CN107260166A - A kind of electric artefact elimination method of practical online brain - Google Patents
A kind of electric artefact elimination method of practical online brain Download PDFInfo
- Publication number
- CN107260166A CN107260166A CN201710381917.4A CN201710381917A CN107260166A CN 107260166 A CN107260166 A CN 107260166A CN 201710381917 A CN201710381917 A CN 201710381917A CN 107260166 A CN107260166 A CN 107260166A
- Authority
- CN
- China
- Prior art keywords
- component
- artefact
- wavelet
- isolated
- eeg signals
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 68
- 210000004556 brain Anatomy 0.000 title claims abstract description 53
- 230000008030 elimination Effects 0.000 title claims abstract description 20
- 238000003379 elimination reaction Methods 0.000 title claims abstract description 20
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 20
- 238000012937 correction Methods 0.000 claims abstract description 14
- 230000006870 function Effects 0.000 claims description 39
- 238000001228 spectrum Methods 0.000 claims description 20
- 239000013598 vector Substances 0.000 claims description 17
- 238000012545 processing Methods 0.000 claims description 16
- 230000002123 temporal effect Effects 0.000 claims description 10
- 230000009466 transformation Effects 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000000844 transformation Methods 0.000 claims description 9
- 238000000926 separation method Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000000819 phase cycle Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 230000017105 transposition Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 238000005070 sampling Methods 0.000 claims 1
- 238000004422 calculation algorithm Methods 0.000 abstract description 22
- 239000000203 mixture Substances 0.000 abstract description 12
- 238000005516 engineering process Methods 0.000 abstract description 5
- 230000007177 brain activity Effects 0.000 abstract description 2
- 230000010365 information processing Effects 0.000 abstract 1
- 238000000537 electroencephalography Methods 0.000 description 49
- 238000012880 independent component analysis Methods 0.000 description 40
- 230000005611 electricity Effects 0.000 description 14
- 238000002474 experimental method Methods 0.000 description 14
- 230000003340 mental effect Effects 0.000 description 12
- 238000010586 diagram Methods 0.000 description 9
- 238000004458 analytical method Methods 0.000 description 7
- 238000011160 research Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000000763 evoking effect Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 210000003739 neck Anatomy 0.000 description 2
- 230000035479 physiological effects, processes and functions Effects 0.000 description 2
- 230000000638 stimulation Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 206010002091 Anaesthesia Diseases 0.000 description 1
- 208000014644 Brain disease Diseases 0.000 description 1
- 230000037005 anaesthesia Effects 0.000 description 1
- 230000003466 anti-cipated effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000002592 echocardiography Methods 0.000 description 1
- 230000002996 emotional effect Effects 0.000 description 1
- 238000013401 experimental design Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 210000003128 head Anatomy 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 210000001595 mastoid Anatomy 0.000 description 1
- 210000003205 muscle Anatomy 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 238000006116 polymerization reaction Methods 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 210000004761 scalp Anatomy 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
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
- A61B5/72—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
-
- 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
- A61B5/7235—Details of waveform analysis
-
- 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
- A61B5/7235—Details of waveform analysis
- A61B5/7253—Details of waveform analysis characterised by using transforms
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Physiology (AREA)
- Psychology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present invention relates to the electric artefact elimination method of the practical online brain of one kind, belong to biomedical information processing technology field.The present invention carries out the real-time EEG signals of the few passage collected down-sampled, notch filter and linear drift correction, then 7 layers of wavelet decomposition are carried out to the EEG signals of few passage respectively using wavelet transform, single pass EEG signals are converted into multichannel, reconstruct wavelet coefficient and as ICA input, the quick obtaining of independent element is realized using FastICA algorithms, the time domain of normal brain activity electric component is different from according to each artefact composition in isolated component, correlative character between frequency domain and sequence, hierarchical clustering algorithm is introduced to cluster each independent element, automatically identify artefact composition generic, remaining component will be reconstructed again after such each artefact component zero setting, obtain clean EEG signals.The present invention solves existing method can not be in the case of few passage, on-line automatic identification and the problem of reject a variety of conventional brains electric artefact.
Description
Technical field
The present invention relates to the electric artefact elimination method of the practical online brain of one kind, belong to biomedical information treatment technology neck
Domain.
Background technology
Brain electricity (electroencephalography, EEG) has contained substantial amounts of psychology, physiology and pathological information, at present
It is widely used in many research necks such as brain diseases diagnosis and brain-computer interface (Brain-computer interface, BCI)
In domain.Normal EEG signals are mainly distributed on δ frequency bands (0-4Hz), θ frequency bands (4-8Hz), α frequency bands (8-13Hz), β frequency bands (13-
20Hz), in five frequency band ranges of γ frequency bands (30-80Hz), and EEG amplitude is very faint, easily by eye electricity
The various puppets such as (electrooculogram, EOG), myoelectricity (Electromyogram, EMG), 50Hz power frequencies and linear drift
There is the performance for having had a strong impact on BCI in the interference of mark, these artefacts.
In some conventional artefact elimination methods, the unavoidable dirt for making brain electricity by artefacts such as EOG, EMG of artefact refusal
Dye；The situation that linear filtering can not overlap suitable for the frequency band of artefact and normal EEG signals；The method of linear regression according to
Rely in a good artefact reference signal, generally we are difficult to find a reference for being suitable for EMG artefacts and non-physiology artefact
Signal.Some researchers are directly rejected the electric medium-high frequency myoelectricity of original brain using the method for bandpass filter or given threshold, made an uproar
Sound and electro-ocular signal, although this method is simple, often also eliminate the useful brain electricity in part while artefact is rejected
Signal, causes pretreating effect undesirable.Blind source separating (Blind signal separation, BBS) technology is that one kind most has
The electric artefact elimination method of the brain of prospect, independent component analysis (Independent component analysis, ICA) is exactly
One kind in BBS methods.
ICA is a kind of statistical method commonly used during signal transacting is studied, and it, which is intended to searching, can maximize each in data point
Measure the linear projection of independence.It is assumed that the N-dimensional EEG signals X=[x collected from electrod-array1,x2,…,xN]T, the signal
It is by M unknown and statistical iteration signal source S=[s1,s2,…,sM]TLinear combination.The target of ICA algorithm is to find to divide
From matrix W so that：S=WX.Some improved ICA algorithms have been implemented and freely used, wherein, FastICA is compared to it
Mainly had the advantage that for its improved ICA algorithm：A, FastICA algorithm the convergence speed are fast, it is possible to increase algorithm it is real-time
Property；B, the algorithm are easy to use, and step parameter need not be set during data decimation；C, it have be similar to neural network algorithm simultaneously
Row, the advantage of distribution, computation complexity are low, and committed memory space is small, is suitable for the analysis of online EEG signals.Generally, ICA will
Seek the quantity of observation signal otherwise less than the quantity of source signal, then occur when the quantity of observation signal is less than source signal quantity
Super complete ICA phenomenons.
At present, the rejecting algorithm of artefact is more ripe in multichannel brain electric signal.And in single/few passage EEG signals
Artefact reject in, because brain wave acquisition number of channels is less, comprising effective information it is less, and lack artefact with reference to letter
Number, highly effective artefact elimination method is there is no at present.In addition, multi-lead EEG signals equipment, due to expensive, build is huge
Greatly, it is cumbersome, tend not to meet convenient collection, the requirement of processing brain electricity.In recent years, in order to solve ICA algorithm be used for it is single/
Super complete problem present in few passage electroencephalogramsignal signal analyzing, passage converter technique is emerging with the research of portable brain electric equipment
Rise, Bogdan Mijovic propose empirical mode decomposition (Empirical mode for single channel EEG signals
Decomposition, EMD) and the electric artefact elimination method of the brains that are combined of ICA, i.e. EMD-ICA methods, and Li Mingai is for few
Passage EEG signals are proposed wavelet transform (Discrete Wavelet Transform, DWT) and independent component analysis
The method being combined, i.e. DWT-ICA methods, this method are not transported online mainly for the electric artefact of eye in EEG signals
With, the present invention is mainly improved on the basis of this method, to improve artefact composition automatic identification and eliminating ability, and
Used online.
Wavelet transformation (Wavelet Transform, WT) is by the way that signal decomposition is obtained into multiple dimensioned wavelet function
Signal is in the local feature of time domain and frequency, and DWT calculating speed is fast, is widely used in brain electricity this non-linear, non-flat
The analysis of steady signal.Traditional ICA algorithm easily produces larger when directly being decomposed to the observation signal containing Gaussian noise
Error, also increases amount of calculation, significantly reduces separating effect, therefore, the present invention introduces DWT, DWT before ICA separation
ICA separating effect is enhanced while few passage brain electricity to be converted into the multichannel brain electric of suitable ICA analyses.Although ICA
Successfully the true brain electricity in the original EEG signals of multichannel can be separated with artefact composition, but artefact in this approach
The identification of component tends to rely on the visual inspection of professional, time-consuming and with very big subjectivity, and simple ICA algorithm is also
The requirement of on-line automatic efficient identification artefact composition is not reached much.
In terms of the on-line automatic identification of artefact, clustering technique is a kind of efficient method, and it is typically based on each independence
Some special characteristics of component, the composition related for isolating artefact from each independent element automatically.Some researcher's bases
Similarity between two independent elements has used K mean algorithms and Fuzzy C-Means Cluster Algorithm to be clustered, automatic identification
And reject artefact.And both clustering algorithms are all a kind of methods of iteration, it is required for the target numbers of given cluster in advance
Determine that the target numbers clustered in the number of times of iteration, actual cluster process are often unknown.Hierarchical clustering (Hierarchical
Clustering, HC) algorithm mainly has two big advantages for above two clustering algorithm：First, hierarchical clustering process
In dendrogram not only include the composition information of each class, and also provided with the formation of height of node in each class each
The intimate information of element in class；Secondly, the number of cluster need not be previously set in whole cluster process.
BCI based on Mental imagery (Motor imagery, MI) is relative to based on Steady State Visual Evoked Potential (Stady-
State visual evoked potential, SSVEP) and the BCI of P300 current potentials for, its synchronization independent of screen
Visual stimulus, it is possible to achieve asynchronous BCI, and subject imagination right-hand man motion intention can be by analyzing subject primary motor area
The EEG signals of C3 and C4 leads are recognized, thus the BCI based on such normal form is easily applied to portable BCI equipment
In.In recent years, emerging portable BCI, portable anesthesia depth monitoring system, portable sleep monitoring, portable emotional state
The application fields such as identification are analyzed generally directed to the EEG signals of online list/few passage.Therefore, the online few logical of practicality is studied
Mental imagery brain electric artefact in road is rejected algorithm and had great importance for the development of portable brain electric equipment.
The content of the invention
The invention provides a kind of combination DWT, ICA and HC electric artefact elimination method of practical online brain, i.e. DWICAC
Method, for solving in the electric Preprocessing Algorithm of current Mental imagery brain, a variety of electric artefacts of conventional brain can not be in single pass feelings
Under condition, on-line automatic identification and the problem of reject.
The technical scheme is that：A kind of electric artefact elimination method of practical online brain, the specific steps of methods described
It is as follows：
A, the original EEG signals to each passage carry out down-sampled, notch filter and linear drift is corrected；
B, the EEG signals to each passage after correction carry out multi-scale wavelet decomposition and wavelet coefficient list branch is reconstructed；
C, wavelet coefficient list branch reconstruction signal progress ICA decomposition respectively to each passage, the time domain for calculating each component are special
Levy, similarity feature between spectrum signature and sequence；
D, the isolated component to each passage carry out hierarchical clustering respectively, and automatic identification simultaneously rejects artefact component, and to residue
Isolated component carry out FastICA inverse transformations reconstruct and wavelet coefficient reconstruct, obtain clean EEG signals.
Down-sampled, notch filter and linear drift correction specific method in the step A is as follows：
It is A1, former in real time to C3, C4 passage respectively using the downsample functions in Matlab Matlab DSPToolBoxes
It is down-sampled that beginning EEG signals carry out 250Hz；
A2, the electric number of the brain using dlsim function pair C3, the C4 passages in Matlab Matlab DSPToolBoxes after down-sampled
According to progress 50HZ notch filter processing；
A3, using in Matlab Matlab DSPToolBoxes detrend function pairs trap processing after signal enter line
Property drift correction；
EEG signals X (t)=[x after being corrected by above-mentioned processing1(t),x2(t) ..., xM(t)]T, wherein：X
(t) it is one group of M dimension random vector, M is brain wave acquisition port number, T is transposition computing, t=1, and 2 ..., N, N is EEG signals X
(t) sampled point number.
In the step B, multi-scale wavelet decomposition and wavelet coefficient list branch reconstruct comprise the following steps that：
B1, selection " db4 " wavelet basis, utilize the wavedec function pairs in Matlab software wavelet transforms tool box
The signal x of a certain passage1(t) 7 layer scattering wavelet decompositions are carried out, signal x is obtained1(t) Coefficients of Approximation component ca7 and details system
Number component cd7, cd6, cd5, cd4, cd3, cd2, cd1；
B2, the Coefficients of Approximation point using the wrcoef functions in wavelet transform tool box respectively to being obtained in step B1
Amount and detail coefficients component carry out single branch reconstruct, obtain corresponding to ca7, cd7, cd6 ..., cd1 wavelet reconstruction signal R in B1
(t)=[r1(t),r2(t),…,r8(t)]T。
In the step C, ICA is decomposed between the temporal signatures of each component, spectrum signature and sequence similarity calculating method
It is as follows：
C1, using the R (t) in step B2 as FastICA input, using in FastICA tool boxes under Matlab softwares
Fastica function pair R (t) carry out ICA decomposition, obtain separation matrix W and 8 isolated component (s1,s2,…,s8)；
C2,8 isolated components obtained to step C1 calculate kurtosis value as the temporal signatures of each isolated component,In formula, siI-th of isolated component, i ∈ [1,8] are represented, E () is statistical expection function,In formula, L is isolated component siData volume,For isolated component siIn l-th of data point m
Power,WithFor isolated component i 4 rank squares and 2 rank squares；
Average band power of 8 isolated components that C3, calculation procedure C1 are obtained in 5 frequency band ranges：fspectral=
In [F (δ) F (θ) F (α) F (β) F (γ)], formula, F () is average band power；
C4, to 8 isolated component s in step C11,s2,…,s8The passage isolated component that sliding window time span is l is set
N is always obtained as a subsequence in dataep8 isolated components of passage finally just divide into individual subsequence, a sequential
Nep8 × N of individual subsequenceepIndividual isolated component, then calculates this 8 × NepSimilarity between the sequence of individual sub- phasesequence component,In formula, NepIt is the subsequence number in a sequential,It is respectively from subsequence j and k
In two isolated components calculatingWithBetween coefficient correlation,
In formula,WithRespectively two isolated componentsWithThe average of middle n data point.
In the step D, hierarchical clustering, the automatic identification of artefact and rejecting, FastICA inverse transformations and wavelet coefficient reconstruct
Specific method it is as follows：
D1, zscore function pair characteristic vectors F in Matlab software signal handling implement casees is utilized to carry out data standard
Change；Calculate the correlation distance in characteristic vector between any vector using linkage functions, using apart from minimum criteria as standard,
Clustering tree is calculated according to interior squared-distance method using linkage functions；
D2, according to step D1 cluster result 8 isolated component s are given automatically1,s2,…,s8Corresponding class label is sticked,
Zero setting processing is carried out to artefact component, remaining component keeps constant, obtains rejecting the isolated component matrix Z (t) of artefact component；
D3, using FastICA inverse transformations by Z (t) carry out projection mapping, obtain wavelet conversion coefficient U (t)=W-1Z (t),
It is reconstructed using the waverec function pair U (t) in wavelet transform tool box, obtains clean EEG signals.
The beneficial effects of the invention are as follows：
1st, solving the electric preprocess method of existing Mental imagery brain needs to train and artefact reference electrode, and can not be online
The problem of realizing.
2nd, the Mental imagery EEG signals of few passage (C3 and C4 passages) can be converted to by wavelet transform many
Signal is used for follow-up FastICA analysis all.
3rd, wavelet conversion coefficient is more stronger than the super-Gaussian of primary signal, and kurtosis is bigger, and ICA is carried out in iteration from wavelet field
There is significant advantage in terms of convergence of algorithm speed, anti-noise ability.
4th, can efficiently certainly using correlative character between time domain of the hierarchical clustering algorithm based on isolated component, frequency spectrum and sequence
It is dynamic to recognize and reject a variety of conventional artefacts.
Brief description of the drawings
Fig. 1 is the electric artefact elimination method flow chart of online brain of the invention；
Fig. 2 is the placement schematic diagram of brain wave acquisition electrode；
Fig. 3 carries out the experiment timing diagram of right-hand man's Mental imagery task for subject；
Fig. 4 is the time-domain diagram and spectrogram of the original EEG signals of C3 and C4 passages, wherein, (a) is that C3 and C4 passages are original
The time-domain diagram of EEG signals, (b) is the spectrogram of the original EEG signals of C3 passages；
Fig. 5 is the time-domain diagram and spectrogram of C3 and C4 passage EEG signals after correction, wherein, (a) is C3 after correction
With the time-domain diagram of C4 passage EEG signals, (b) is the spectrogram of C3 passage EEG signals after correction；
Fig. 6 is C3 channel Wavelet coefficient list branch reconstruction signals；
Fig. 7 is the oscillogram of each isolated component after C3 passages ICA is decomposed；
Fig. 8 is the temporal signatures and spectrum signature of eye electric component after C3 passages ICA is decomposed, wherein, (a) is eye electric component
Temporal signatures, (b) is the spectrum signature of eye electric component；
Fig. 9 is the time domain beamformer and spectrum signature of myoelectricity component after C3 passages ICA is decomposed, wherein, (a) is myoelectricity component
Time domain beamformer, (b) be myoelectricity component spectrum signature；
Figure 10 is the dendrogram of C3 passages each component progress hierarchical clustering；
Figure 11 is the EEG signals oscillogram of C3 and C4 passages after pretreatment.
Embodiment
The invention will be further described with specific embodiment below in conjunction with the accompanying drawings.
Embodiment 1：As shown in figs. 1-11, a kind of electric artefact elimination method of practical online brain, it is said that experimental design sum
It is bright as follows：
Set forth herein preprocess method have altogether and surveyed in right-hand man's Mental imagery brain electricity of 10 Healthy subjects
Examination, wherein subject is 6 male 4 female；It is dextro manuality, the range of age：19 years old~25 years old；Undergraduate course and master's educational background, without influence brain
Medical history and the brain wave acquisition experience of function, all subjects endorsed experimental study informed consent form before experiment.
This experiment data acquisition equipment for 16 lead EEG amplifiers (Mipower-UC, EEG Collection V2,
Tsing-Hua University's nerve engineering experiment room), 0Hz~250Hz signal bands, sample frequency is 1000Hz, 24 A/D converters, nothing
Notch filter；16 lead brains of lead of adopting international standards 10-20 system customizations electric cap (Ag-AgCL powder electrodes, Wuhan Green
Tyke Science and Technology Ltd.).The placement schematic diagram of brain wave acquisition electrode is as shown in Fig. 2 the main collection motor area C3 of experiment
With the EEG signals of two electrodes of C4, respectively x1And x (t)2(t)；Left mastoid process M1 electrodes for reference, Fpz is ground connection
Electrode；To ensure that the impedance between electrode and subject scalp is less than 5k Ω before experiment.
Experiment is carried out in quiet environment, and experiment timing diagram is as shown in figure 3, in experimentation, subject is sitting in comfortable
On seat, about 1 meter of eye distance computer screen, experiment start time has buzzing prompting every time, while screen centre occurs
"+", points out this experiment of subject to have begun to, the state continues 2s；Start prompting and terminate rear "+" disappearance, while screen is hit exactly
Centre can occur task picture prompting at random, point out subject followed by which kind of Mental imagery task, the state continues 1.5s；Carry
Screen centre occurs " * " after diagram piece disappears, the left hand before prompting subject is proceeded by suggested by picture/right hand fortune
Dynamic imagination task, this task time continues 6s, and task can enter resting state after terminating, and screen white screen, the state continues
Time is 3s.The stimulation normal form is by being widely used in the stimulation software E-prime (versions of biomedical engineering field
Version1.1) complete.The time of having a rest that subject is carried out between testing every time is 5 minutes, and each subject carries out 5 experiments,
Experiment includes 6 sequential every time, and all experiments of 10 subjects are completed in one day.
The present invention have chosen certain during subject 5 is tested at the 3rd time and once imagine that eeg data during right hand motion is used for subsequently
Analysis.
Comprised the following specific steps that for the method that above-mentioned eeg data result carries out the electric artefact rejecting of online brain：
A, the original EEG signals to C3, C4 passage carry out down-sampled, notch filter and linear drift is corrected；
B, the EEG signals to latter two passage of correction carry out multi-scale wavelet decomposition and wavelet coefficient list branch is reconstructed；
C, wavelet coefficient list branch reconstruction signal progress ICA decomposition respectively to two passages, calculate the time domain of each component
Correlation between feature, spectrum signature and sequence；
D, the isolated component to two passages carry out hierarchical clustering, and automatic identification simultaneously rejects artefact component, and to remaining
Isolated component carries out FastICA inverse transformations reconstruct and wavelet coefficient reconstruct, obtains clean EEG signals.
Down-sampled, notch filter and linear drift correction specific method in the step A is as follows：
A1, the original brain using the downsample functions in Matlab Matlab DSPToolBoxes respectively to C3, C4 passage
Electric signal progress 250Hz is down-sampled, to reduce data total amount, improves the real-time of data processing；
The time domain beamformer and spectrogram of latter two down-sampled passage are as shown in Figure 4, it can be seen that the number of two passages
According to obvious data wander and 50Hz Hz noise phenomenons；
A2, the electric number of the brain using dlsim function pair C3, the C4 passages in Matlab Matlab DSPToolBoxes after down-sampled
According to 50HZ notch filter processing is carried out, 50Hz Hz noises are rejected；
A3, using in Matlab Matlab DSPToolBoxes detrend function pairs trap processing after signal enter line
Property correction, eliminate the artefact that brings of data linear drift；
After A2 and the correction of the steps of A3 two, the time domain beamformer and spectrogram of two passages are as shown in Figure 5, it can be seen that
The data of two passages have eliminated 50Hz Hz noises and data drift phenomenon, but also there is serious EOG in data
With the artifacts such as EMG；
EEG signals X (t)=[x after being corrected by above-mentioned processing1(t),x2(t)]T∈RM×N, wherein：X(t)
It is one group of M dimension random vector, M is brain wave acquisition port number, and M is 2 here；T is transposition computing, x1And x (t)2(t) represent respectively
The EEG signals of C3 and C4 passages；T=1,2 ..., N, the N number of sampled point for being EEG signals X (t), N is 1500 here；By step
The eeg data that two passages distinguish 6s is obtained after rapid A processing.Analysis below is by taking C3 passages as an example, the processing side of C4 passages
Method and step are similarly.
In the step B, multi-scale wavelet decomposition and wavelet coefficient list branch reconstruct comprise the following steps that：
B1, selection " db4 " wavelet basis, utilize the wavedec function pairs in Matlab software wavelet transforms tool box
The signal x of C3 passages1(t) 7 layer scattering wavelet decompositions are carried out, signal x is obtained1(t) Coefficients of Approximation component ca7 and detail coefficients
Component cd1, cd2, cd3, cd4, cd5, cd6, cd7, now, single pass signal have reformed into the signal of multichannel；
The present invention have chosen the decomposition and reconstruction of Daubechies small echo C3 passage EEG signals.It is flat by maximum frequency response
Smooth frequency is portrayed, and the Daubechies wavelet basis of same order is not used in some current researchs；However, research shows, this
The 4 rank Daubechies small echos (db4) that invention is finally chosen preferably can correctly express EEG signal and spike.
B2, the low-and high-frequency wavelet systems obtained using the wrcoef function pairs previous step decomposition in wavelet transform tool box
Number carries out single branch reconstruct, obtains corresponding to wavelet coefficient ca7, cd7, cd6 ..., cd1 reconstruction signal R (t)=[r in B11
(t),r2(t),…,r8(t)]T, wherein, 8 wavelet reconstruction signals and x1(t) identical, its waveform such as Fig. 6 institutes of data points
Show.
In the step C, ICA is decomposed between the temporal signatures of each component, spectrum signature and sequence correlation calculations side
Method is as follows：
C1, using the wavelet reconstruction signal R (t) in step B2 as FastICA input, using under Matlab softwares
Fastica functions in FastICA tool boxes, separation matrix W is obtained by continuous iteration, set herein iteration precision as
0.0001.When weight vector reaches this precision, just terminate iterative process, weights now are final weight vector.Through
Iteration is crossed, the separation matrix W obtained herein is：
8 isolated component (s that R (t) is obtained after FastICA is decomposed1,s2,…,s8), its waveform is as shown in Figure 7；
C2, FastICA is decomposed respectively after obtained 8 isolated components calculate kurtosis value as each isolated component when
Characteristic of field,In formula, siRepresent i-th of isolated component, E () statistical expection
Function,In formula, L is isolated component siData volume, L=1500 here,For isolated component
siIn l-th of data point m powers, when m be 2 and 4 when just obtainWithIf kurt (si) value it is bigger, then table
Show that the DYNAMIC DISTRIBUTION of signal is more precipitous.Shown in the kurtosis value of each isolated component such as Fig. 8 (a), it can be seen that s1And s2
Kurtosis value come front two (kurt (s1)=11.08, kurt (s2)=8.878), the kurtosis value of noticeably greater than other components,
s3Kurtosis value be -0.7788, be the component that kurtosis value is uniquely negative value in 8 isolated components, and conventional numerous studies table
Bright, eye electricity artefact often has highest kurtosis value in the artefact composition of EEG signals, and noise artefact is used as sub-Gaussian signals, peak
Angle value is often negative value.
Average band power of 8 isolated components in 5 frequency band ranges after C3, respectively calculating FastICA decomposition：
fspectralIn=[F (δ) F (θ) F (α) F (β) F (γ)], formula, F () is average band power, and it passes through in Matlab softwares
Pwelch functions are calculated, using average band power as each isolated component spectrum signature；Wherein, s1Spectrum distribution such as
Shown in Fig. 8 (b), as can be seen from the figure：s1Frequency it is relatively low, be mainly distributed in the range of 0-12Hz, and with higher width
Value, s2Also show and s1Same spectrum signature；And there was only the electric artefact of eye in the artefact composition of EEG signals often frequency is most
Low, amplitude is larger；s6Time domain beamformer and spectrum distribution respectively as shown in Fig. 8 (a) and 8 (b), can from Fig. 8 (b)
Go out：s6Frequency spectrum be mainly distributed in 30-80Hz frequency band range, and in the range of this have of a relatively high amplitude, and
In the artefact composition of EEG signals, Muscle artifacts caused by the body kinematicses such as head rotation, hand shake often frequency
Highest (generally >=30Hz), amplitude are relatively large；
C4, in 8 isolated component s1,s2,…,s8It is middle that the C3 that sliding window time span is 1s (i.e. 250 sampled points) is set
Passage isolated component data are as a subsequence, and 8 isolated components of C3 passages finally just divide into 6 subsequences in a sequential
48 (6 × 8) individual isolated components, then calculate similarity between the sequence of this 48 sub- phasesequence components,
In formula, NepIt is the subsequence number in a sequential, N hereepFor 6,Calculated respectively from subsequence j and k
Two isolated components (With) between coefficient correlation,In formula,WithRespectively
For two isolated componentsWithThe average of middle n data point, n is 250 here；In this 48 sub- phasesequence components, when some
Subsequence component and most of subsequence components (the subsequence component for being set as 60% here) all show extremely low coefficient correlation
When (Wherein：| | for the computing that takes absolute value), the least correlativing coefficient in these subsequence components is taken as original
8 isolated component s1,s2,…,s8Similarity between the sequence of middle correspondence isolated component；When some subsequence components and most of sub- sequences
When row component (the subsequence component for being set as 60% here) all shows very high coefficient correlationTake these
Maximum correlation coefficient in subsequence component is used as 8 isolated component s of original1,s2,…,s8It is similar between the sequence of middle correspondence isolated component
Degree, so as to obtain 8 isolated component s of C3 passages in a sequential1,s2,…,s8Sequence between similarity characteristic vector.
Artefact occurs at random during brain wave acquisition, it is difficult to it is anticipated that and the sometimes only meeting in some subsequences
Go out to lose face the random artefact such as electric artefact or myoelectricity.And subsequence component and other subsequences without artefact comprising artefact composition
Component is often extremely low without similar feature or similarity, on the other hand, brain E-serial component that Mental imagery induces and its
The subsequence component of its useful EEG signals shows high similarity.
By above-mentioned calculating, the characteristic vector F=[f of one 7 dimension are obtained1,f2,f3,f4,f5,f6,f7]T, wherein f1To be each
The temporal signatures of individual isolated component, f2,f3,f4,f5,f6For spectrum signature, f7For similarity feature；
In the step D, to the 8 isolated component s obtained in C11,s2,…,s8Hierarchical clustering is carried out, automatic identification is simultaneously
Artefact is rejected, the specific side of FastICA inverse transformations reconstruct and wavelet coefficient reconstruct is then carried out to remaining isolated component successively
Method is as follows：
D1, first with the spy calculated in the zscore function pair steps C in Matlab software signal handling implement casees
Levy vectorial F and carry out data normalization so that the average value of eigenvectors matrix is zero, and standard deviation is 1；Then linkage is utilized
Function calculates the correlation distance between any vector in characteristic vector, here it is actual calculate be Euclidean between two vectors away from
From, next with apart from minimum criteria using linkage functions according to interior squared-distance method calculate clustering tree, finally utilize
Dendrogram functions draw out hierarchical clustering dendrogram as shown in Figure 10 according to the output of linkage functions；
D2, as shown in Figure 10, eye electric component s1And s2Due to being different from phase between time domain, frequency spectrum and the sequence of other isolated components
Close property feature and by auto-polymerization be a class, according to conventional research can with it is further seen that：s1For blink artefact, s2It is dynamic for eye
Artefact；Myoelectricity component s6One is individually polymerized to due to being different from similarity feature between the frequency spectrum and sequence of other isolated components
Class；Normal EEG signals component s4、s5、s7、s8A class is polymerized to first, although noise component(s) s3Be different from uniqueness when
Characteristic of field, but its general characteristic is similar with brain electricity, so last and normal brain activity voltolisation is for a class；8 are given while cluster
Individual isolated component sticks corresponding label, to artefact component s1、s2、s3And s6Zero setting processing is carried out, remaining component keeps constant,
Obtain the independent element matrix Z (t) of no artefact composition；
D3, using FastICA inverse transformations by Z (t) carry out projection mapping, obtain wavelet conversion coefficient U (t), i.e.,：
U (t)=W-1Z(t)
Wavelet basis function " db4 " is selected, using being picked in the waverec function pair steps D2 in wavelet transform tool box
Except the wavelet coefficient U (t) of various artefacts reconstructs EEG signals, as shown in figure 11, it can be seen that the EOG in EEG signals is pseudo-
Mark, EMG artefacts, 50Hz Hz noises, noise jamming and data drift phenomenon have all been eliminated well.
In order to further verify that the present invention rejects the beneficial effect of artefact, respectively by the method in the present invention and other methods
Performance during for this experiment is contrasted, and 10 subjects are carried using Li Mingai DWT-ICA, Bogdan Mijovie proposed
The EMD-ICA and DWICAC artefacts elimination method proposed by the present invention that go out and calculate signal after pretreatment frequency band energy it is special
Levy, the classification accuracy for being finally based on SVMs (Support vector machine, SVM) grader is as shown in table 1,
SVM is the machine learning method based on Statistical Learning Theory, is adapted to small sample classification problem, and generalization ability is strong, can fit well
The single test of Mental imagery EEG signals feature is answered to classify.Because the present invention has only used the eeg data of C3 and Cz passages to use
The frequency band energy feature of EEG signals is only extracted when analysis, classification, intrinsic dimensionality is relatively low, therefore the last choosing of the present invention
The linear SVM based on linear kernel is selected, error penalty factor value is 1.
Table 1 uses EMD-ICA, DWT-ICA, DWICAC artefact elimination method and the classification accuracy based on SVM classifier
It can be found that DWICAC methods proposed by the present invention are relative to EMD-ICA methods and DWT-ICA artefacts from table 1
For elimination method, almost automatic identification and all conventional artefacts (EOG, EMG and noise etc.) in brain electricity can be rejected；And dividing
This method is also significantly better than other two methods in terms of class accuracy rate, wherein：The present invention is for right-hand man's Mental imagery brain electricity
Average classification accuracy is 84.64%, and EMD-ICA methods and DWT-ICA methods 8.23% and 5.26%, best result are higher by respectively
Class accuracy rate is 89.27%, and EMD-ICA methods and DWT-ICA methods 7.64% and 3.55% are higher by respectively；It is above-mentioned to analyze into one
Step embodies superiority of the DWICAC methods in the online artefact of few passage Mental imagery brain electricity is rejected.
Above in conjunction with accompanying drawing to the present invention embodiment be explained in detail, but the present invention be not limited to it is above-mentioned
Embodiment, can also be before present inventive concept not be departed from the knowledge that those of ordinary skill in the art possess
Put that various changes can be made.
Claims (5)
1. a kind of electric artefact elimination method of practical online brain, it is characterised in that：Comprise the following steps that：
A, the original EEG signals to each passage carry out down-sampled, notch filter and linear drift is corrected；
B, the EEG signals to each passage after correction carry out multi-scale wavelet decomposition and wavelet coefficient list branch is reconstructed；
C, wavelet coefficient list branch reconstruction signal progress ICA decomposition respectively to each passage, calculate temporal signatures, the frequency of each component
Similarity feature between spectrum signature and sequence；
D, the isolated component to each passage carry out hierarchical clustering respectively, and automatic identification simultaneously rejects artefact component, and to remaining only
Vertical component carries out FastICA inverse transformations reconstruct and wavelet coefficient reconstruct, obtains clean EEG signals.
2. the electric artefact elimination method of practical online brain according to claim 1, it is characterised in that：In the step A
Down-sampled, notch filter and linear drift correction specific method are as follows：
A1, using the downsample functions in Matlab Matlab DSPToolBoxes respectively to C3, C4 passage in real time original brain
It is down-sampled that electric signal carries out 250Hz；
A2, entered using eeg data of dlsim function pair C3, the C4 passages in Matlab Matlab DSPToolBoxes after down-sampled
The processing of row 50HZ notch filters；
A3, using in Matlab Matlab DSPToolBoxes detrend function pairs trap processing after signal linearly floated
Shift correction；
EEG signals X (t)=[x after being corrected by above-mentioned processing1(t),x2(t) ..., xM(t)]T, wherein：X (t) is
One group of M ties up random vector, and M is brain wave acquisition port number, and T is transposition computing, and t=1,2 ..., N, N adopts for EEG signals X's (t)
Sampling point number.
3. the electric artefact elimination method of practical online brain according to claim 1, it is characterised in that：It is many in the step B
Multi-scale wavelet is decomposed and comprising the following steps that wavelet coefficient list branch is reconstructed：
B1, selection " db4 " wavelet basis, it is a certain using the wavedec function pairs in Matlab software wavelet transforms tool box
The signal x of passage1(t) 7 layer scattering wavelet decompositions are carried out, signal x is obtained1(t) Coefficients of Approximation component ca7 and detail coefficients point
Measure cd7, cd6, cd5, cd4, cd3, cd2, cd1；
B2, using the wrcoef functions in wavelet transform tool box respectively to the Coefficients of Approximation component that is obtained in step B1 and
Detail coefficients component carries out single branch reconstruct, obtains corresponding to ca7, cd7, cd6 ... in B1, and cd1 wavelet reconstruction signal R (t)=
[r1(t),r2(t),…,r8(t)]T。
4. the electric artefact elimination method of practical online brain according to claim 1, it is characterised in that：In the step C,
ICA is decomposed between the temporal signatures of each component, spectrum signature and sequence, and similarity calculating method is as follows：
C1, using the R (t) in step B2 as FastICA input, using in FastICA tool boxes under Matlab softwares
Fastica function pair R (t) carry out ICA decomposition, obtain separation matrix W and 8 isolated component (s1,s2,…,s8)；
C2,8 isolated components obtained to step C1 calculate kurtosis value as the temporal signatures of each isolated component,In formula, siI-th of isolated component, i ∈ [1,8] are represented, E () is statistical expection function,In formula, L is isolated component siData volume,For isolated component siIn l-th of data point m
Power,WithRespectively isolated component si2 rank central moments and 4 rank central moments；
Average band power of 8 isolated components that C3, calculation procedure C1 are obtained in 5 frequency band ranges：fspectral=[F (δ)
F (θ) F (α) F (β) F (γ)], in formula, F () is average band power；
C4, to 8 isolated component s in step C11,s2,…,s8The passage isolated component data that sliding window time span is l are set
As a subsequence, N is always obtainedep8 isolated components of passage finally just divide into N in individual subsequence, a sequentialepIt is individual
8 × N of subsequenceepIndividual isolated component, then calculates this 8 × NepSimilarity between the sequence of individual sub- phasesequence component,In formula, NepIt is the subsequence number in a sequential,It is respectively from subsequence j and k
In two isolated components calculatingWithBetween coefficient correlation,
In formula,WithRespectively two isolated componentsWithThe average of middle n data point.
By above-mentioned calculating, the characteristic vector F=[f of one 7 dimension are obtained1,f2,f3,f4,f5,f6,f7]T, wherein f1It is only for each
The temporal signatures of vertical component, f2,f3,f4,f5,f6For spectrum signature, f7For similarity feature.
5. the electric artefact elimination method of practical online brain according to claim 1, it is characterised in that：In the step D, layer
The specific method of secondary cluster, the automatic identification of artefact and rejecting, FastICA inverse transformations and wavelet coefficient reconstruct is as follows：
D1, zscore function pair characteristic vectors F in Matlab software signal handling implement casees is utilized to carry out data normalization；Profit
The correlation distance in characteristic vector between any vector is calculated with linkage functions, as standard, to be utilized apart from minimum criteria
Linkage functions calculate clustering tree according to interior squared-distance method；
D2, according to step D1 cluster result 8 isolated component s are given automatically1,s2,…,s8Corresponding class label is sticked, to puppet
Mark component carries out zero setting processing, and remaining component keeps constant, obtains rejecting the isolated component matrix Z (t) of artefact component；
D3, using FastICA inverse transformations by Z (t) carry out projection mapping, obtain wavelet conversion coefficient U (t)=W-1Z (t), is utilized
Waverec function pair U (t) in wavelet transform tool box are reconstructed, and obtain clean EEG signals.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710381917.4A CN107260166A (en) | 2017-05-26 | 2017-05-26 | A kind of electric artefact elimination method of practical online brain |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710381917.4A CN107260166A (en) | 2017-05-26 | 2017-05-26 | A kind of electric artefact elimination method of practical online brain |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107260166A true CN107260166A (en) | 2017-10-20 |
Family
ID=60064677
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710381917.4A Pending CN107260166A (en) | 2017-05-26 | 2017-05-26 | A kind of electric artefact elimination method of practical online brain |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107260166A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107981862A (en) * | 2017-11-30 | 2018-05-04 | 华南理工大学 | A kind of brain signal online denoising method under magnetic nuclear resonance environment |
| CN108388855A (en) * | 2018-02-11 | 2018-08-10 | 宁波工程学院 | A kind of pin electrode EMG Signal Decomposition Based method |
| CN108520239A (en) * | 2018-04-10 | 2018-09-11 | 哈尔滨理工大学 | A kind of Method of EEG signals classification and system |
| CN108937968A (en) * | 2018-06-04 | 2018-12-07 | 安徽大学 | The Conduction choice method of emotion EEG signals based on independent component analysis |
| CN109009092A (en) * | 2018-06-15 | 2018-12-18 | 东华大学 | A method of removal EEG signals noise artefact |
| CN109157214A (en) * | 2018-09-11 | 2019-01-08 | 河南工业大学 | A method of the online removal eye electricity artefact suitable for single channel EEG signals |
| CN109497997A (en) * | 2018-12-10 | 2019-03-22 | 杭州妞诺科技有限公司 | Based on majority according to the seizure detection and early warning system of acquisition |
| CN109620218A (en) * | 2019-01-29 | 2019-04-16 | 杭州妞诺科技有限公司 | Brain wave intelligence screening method and system |
| CN110074783A (en) * | 2019-05-17 | 2019-08-02 | 杭州师范大学 | The cerebral cortex excitability of transcranial magnetic stimulation induction signal and imaging and quantization method |
| CN110859616A (en) * | 2019-12-12 | 2020-03-06 | 科大讯飞股份有限公司 | Cognitive assessment method, device and equipment of object and storage medium |
| CN111685759A (en) * | 2020-05-12 | 2020-09-22 | 河北大学 | P, T characteristic wave detection method of electrocardiosignals |
| CN111856958A (en) * | 2020-07-27 | 2020-10-30 | 西北大学 | Intelligent household control system, control method, computer equipment and storage medium |
| CN112464902A (en) * | 2020-12-16 | 2021-03-09 | 杭州电子科技大学 | Electroencephalogram blink artifact detection method based on multi-channel multi-dimensional feature optimization |
| CN112741637A (en) * | 2020-12-23 | 2021-05-04 | 杭州国辰迈联机器人科技有限公司 | P300 electroencephalogram signal extraction method, cognitive rehabilitation training method and system |
| CN113011506A (en) * | 2021-03-24 | 2021-06-22 | 华南理工大学 | Texture image classification method based on depth re-fractal spectrum network |
| CN114469135A (en) * | 2022-01-17 | 2022-05-13 | 北京脑陆科技有限公司 | Artifact removing method and device for multi-lead electroencephalogram signal and brain-computer interface |
| CN116602691A (en) * | 2023-07-14 | 2023-08-18 | 北京元纽科技有限公司 | Denoising method and device for electroencephalogram signals, electronic equipment and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006271876A (en) * | 2005-03-30 | 2006-10-12 | Kyoto Univ | Spike sorting apparatus, method and program, and recording medium |
| CN102697493A (en) * | 2012-05-03 | 2012-10-03 | 北京工业大学 | Method for rapidly and automatically identifying and removing ocular artifacts in electroencephalogram signal |
| CN102835955A (en) * | 2012-09-08 | 2012-12-26 | 北京工业大学 | Method of automatically removing ocular artifacts from electroencephalogram signal without setting threshold value |
| CN103006211A (en) * | 2013-01-17 | 2013-04-03 | 西安电子科技大学 | Map mapping device based on brain electrical activity network analysis |
| CN103690163A (en) * | 2013-12-21 | 2014-04-02 | 哈尔滨工业大学 | ICA (independent component analysis) and HHT (Hilbert-Huang transform) fusion based automatic electrooculogram interference eliminating method |
| WO2015040532A2 (en) * | 2013-09-19 | 2015-03-26 | Tata Consultancy Services Limited | System and method for evaluating a cognitive load on a user corresponding to a stimulus |
| CN105326499A (en) * | 2015-08-19 | 2016-02-17 | 兰州大学 | Portable electroencephalogram collection system |
-
2017
- 2017-05-26 CN CN201710381917.4A patent/CN107260166A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006271876A (en) * | 2005-03-30 | 2006-10-12 | Kyoto Univ | Spike sorting apparatus, method and program, and recording medium |
| CN102697493A (en) * | 2012-05-03 | 2012-10-03 | 北京工业大学 | Method for rapidly and automatically identifying and removing ocular artifacts in electroencephalogram signal |
| CN102835955A (en) * | 2012-09-08 | 2012-12-26 | 北京工业大学 | Method of automatically removing ocular artifacts from electroencephalogram signal without setting threshold value |
| CN103006211A (en) * | 2013-01-17 | 2013-04-03 | 西安电子科技大学 | Map mapping device based on brain electrical activity network analysis |
| WO2015040532A2 (en) * | 2013-09-19 | 2015-03-26 | Tata Consultancy Services Limited | System and method for evaluating a cognitive load on a user corresponding to a stimulus |
| CN103690163A (en) * | 2013-12-21 | 2014-04-02 | 哈尔滨工业大学 | ICA (independent component analysis) and HHT (Hilbert-Huang transform) fusion based automatic electrooculogram interference eliminating method |
| CN105326499A (en) * | 2015-08-19 | 2016-02-17 | 兰州大学 | Portable electroencephalogram collection system |
Non-Patent Citations (1)
| Title |
|---|
| 王娆芬等: "基于ICA和FCM算法的操作员脑电信号眼电伪迹自动去除研究", 《PROCEEDINGS OF THE 30TH CHINESE CONTROL CONFERENCE》 * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107981862A (en) * | 2017-11-30 | 2018-05-04 | 华南理工大学 | A kind of brain signal online denoising method under magnetic nuclear resonance environment |
| CN107981862B (en) * | 2017-11-30 | 2020-06-19 | 华南理工大学 | Online denoising method for brain signals in nuclear magnetic resonance environment |
| CN108388855A (en) * | 2018-02-11 | 2018-08-10 | 宁波工程学院 | A kind of pin electrode EMG Signal Decomposition Based method |
| CN108388855B (en) * | 2018-02-11 | 2022-02-11 | 宁波工程学院 | Needle electrode electromyographic signal decomposition method |
| CN108520239B (en) * | 2018-04-10 | 2021-05-07 | 哈尔滨理工大学 | Electroencephalogram signal classification method and system |
| CN108520239A (en) * | 2018-04-10 | 2018-09-11 | 哈尔滨理工大学 | A kind of Method of EEG signals classification and system |
| CN108937968A (en) * | 2018-06-04 | 2018-12-07 | 安徽大学 | The Conduction choice method of emotion EEG signals based on independent component analysis |
| CN109009092A (en) * | 2018-06-15 | 2018-12-18 | 东华大学 | A method of removal EEG signals noise artefact |
| CN109009092B (en) * | 2018-06-15 | 2020-06-02 | 东华大学 | Method for removing noise artifact of electroencephalogram signal |
| CN109157214A (en) * | 2018-09-11 | 2019-01-08 | 河南工业大学 | A method of the online removal eye electricity artefact suitable for single channel EEG signals |
| CN109497997A (en) * | 2018-12-10 | 2019-03-22 | 杭州妞诺科技有限公司 | Based on majority according to the seizure detection and early warning system of acquisition |
| CN109620218A (en) * | 2019-01-29 | 2019-04-16 | 杭州妞诺科技有限公司 | Brain wave intelligence screening method and system |
| CN110074783B (en) * | 2019-05-17 | 2021-07-20 | 杭州师范大学 | Cerebral cortex excitability of transcranial magnetic stimulation induced signal and imaging and quantifying method |
| CN110074783A (en) * | 2019-05-17 | 2019-08-02 | 杭州师范大学 | The cerebral cortex excitability of transcranial magnetic stimulation induction signal and imaging and quantization method |
| CN110859616A (en) * | 2019-12-12 | 2020-03-06 | 科大讯飞股份有限公司 | Cognitive assessment method, device and equipment of object and storage medium |
| CN111685759A (en) * | 2020-05-12 | 2020-09-22 | 河北大学 | P, T characteristic wave detection method of electrocardiosignals |
| CN111856958A (en) * | 2020-07-27 | 2020-10-30 | 西北大学 | Intelligent household control system, control method, computer equipment and storage medium |
| CN112464902A (en) * | 2020-12-16 | 2021-03-09 | 杭州电子科技大学 | Electroencephalogram blink artifact detection method based on multi-channel multi-dimensional feature optimization |
| CN112464902B (en) * | 2020-12-16 | 2024-02-09 | 杭州电子科技大学 | Electroencephalogram blink artifact detection method based on multichannel multidimensional feature optimization |
| CN112741637A (en) * | 2020-12-23 | 2021-05-04 | 杭州国辰迈联机器人科技有限公司 | P300 electroencephalogram signal extraction method, cognitive rehabilitation training method and system |
| CN113011506A (en) * | 2021-03-24 | 2021-06-22 | 华南理工大学 | Texture image classification method based on depth re-fractal spectrum network |
| CN113011506B (en) * | 2021-03-24 | 2023-08-25 | 华南理工大学 | Texture image classification method based on deep fractal spectrum network |
| CN114469135A (en) * | 2022-01-17 | 2022-05-13 | 北京脑陆科技有限公司 | Artifact removing method and device for multi-lead electroencephalogram signal and brain-computer interface |
| CN116602691A (en) * | 2023-07-14 | 2023-08-18 | 北京元纽科技有限公司 | Denoising method and device for electroencephalogram signals, electronic equipment and storage medium |
| CN116602691B (en) * | 2023-07-14 | 2023-10-10 | 北京元纽科技有限公司 | Denoising method and device for electroencephalogram signals, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107260166A (en) | A kind of electric artefact elimination method of practical online brain | |
| Chen et al. | A novel EEMD-CCA approach to removing muscle artifacts for pervasive EEG | |
| Javadi et al. | Classification of ECG arrhythmia by a modular neural network based on mixture of experts and negatively correlated learning | |
| CN106709469B (en) | Automatic sleep staging method based on electroencephalogram and myoelectricity multiple characteristics | |
| Lestari et al. | Wavelet based-analysis of alpha rhythm on eeg signal | |
| CN110269609B (en) | Method for separating ocular artifacts from electroencephalogram signals based on single channel | |
| Fathima et al. | Wavelet based features for epileptic seizure detection | |
| Shedeed | A new method for person identification in a biometric security system based on brain EEG signal processing | |
| CN107007291A (en) | Intense strain intensity identifying system and information processing method based on multi-physiological-parameter | |
| Srivastava et al. | DWT-based feature extraction from ECG signal | |
| Quazi et al. | Artifacts removal from EEG signal: FLM optimization-based learning algorithm for neural network-enhanced adaptive filtering | |
| Liu et al. | Removal of EMG Artifacts from Multichannel EEG Signals Using Combined Singular Spectrum Analysis and Canonical Correlation Analysis. | |
| Desai et al. | Decision support system for arrhythmia beats using ECG signals with DCT, DWT and EMD methods: A comparative study | |
| CN106236080A (en) | Based on the removing method of myoelectricity noise in multichannel EEG signals | |
| CN112426162A (en) | Fatigue detection method based on electroencephalogram signal rhythm entropy | |
| Degirmenci et al. | Emotion recognition from EEG signals by using empirical mode decomposition | |
| CN113723557A (en) | Depression electroencephalogram classification system based on multiband time-space convolution network | |
| Gu et al. | AOAR: an automatic ocular artifact removal approach for multi-channel electroencephalogram data based on non-negative matrix factorization and empirical mode decomposition | |
| Al-Taee et al. | Optimal feature set for finger movement classification based on sEMG | |
| CN116340824A (en) | Electromyographic signal action recognition method based on convolutional neural network | |
| CN113967022B (en) | Individual self-adaption-based motor imagery electroencephalogram characteristic characterization method | |
| CN113269048B (en) | Motor imagery electroencephalogram signal classification method based on deep learning and mixed noise data enhancement | |
| Assi et al. | Kmeans-ICA based automatic method for ocular artifacts removal in a motorimagery classification | |
| Hsu | Wavelet-coherence features for motor imagery EEG analysis posterior to EOG noise elimination | |
| Gurrala et al. | Artifact removal from sleep hipnograms using complete ensemble empirical decomposition with adaptive noise |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171020 |
|
| RJ01 | Rejection of invention patent application after publication |