 sampled points, this
In λ represent the intensity of myoelectricity noise jamming EEG signals, λ=3 in Fig. 3 b.According to the step in example one, the inventive method obtains
To the EEG signals removing the reconstruction of myoelectricity noise as shown in Figure 3 cEEMD-ICA manifold
Road processing method is similar with EEMD-CCA band-wise processing mode, and two kinds of methods set identical threshold value.Fig. 3 d show EEMD-
ICA band-wise processing method removes the EEG signals that myoelectricity noise is rebuild
In order to effectively assess the denoising effect of two kinds of algorithms of EEMD-CCA and EEMD-ICA, select two performance indications conducts
Evaluation index, is relative root-mean-square error (RRMSE) and correlation coefficient (CC) respectively.
Signal to noise ratio defines: SNR=RMS (XEEG)/RMS(λ·XEMG), wherein RMS represents root-mean-square,WithCan be seen that λ and noise
Being inversely proportional to than SNR, λ is the biggest, and signal to noise ratio is the lowest.
Additionally, root-mean-square error relatively is defined as follows:RRMSE value
The least show that denoising effect is the best.
Second individual character energy index is correlation coefficient CC.It is used herein as correlation coefficient as performance indications, can intuitively show EEG
Structural similarity before and after signal denoising, the value the highest explanation data of correlation coefficient are recovered the best.Fig. 3 e, Fig. 3 f illustrate two
The method of kind eliminating myoelectricity noise and retaining the effect that brain is electric under different signal to noise ratios, the most all can be seen that in the inventive method
EEMD ICA processing method stably it is better than in different signal to noise ratios.
2. half simulation EEG signals
In order to verify that the present invention has more preferable denoising effect than EEMD-ICA further, will use by true EEG signals
The EEG signals being mixed to get with true electromyographic signal, compares the denoising effect of two kinds of methods.
Multichannel by myoelectricity noise jamming half simulation EEG signals X of first simulation N=192(t)=[x1(t),x2
(t),…,x19(t)]T, the most as shown in Figure 4 b, during λ=1.5, X2T () is XEEG(t) and XEMGThe mixed signal of (t): X2(t)=
XEEG(t)+λXEMG(t), X as shown in fig. 4 aEEG(t) and XEMGT () represents the 19 passages truly clean EEG signals of simulation respectively
With myoelectricity noise, and XEEG(t)=[xEEG1(t),xEEG2(t),…,xEEG19(t)]T, XEMG(t)=[xEMG1(t),xEMG2
(t),…,xEMG19(t)]T。
Signal sampling frequency is 1000Hz, takes the analogue signal of totally 10 seconds, the most a total of S2=10000 sampled points.
According to the step in example one, the inventive method obtains the EEG signals removing the reconstruction of myoelectricity noise as illustrated in fig. 4 cFig. 4 d show EEMD-ICA band-wise processing method and removes the brain that myoelectricity noise is rebuild
The signal of telecommunicationShow as evaluation index, Fig. 4 e and Fig. 4 f with performance indications RRMSE and CC
Two kinds of methods eliminating myoelectricity noise and retaining the effects of brain electricity under different signal to noise ratios, the most all can be seen that in the present invention
Method is stably better than EEMD-ICA processing method in different signal to noise ratios.
3. actual measurement EEG signals
At the Part III of this example, use true eeg data as experimental subject, use EEMD-CCA multichannel to calculate
Method processes, and passes judgment on the denoising effect of inventive method.Fig. 5 a is the true epilepsy EEG signals of one section of 21 passage, this letter
Number sample frequency be 250Hz, the sampling time is 10s, altogether 2500 sampled points.This segment signal quilt is can be seen that from Fig. 5 a
Eye electricity and two kinds of noises of myoelectricity are disturbed.Eye electricity substantially can be seen near the 2.5 of Fp1 and Fp2 passage, 3.5,6 and 7.5s
Measure.Same myoelectricity interference can be at the 0s-3.9s of F7, T3, T5, C3, T1 passage and the 5s-of F8, T4, F4, C4, P4
It is observed at 10s.Epilepsy can be observed at passage T2, F8, T4, T6, and part outbreak region is tight by myoelectricity noise
Heavily disturbing, this influences whether the analysis of follow-up EEG signals and works epilepsy zone location, therefore eliminates myoelectricity noise
Interference is extremely necessary.
According to the step of example one, above-mentioned true EEG signals is processed, use the removal flesh that the inventive method obtains
The EEG signals such as Fig. 5 b rebuild after electrical noise.By comparison diagram 5a, Fig. 5 b, it appeared that myoelectricity noise is not only disappeared by the present invention
That removes is the cleanest, moreover it is possible to intactly retain the Key detail information in EEG signals, such as, by flesh in passage F8, T4, T6
The epilepsy brain electricity part of electrical noise interference has eliminated myoelectricity noise, and eye electrical noise therein is by complete preservation
Get off.
To sum up, the present invention can solve the problem that the difficult problem that myoelectricity noise eliminates in the case of multichannel, and additionally the present invention is one to finish
Full automatic monitoring the method eliminating myoelectricity noise, it is to avoid the impact that human intervention causes.The method is applicable to clinic and examines
In the disconnected multichannel brain electric equipment with Neuroscience Research, compared with the multichannel brain electric processing method of EEMD-ICA, it is possible to take
Obtain more preferable denoising effect, significant to the research real bioelectrical activity of brain further.
Claims (1)
1., based on a removing method for myoelectricity noise in multichannel EEG signals, it is characterized in that comprising the steps:
Step one: gathered and recorded the EEG signals of t N channel by EEG measuring equipment, be designated as: X (t)=[x1(t),x2
(t),…,xn(t),…,xN(t)]T, xnT () is the EEG signals of t the n-th passage, T is the transposition of matrix;1≤n≤N;
Step 2: application population mean empirical mode decomposition is by EEG signals x of described n-th passagenT () is decomposed into P eigen mode
State component, is designated as: In(t)=[i1(n)(t),i2(n)(t),…,ip(n)(t),…,iP(n)(t)]T;ip(n)T () is that t n-th is led to
EEG signals x in roadnPth the intrinsic modal components of (t);1≤p≤P;Thus obtain EEG signals X (t) of t N channel
Intrinsic mode Component Matrices, be designated as: I (t)=[I1(t),I2(t),…,In(t),…,IN(t)]T;
Step 3: ask for EEG signals x of described n-th passagenPth intrinsic modal components i of (t)p(n)The auto-correlation system of (t)
Numerical value Rp(n), as described autocorrelation coefficient Rp(n)During less than threshold θ, it is determined that described pth intrinsic modal components ip(n)T () is for containing
There is the intrinsic modal components of myoelectricity noise;Thus pick out from described intrinsic mode Component Matrices I (t) and all make an uproar containing myoelectricity
Sound intrinsic modal components, and form containing the intrinsic mode Component Matrices of myoelectricity noise, be designated as M (t)=[m1(t),m2
(t),…,mB(t)]T;B represent containing myoelectricity noise the sum of intrinsic modal components;
Step 4: described intrinsic mode Component Matrices M (t) containing myoelectricity noise is carried out blind source signal with canonical correlation analysis
Separation, obtain hybrid matrix A, solve mixed matrix W and source signal matrix Y (t)=[y1(t),y2(t),…,yb(t),…,yB
(t)]T;ybT () represents the b canonical variable, and have: M (t)=AY (t) or Y (t)=WM (t);1≤b≤B;
Step 5: ask for the b exemplary component y in described source signal matrix Y (t)bThe autocorrelation coefficient values r of (t)b, when described
Auto-correlation system rbDuring less than set threshold value e, it is determined that described the b exemplary component ybT () divides for the typical case containing myoelectricity noise
Amount;And zero will be set to containing the exemplary component of myoelectricity noise;Thus make an uproar all in described source signal matrix Y (t) containing myoelectricity
The exemplary component of sound is all set to zero, obtains not containing the source signal matrix of myoelectricity noise
Step 6: utilize formula (1) to obtain not containing the intrinsic mode Component Matrices of myoelectricity noise
Step 7: by the described intrinsic mode Component Matrices not containing myoelectricity noiseIn each intrinsic modal components according to respectively
Position before selecting in comfortable described intrinsic mode Component Matrices I (t), replaces in described intrinsic mode Component Matrices I (t) corresponding
Intrinsic modal components;Thus obtain removing intrinsic mode Component Matrices I ' (t) after making an uproar=[I '1(t),I′2(t),…,I′n
(t),…,I′N(t)]T;
Step 8: utilize formula (2) to obtain removing the clean EEG signals of the n-th passage after making an uproarThus obtain after removal makes an uproar
N channel EEG signals
In formula (2), i 'p(n)T () represents EEG signals x of the n-th passagenThe removal of (t) make an uproar after pth intrinsic modal components.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610692634.7A CN106236080B (en) | 2016-08-19 | 2016-08-19 | The removing method of myoelectricity noise in EEG signals based on multichannel |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610692634.7A CN106236080B (en) | 2016-08-19 | 2016-08-19 | The removing method of myoelectricity noise in EEG signals based on multichannel |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106236080A true CN106236080A (en) | 2016-12-21 |
CN106236080B CN106236080B (en) | 2019-03-08 |
Family
ID=57592616
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610692634.7A Active CN106236080B (en) | 2016-08-19 | 2016-08-19 | The removing method of myoelectricity noise in EEG signals based on multichannel |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106236080B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106805945A (en) * | 2017-01-22 | 2017-06-09 | 合肥工业大学 | The removing method of Muscle artifacts in a kind of EEG signals of a small number of passages |
CN109598222A (en) * | 2018-11-26 | 2019-04-09 | 南开大学 | Wavelet neural network Mental imagery brain electricity classification method based on the enhancing of EEMD data |
CN110780162A (en) * | 2019-10-10 | 2020-02-11 | 国网天津市电力公司电力科学研究院 | Method for extracting partial discharge signal of primary and secondary fusion power distribution switch and detection device |
WO2020088083A1 (en) * | 2018-10-31 | 2020-05-07 | 安徽华米信息科技有限公司 | Noise detection method and apparatus |
CN111631710A (en) * | 2020-06-22 | 2020-09-08 | 中国科学技术大学 | Method for eliminating myoelectric artifacts in state-related dynamic electroencephalogram signals |
CN113509169A (en) * | 2021-08-05 | 2021-10-19 | 成都乐享智家科技有限责任公司 | Multi-parameter-based non-contact sleep apnea detection system and method |
WO2023042431A1 (en) * | 2021-09-17 | 2023-03-23 | ソニーグループ株式会社 | Measurement device and measurement method |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101474070A (en) * | 2009-01-21 | 2009-07-08 | 电子科技大学 | Method for removing ocular artifacts in brain-electrical signal |
WO2009087486A2 (en) * | 2007-12-31 | 2009-07-16 | Emotiv Systems Pty Ltd | Biosensor noise reduction |
CN101690659A (en) * | 2009-09-29 | 2010-04-07 | 华东理工大学 | Brain wave analysis method |
CN101869477A (en) * | 2010-05-14 | 2010-10-27 | 北京工业大学 | Self-adaptive EEG signal ocular artifact automatic removal method |
EP2630910A1 (en) * | 2012-02-24 | 2013-08-28 | AIT Austrian Institute of Technology GmbH | Method for detecting artefacts |
CN103845052A (en) * | 2014-02-20 | 2014-06-11 | 清华大学 | Human body faint early warning method based on acquired electroencephalogram signals |
WO2015039744A1 (en) * | 2013-09-18 | 2015-03-26 | Universitätsklinikum Erlangen | Method and device for measuring cerebral perfusion |
CN104688220A (en) * | 2015-01-28 | 2015-06-10 | 西安交通大学 | Method for removing ocular artifacts in EEG signals |
US20150327813A1 (en) * | 2010-08-02 | 2015-11-19 | Chi Yung Fu | Method for processing brainwave signals |
CN105342605A (en) * | 2015-12-09 | 2016-02-24 | 西安交通大学 | Method for removing myoelectricity artifacts from brain electrical signals |
CN105496363A (en) * | 2015-12-15 | 2016-04-20 | 浙江神灯生物科技有限公司 | Method for classifying sleep stages on basis of sleep EGG (electroencephalogram) signal detection |
-
2016
- 2016-08-19 CN CN201610692634.7A patent/CN106236080B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2009087486A2 (en) * | 2007-12-31 | 2009-07-16 | Emotiv Systems Pty Ltd | Biosensor noise reduction |
CN101474070A (en) * | 2009-01-21 | 2009-07-08 | 电子科技大学 | Method for removing ocular artifacts in brain-electrical signal |
CN101690659A (en) * | 2009-09-29 | 2010-04-07 | 华东理工大学 | Brain wave analysis method |
CN101869477A (en) * | 2010-05-14 | 2010-10-27 | 北京工业大学 | Self-adaptive EEG signal ocular artifact automatic removal method |
US20150327813A1 (en) * | 2010-08-02 | 2015-11-19 | Chi Yung Fu | Method for processing brainwave signals |
EP2630910A1 (en) * | 2012-02-24 | 2013-08-28 | AIT Austrian Institute of Technology GmbH | Method for detecting artefacts |
WO2015039744A1 (en) * | 2013-09-18 | 2015-03-26 | Universitätsklinikum Erlangen | Method and device for measuring cerebral perfusion |
CN103845052A (en) * | 2014-02-20 | 2014-06-11 | 清华大学 | Human body faint early warning method based on acquired electroencephalogram signals |
CN104688220A (en) * | 2015-01-28 | 2015-06-10 | 西安交通大学 | Method for removing ocular artifacts in EEG signals |
CN105342605A (en) * | 2015-12-09 | 2016-02-24 | 西安交通大学 | Method for removing myoelectricity artifacts from brain electrical signals |
CN105496363A (en) * | 2015-12-15 | 2016-04-20 | 浙江神灯生物科技有限公司 | Method for classifying sleep stages on basis of sleep EGG (electroencephalogram) signal detection |
Non-Patent Citations (1)
Title |
---|
彭志红: "改进独立分量分析在脑电信号伪迹消除中的应用研究", 《PROCEEDINGS OF THE 29TH CHINESE CONTROL CONFERENCE》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106805945A (en) * | 2017-01-22 | 2017-06-09 | 合肥工业大学 | The removing method of Muscle artifacts in a kind of EEG signals of a small number of passages |
CN106805945B (en) * | 2017-01-22 | 2019-06-18 | 合肥工业大学 | It is a kind of minority channel EEG signals in Muscle artifacts removing method |
WO2020088083A1 (en) * | 2018-10-31 | 2020-05-07 | 安徽华米信息科技有限公司 | Noise detection method and apparatus |
CN109598222A (en) * | 2018-11-26 | 2019-04-09 | 南开大学 | Wavelet neural network Mental imagery brain electricity classification method based on the enhancing of EEMD data |
CN109598222B (en) * | 2018-11-26 | 2023-04-07 | 南开大学 | EEMD data enhancement-based wavelet neural network motor imagery electroencephalogram classification method |
CN110780162A (en) * | 2019-10-10 | 2020-02-11 | 国网天津市电力公司电力科学研究院 | Method for extracting partial discharge signal of primary and secondary fusion power distribution switch and detection device |
CN110780162B (en) * | 2019-10-10 | 2022-11-08 | 国网天津市电力公司电力科学研究院 | Method for extracting partial discharge signal of primary and secondary fusion power distribution switch and detection device |
CN111631710A (en) * | 2020-06-22 | 2020-09-08 | 中国科学技术大学 | Method for eliminating myoelectric artifacts in state-related dynamic electroencephalogram signals |
CN113509169A (en) * | 2021-08-05 | 2021-10-19 | 成都乐享智家科技有限责任公司 | Multi-parameter-based non-contact sleep apnea detection system and method |
WO2023042431A1 (en) * | 2021-09-17 | 2023-03-23 | ソニーグループ株式会社 | Measurement device and measurement method |
Also Published As
Publication number | Publication date |
---|---|
CN106236080B (en) | 2019-03-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Chen et al. | A novel EEMD-CCA approach to removing muscle artifacts for pervasive EEG | |
CN106236080A (en) | Based on the removing method of myoelectricity noise in multichannel EEG signals | |
CN102835955B (en) | Method of automatically removing ocular artifacts from electroencephalogram signal without setting threshold value | |
Sadasivan et al. | SVD based technique for noise reduction in electroencephalographic signals | |
CN107260166A (en) | A kind of electric artefact elimination method of practical online brain | |
CN104720797B (en) | One kind is based on myoelectricity noise cancellation method in single pass EEG signals | |
CN110269609B (en) | Method for separating ocular artifacts from electroencephalogram signals based on single channel | |
CN105342605B (en) | A kind of method of Muscle artifacts in removal EEG signals | |
CN104688220B (en) | Method for removing ocular artifacts in EEG signals | |
CN106805945B (en) | It is a kind of minority channel EEG signals in Muscle artifacts removing method | |
Gruszczyńska et al. | Epilepsy identification based on EEG signal using RQA method | |
CN108309290A (en) | The automatic removal method of Muscle artifacts in single channel EEG signals | |
CN109480872B (en) | Driving fatigue detection method based on electroencephalogram signal frequency band energy ratio characteristics | |
CN103110418A (en) | Electroencephalogram signal characteristic extracting method | |
CN109820503A (en) | The synchronous minimizing technology of a variety of artefacts in single channel EEG signals | |
CN109480832A (en) | The removing method of Muscle artifacts in a kind of single pass EEG signals | |
CN112426162A (en) | Fatigue detection method based on electroencephalogram signal rhythm entropy | |
CN112294340A (en) | Method, system, storage medium and computer equipment for quickly and automatically removing myoelectric artifacts | |
Moser et al. | Classification and detection of single evoked brain potentials using time-frequency amplitude features | |
Islam et al. | Probability mapping based artifact detection and wavelet denoising based artifact removal from scalp EEG for BCI applications | |
CN113842115A (en) | Improved EEG signal feature extraction method | |
CN106859641B (en) | A method of based on nuclear-magnetism artefact in automatic ICA removal EEG signal | |
Feng et al. | A Novel SSA-CCA Framework forMuscle Artifact Removal from Ambulatory EEG | |
CN111671421A (en) | Electroencephalogram-based children demand sensing method | |
CN115017996B (en) | Mental load prediction method and system based on multiple physiological parameters |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |