CN104605845B - Electroencephalogram signal processing method based on DIVA model - Google Patents
Electroencephalogram signal processing method based on DIVA model Download PDFInfo
- Publication number
- CN104605845B CN104605845B CN201510050938.9A CN201510050938A CN104605845B CN 104605845 B CN104605845 B CN 104605845B CN 201510050938 A CN201510050938 A CN 201510050938A CN 104605845 B CN104605845 B CN 104605845B
- Authority
- CN
- China
- Prior art keywords
- brain
- dipole
- data
- signal processing
- signal
- 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.)
- Expired - Fee Related
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 18
- 238000000034 method Methods 0.000 claims abstract description 42
- 238000012545 processing Methods 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 3
- 210000004556 brain Anatomy 0.000 claims description 25
- 239000011159 matrix material Substances 0.000 claims description 25
- 210000004761 scalp Anatomy 0.000 claims description 16
- 230000004913 activation Effects 0.000 claims description 14
- 239000000203 mixture Substances 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 13
- 210000003128 head Anatomy 0.000 claims description 12
- 238000005457 optimization Methods 0.000 claims description 11
- 238000009826 distribution Methods 0.000 claims description 10
- 230000005684 electric field Effects 0.000 claims description 10
- 230000004927 fusion Effects 0.000 claims description 10
- 230000009466 transformation Effects 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 5
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 4
- 230000005672 electromagnetic field Effects 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 4
- 238000002156 mixing Methods 0.000 claims description 4
- 210000002615 epidermis Anatomy 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005481 NMR spectroscopy Methods 0.000 claims 1
- 230000005611 electricity Effects 0.000 claims 1
- 238000002922 simulated annealing Methods 0.000 claims 1
- 238000004458 analytical method Methods 0.000 abstract description 8
- 238000011160 research Methods 0.000 abstract description 6
- 238000012880 independent component analysis Methods 0.000 abstract description 4
- 210000005036 nerve Anatomy 0.000 abstract description 4
- 230000007547 defect Effects 0.000 abstract description 2
- 238000002474 experimental method Methods 0.000 abstract description 2
- 230000035945 sensitivity Effects 0.000 abstract description 2
- 238000007499 fusion processing Methods 0.000 abstract 1
- 230000035479 physiological effects, processes and functions Effects 0.000 abstract 1
- 230000006870 function Effects 0.000 description 12
- 238000004088 simulation Methods 0.000 description 11
- 238000013507 mapping Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 8
- 210000000056 organ Anatomy 0.000 description 7
- 230000001537 neural effect Effects 0.000 description 5
- 238000003062 neural network model Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 210000003625 skull Anatomy 0.000 description 3
- 230000002087 whitening effect Effects 0.000 description 3
- 238000007476 Maximum Likelihood Methods 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 210000003710 cerebral cortex Anatomy 0.000 description 2
- 210000001175 cerebrospinal fluid Anatomy 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000004615 ingredient Substances 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000007935 neutral effect Effects 0.000 description 2
- 238000003786 synthesis reaction Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 206010029216 Nervousness Diseases 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000000137 annealing Methods 0.000 description 1
- 235000019606 astringent taste Nutrition 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000007177 brain activity Effects 0.000 description 1
- 210000001638 cerebellum Anatomy 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000004020 conductor Substances 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 210000004704 glottis Anatomy 0.000 description 1
- 230000011731 head segmentation Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000003387 muscular Effects 0.000 description 1
- 230000001379 nervous effect Effects 0.000 description 1
- 230000036403 neuro physiology Effects 0.000 description 1
- 238000002610 neuroimaging Methods 0.000 description 1
- 230000000926 neurological effect Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 210000005182 tip of the tongue Anatomy 0.000 description 1
- 210000001519 tissue Anatomy 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
Landscapes
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses an electroencephalogram signal processing method based on a DIVA model. According to the electroencephalogram signal processing method, fMIR data generated through the DIVA model in a simulating mode are used for carrying out positioning analysis on electroencephalogram signals, the complexity of actual computing is simplified through an independent component analysis method, and the defects that non-intrusive electroencephalogram signals are low in resolution ratio and high in interference are overcome. The fMIR data generated through the DIVA model are used for carrying out fusion processing on electroencephalogram data, and the problems that the electroencephalogram signals are low in space resolution ratio, high in signal interference and low in signal-noise ratio are resolved greatly. Through the pre-processing of ICA, the computing complexity is reduced, and the sensitivity of an equivalent dipole positioning algorithm to the noise is overcome greatly. Finally, the electroencephalogram signal processing method is used for processing actual experiment data, and the obtained conclusion meets the physiology reality. A feasible resolution scheme is provided for the electroencephalogram signal processing problem in a Chinese nerve analysis system, and a foundation is laid for the research related to generation and obtaining of Chinese phonetic symbols in future.
Description
Technical field
The invention discloses a kind of brain-electrical signal processing method based on diva model, it is related to EEG Processing technology neck
Domain.
Background technology
Brain-computer interface (brain computer interface, bci) be a kind of realized based on EEG signals human brain with
The system that computer or other electronic equipment are communicated and controlled, it is one and leads to not against peripheral nervouss and muscular tissue etc.
The communication system of normal brain output channel.In other words, bci is that the direct of foundation between human brain and computer exchanges and control
Passage processed, by this passage, people just directly can express idea or commanding apparatus by brain, without language or limbs
Action.
The computer MSR Information system that Fu Langkegangse (frank guenther) professor of Boston University leads is in brain-computer interface
Succeed in developing in technical foundation and be developed into a kind of neural analysis system (neuralynx system).Neural analysis system is by brain
Computer interface (bci) and speech synthesis system diva (directions into velocities of articulators)
Model two parts form.This system can thinking process " reading " just in people's brain out, be then converted into normal
Language is stated in real time.
Diva model is then a kind of neutral net with regard to speech production and acquisition with biological significance.Diva model
It is made up of feedforward subsystem, feedback control subsystem and front field (maeda) simulation sound channel.Model is by certain rule
While using voice as input, produce a rate of articulation and the time-varying series of organ site change, apply this sequence
Row, the ideal required for system can be obtained by is pronounced.
One key character of diva model is exactly that each to model component and cerebral cortex relevant range is corresponded.
Model component is root on the basis of relevant neuroanatomy and neuro physiology research with the mapping relations in cerebral cortex region
Determine according to mni (montreal neurological institute) standard reference system.Carrying out sending out according to parameter preset
During sound task, diva model can produce the fmir data of reaction brain area state of activation.So, diva model is actually just constituted
One in order to explain from the related data of various researchs and to carry out concordance description to regard to voice nerve processing procedure
Basic framework.
During voice correlational study, the neuroimaging art of commonly used non-intrusion type exists as the interface between brain machine
Tested on the person.Although the device of this non-intrusion type is conveniently worn on human body, because skull declines to signal
Subtract effect and the dispersion of electromagnetic wave that neuron is sent and blurring effect, the resolution that recorded signal is not high.This letter
Number ripple still can be detected, but Signal-to-Noise is relatively low, and the requirement to post processing is higher.
Content of the invention
The technical problem to be solved is: for the defect of prior art, provide a kind of based on diva model
Brain-electrical signal processing method.The fmir data that the method is generated by diva modeling carries out positioning analysises to EEG signals,
And simplified real using Independent Component Analysis (independent component correlation algorithm, ica)
The complexity that border calculates, overcomes the shortcoming that non-intrusion type EEG signals resolution is low, interference is big.
The present invention is to solve above-mentioned technical problem to employ the following technical solutions:
A kind of brain-electrical signal processing method based on diva model, the functional core being produced using diva neural network model
MR data carries out fusion treatment to EEG signals, comprises the following steps:
Step one, the EEG signals data passed through in non-intrusion type brain-computer interface extraction phonation, to the brain collecting
Electrical signal data carries out pretreatment;
Step 2, based on diva model provide user interface channel parameters are configured, make diva model carry out mould
Send out sound, obtain fmir data;
Will fmir data input count drawing instrument spm in be analyzed, using 12 parameters affine transformation to fmir number
According to being normalized, then the image after processing and high-resolution structural images are carried out registering, and be normalized to the sky of mni
Between, then do space smoothing process using three-dimensional Gaussian function pair through the data of above-mentioned process;
Step 3, the EEG signals data processing through step 2 is carried out with preliminary albefaction or spheroidising, remove each
Dependency between observation signal, is then processed to EEG signals data using ica method, draws optimal transform matrix,
And then recover signal source matrix, isolate effective composition from EEG signals data, each effective composition corresponds to an idol
Extremely son becomes;
Step 4, combine head model and its coefficient of conductivity, optimum idol is solved to the effective composition that each extracts
Extremely son configuration is so that the scalp Electric Field Distribution being produced by above-mentioned dipole and the scalp Electric Field Distribution measuring are under mean square meaning
Minimum, using the activation point position in fmir data as find dipole position seed point, fmir data limit source can
Globally optimal solution can be solved in space;
Step 5, by after fusion treatment result output.
As present invention further optimization scheme, in described step 3, using ica method, EEG signals data is carried out
Process specific as follows:
Set the Scalp Potential x of recordiAs observation vector, wherein, i=1,2 ... .., m, m represent the quantity of electrode,
Observation vector xiIt is n time upper independent signal source sjLinear mixing, wherein, j=1,2 ..., n, each sjIt is all system
Count independence and have its fixing spatial information weight aj, ajIt is the jth row of lead-field matrix a;
Under above-mentioned setting, directly find the transformation matrix w of optimum using ica method, and recover signal source matrix m;Make
With the signal y estimating, observation signal x is rebuild:
X=w-1y (1)
J-th time series of x can be expressed from the next:
Wherein,It is w-1Line n m row element.
As present invention further optimization scheme, in described step 3, using ica method, EEG signals data is carried out
Process, carry out following settings:
301st, the signal that independent signal source produces is statistical iteration;
302nd, the EEG signals data observing is instantaneous linear mixed signal;
303rd, the quantity in independent signal source is less than the quantity of electrode.
As present invention further optimization scheme, the detailed process of described step 4 is as follows:
Set the electromagnetic field observation signal of brain epidermis and the source signal of the internal any position of brain approx linearly closes
System, is expressed from the next:
X=as+n (3)
Wherein, x is the observation signal of scalp electrode record, and a is lead-field matrix, and s is dipole vector, and n represents each
The vector of the noise composition in electrode receipt signal;
Lead-field matrix a is the nonlinear function of dipole position, brain geometry and the medium coefficient of conductivity, it
Each row expression contribution to left end observation signal x positioned at the source of the unit strength of a certain position;
Optimization problem be equivalent to solve following formula:
Wherein, c is signal to noise ratio normalization matrix, and λ is regularization parameter;
Object function is associated with the spatial property of solution, objective function is:
In formula, p item is a kind of space constraint to solution, that is, combine the possible space position that fmir limits source, fixed by following formula
Justice:
Wherein, l is the quantity of dipole;riIt is the position vector of i-th dipole;siIt is the space letter of fmir activation point
Breath;C and diIt is all constant, the position of dipole is related to the activation point position of fmir, the activation point position of fmir is
Find the seed point of dipole position;
Application ultimate range limits position and the dipole moment of matching dipole, here it is the equivalent dipole limited by fmir
Model, abbreviation fc-ecd model.One basic assumption of fc-ecd is: the Electric Field Distribution being observed can be by one or several
Individual dipole, produces, and the activation in a certain size region simultaneously can be by single dipole subrepresentation.
After setting up object function, the parameter of dipole is regarded as one of higher dimensional space node, then using simulation
Annealing algorithm seeks globally optimal solution.
As present invention further optimization scheme, the preprocessing process described in step one includes: removal nictation artefact,
Ocular movement artefact, low-pass filtering, bad electrode reset, average and baseline correction.
As present invention further optimization scheme, the Gaussian function described in step 2 needs to meet its halfwidth and is
12mm*12mm*24mm.
As present invention further optimization scheme, the dipole arrangement described in step 4 includes position, direction and big
Little.
As present invention further optimization scheme, exporting the result after fusion treatment described in step 5, is logical
Cross and exported using fit workbox in matlab.
As present invention further optimization scheme, described constant diValue be 5mm.
The present invention adopts above technical scheme compared with prior art, has following technical effect that and is produced using diva model
Raw fmir data carries out fusion treatment to eeg data, overcomes that EEG signals spatial resolution is low, signal interference is big, noise
Than very low problem.By the pretreatment of ica, reduce the complexity of computing, be largely overcoming equivalent dipole
Location algorithm is for the sensitivity of noise.Finally, using this method, true experimental data is processed, the conclusion obtaining meets life
Of science true.This method is that the EEG Processing problem in Chinese neural analysis system provides feasible solution, is
Chinese speech is generated and is laid a good foundation with acquisition correlational study from now on.
Brief description
Fig. 1 is the structural representation of diva neural network model;
Fig. 2 is Chinese neural analysis system structure diagram;
Fig. 3 is existing independent component analysis schematic diagram;
Fig. 4 is preferable 4 layers of head model.
Specific embodiment
Below in conjunction with the accompanying drawings technical scheme is described in further detail:
The technical problem to be solved in the present invention is to provide a kind of brain-electrical signal processing method based on diva model.Including with
Lower step:
Step 1, the EEG signals (eeg) passed through in non-intrusion type brain-computer interface extraction phonation.To the eeg collecting
Data carries out pretreatment, including removal nictation artefact, ocular movement artefact, low-pass filtering, the reset of bad electrode, average and baseline
Correction;
Step 2, based on diva model provide user interface channel parameters are configured, so that model is simulated respectively
Pronunciation, obtains fmir data.Then it is analyzed in data result input statistics drawing instrument spm simulation being produced, adopt
The affine transformation of 12 parameters is normalized to data.Again image and high-resolution structural images are carried out registering, and
It is normalized to the space of mni.Then after utilizing the three-dimensional Gaussian function pair standardization that halfwidth (fwhm) is 12mm*12mm*24mm
Data do space smoothing process;
Step 3, preliminary albefaction or spheroidising are carried out to data, except the dependency between each observation signal.Then profit
With ica method, eeg data is processed, effective component extracting;
In the measurement of eeg signal, the signal being observed is actually by some relatively independent source signal superpositions
Become.Therefore, we be will be helpful to using the independent element that ica method decomposites observation signal and hold really significant brain
Activity.In eeg signal processing, can be by the Scalp Potential x of recordi, i=1,2 ... .., m represent as observation vector, m
The quantity of electrode.Assume observation signal vector xiIt is n time upper independent signal source sj, j=1,2 ..., the linear mixing of n.
Each sjIt is all statistical iteration, and have its fixing spatial information weight aj.ajIt is the jth row of lead-field matrix a.?
Under statement above, ica method directly finds the transformation matrix w of optimum, and the principle using ica recovers signal source matrix m.Cause
This, reconstruction observation signal x being carried out from the signal y estimating can be carried out by following formula:
X=w-1y (1)
Therefore, j-th time series of x can be expressed from the next:
Wherein,It is w-1Line n m row element.In the data handling procedure of eeg, in order that using ica method,
Typically all make the following assumptions:
1st, the signal that signal source produces is statistical iteration;
2nd, the EEG signals observing are instantaneous linear mixed signal;
3rd, the quantity in independent signal source is less than the quantity of electrode
As shown in formula (2), ica method isolates effective composition from numerous eeg measurement signals, and each effectively becomes
A corresponding dipole is divided to become.
Step 4, to each extract effective components utilising fmir data carry out fusion treatment;
The electromagnetic field of human brain can be explained by the Max Wei Er equation of half stable state.Theoretical, brain epidermis according to this
The internal any position of electromagnetic field observation signal and brain source signal approx linear, can be expressed from the next:
X=as+n (3)
Wherein, x is the observation signal of scalp electrode record, and a is referred to as lead-field matrix (lead field matrix), and s is
Dipole vector, n represents the vector of the noise composition in each electrode receipt signal.Lead-field matrix a is dipole position, big
Brain geometry and the nonlinear function of the medium coefficient of conductivity, its each row represent the unit strength positioned at a certain position
The contribution to left end observation signal x for the source.
Eeg Reverse Problem refers to known head model and the coefficient of conductivity, finds optimum dipole arrangement and (includes position
Put, direction, size) so that the scalp Electric Field Distribution being produced by these dipoles and the scalp Electric Field Distribution measuring are mean square
Minimum under meaning.It is equivalent to solve following optimization problem:
Wherein, c is signal to noise ratio normalization matrix, and λ is regularization parameter.More generally, can be object function and solution
Certain spatial property connects, and with objective function can be accordingly:
In formula, p item is certain space constraint to solution, that is, combines the possible space position that fmir limits source, permissible
It is defined by the formula:
Wherein, l is the quantity of dipole;riIt is the position vector of i-th dipole;siIt is the space letter of fmir activation point
Breath;c,diIt is constant, herein, di=5mm.The position of dipole is related to the activation point position of fmir, and fmir swashs
Point position alive can be regarded as finding the seed point of dipole position.Meanwhile, application ultimate range limits the position of matching dipole
Put and dipole moment, here it is the equivalent dipole model limited by fmir, abbreviation fc-ecd model.One of fc-ecd is substantially false
If being: the Electric Field Distribution being observed can be produced by one or several dipoles, and the activation in a certain size region can simultaneously
With by single dipole subrepresentation.
After setting up object function, the parameter of dipole is regarded as one of higher dimensional space node, is moved back using simulation
Fiery algorithm seeks globally optimal solution.
Step 5, in matlab using fit workbox by after fusion treatment result export.
A specific embodiment of the present invention be given below:
Step 1, the EEG signals (eeg) passed through in non-intrusion type brain-computer interface extraction Chinese vowel phonation.To adopting
The eeg data collecting carries out pretreatment, resets, puts down including removal nictation artefact, ocular movement artefact, low-pass filtering, bad electrode
All and baseline correction;
Step 2, the user interface being provided based on diva model by first three formant frequency value for 805hz, 1265hz and
2770hz, makes model carry out Chinese vowel simulation pronunciation respectively, obtains fmir data.Then data result simulation being produced
It is analyzed in input statistics drawing instrument spm, the affine transformation using 12 parameters is normalized to data.To scheme again
As carrying out registering with high-resolution structural images, and it is normalized to the space of mni.Then halfwidth (fwhm) is utilized to be 12mm*
Data after the three-dimensional Gaussian function pair standardization of 12mm*24mm is done space smoothing and is processed;
Step 3, preliminary albefaction or spheroidising are carried out to data, except the dependency between each observation signal.Using only
Vertical component analyzing method (ica) draws optimal transform matrix, and then recovers signal source matrix s.From numerous eeg measurement signals
Isolate effective composition, each effective composition corresponding dipole becomes.
Step 4, on the premise of known head model and its coefficient of conductivity, the effective composition that each extracts is asked
The optimum dipole arrangement of solution (including position, direction, size) is so that the scalp Electric Field Distribution that produced by these dipoles and survey
The scalp Electric Field Distribution measured is minimum under mean square meaning.Using the activation point position of fmir as the kind being searching dipole position
Sub-, limit in fmir and in the possible space in source, solve globally optimal solution.
Step 5, in matlab using fit workbox by after fusion treatment result export.
As shown in figure 1, diva neural network model is mainly by feedforward subsystem, feedback control subsystem and sound channel
Constituted.Described feedforward subsystem includes: a part for voice mapping ensemblen, cerebellum, phonatory organ speed and position are reflected
Penetrate collection (part);Feedback control subsystem includes a part for voice mapping ensemblen, audition error map collection, hearing status mapping
Collection, somatesthesia error map collection, somatesthesia position mapping ensemblen, phonatory organ speed and position mapping ensemblen;Sound channel adopts front field sound channel mould
Type, is divided into 8 ingredients simulation sound channel: the position of tongue, the shape of tongue, the tip of the tongue, lip height, lip be convex and jaw,
The opening and closing degree of the height of throat and glottis.The effect of voice mapping ensemblen is mould according to the mapping relations having existed in model
The input of type carries out mapping code;The effect of hearing status and error map collection is to adjust fundamental frequency f0 by coding with first three altogether
Shake peak frequency f1~f3 position describing current pronunciation;The effect of phonatory organ speed and position mapping ensemblen is determining diva
The position of each phonatory organ in the sound channel framework of neural network model;The effect of somatesthesia state and error map collection is adjustment pronunciation
The position of all parts of organ and parameter are currently pronounced to adjust.
As shown in Fig. 2 Chinese neural analysis system is made up of two parts: torsion free modules and speech synthesis system
Diva model.In bci, EEG signals obtain by using the wireless nerve electrode that person wears, and the nerve signal detecting is then
It is used for driving continuous " moving " of voice operation demonstrator, provide real-time voice output for user.Voice operation demonstrator diva model
It is then a kind of neutral net with regard to speech production and acquisition with biological significance.Diva model is by feedforward subsystem
System, feedback control subsystem and front field (maeda) simulation sound channel are formed.Model is by certain rule using voice as defeated
While entering, produce a rate of articulation and the time-varying series of organ site change, apply this sequence, system just can obtain
To required ideal pronunciation.
As shown in figure 3, ica method be based on it is assumed hereinafter that: signal source produce signal be statistical iteration;Observe
EEG signals be instantaneous linear mixed signal independent signal source quantity be less than electrode quantity.
Generally, the data being obtained all has dependency, thus be usually required data is carried out preliminary white
Change or spheroidising, because whitening processing can remove the dependency between each observation signal, thus simplifying follow-up isolated component
Extraction process, and it is generally the case that data carries out whitening processing compared with not carrying out whitening processing to data, algorithm
Better astringency.They are separated by input signal to pass through solution mixing system afterwards, export effective ingredient.The pretreatment of ica can
To be effectively reduced the complexity of problem, and algorithm is simple.The research of ica algorithm can be divided into the iteration based on information theory criterion
Method of estimation and be based on statistical algebraic method two big class, from principle for, they are all the independence that make use of source signal
Property and non-Gaussian system.Based in information-theoretical technique study, scholars from maximum entropy, Minimum mutual information, maximum likelihood and bear
Entropy maximization angularly proposes a series of algorithm for estimating.As fastica algorithm, infomax algorithm, maximum- likelihood estimation
Deng.The Higher-Order Cumulants such as second-order cumulant, fourth order cumulant are mainly had based on statistical method.
As shown in figure 4, this patent adopts preferable 4 layers of head model as the simulation of true brain head model, will
Head segmentation becomes 4 parts: brain, cerebrospinal fluid, skull and scalp, and it provides a simplified model of brain volume conductor.Very
Many research worker using data such as actual image datas (ct/mri), using fem (finite element method) or
Bem (boundary element method) method, is modeled to head model;Then, more pre- by segmentation, registration etc.
Head is divided into 4 parts (brain, cerebrospinal fluid, skull and scalp) by processing procedure again.But, these processes need substantial amounts of calculating
And long time.And adopt 4 layer models a maximum benefit be exactly save the calculating time on the premise of, using the teaching of the invention it is possible to provide
One approximate well for true head model.There are some researches show: even if 4 layers of spherical model and more real FEM (finite element) model
Compare, an effective simulation tool of still can yet be regarded as, it can provide rational head table within 10%~20% for the error
Current potential is estimated.As shown in figure 4, the radius of adopt herein 4 layers of spherical model is 79mm, 81mm, 85mm and 88mm successively, its conduction
Coefficient is respectively 0.461s/m, 1.39s/m, 0.0058s/m and 0.461s/m.Eeg forward problem can be carried out according to formula (3)
Calculate.The solution of forward problem can not only provide the head table Potential Distributing of simulation in the emulation experiment below, and can be
The effect that auxiliary is inferred is played in the iterative process solving Reverse Problem.
Above in conjunction with accompanying drawing, embodiments of the present invention are explained in detail, but the present invention is not limited to above-mentioned enforcement
Mode, in the ken that those of ordinary skill in the art possess, can also be on the premise of without departing from present inventive concept
Make a variety of changes.
Claims (9)
1. a kind of brain-electrical signal processing method based on diva model is it is characterised in that utilize the feature that diva model produces
Nuclear magnetic resonance data carries out fusion treatment to EEG signals, comprises the following steps:
Step one, the EEG signals data passed through in non-intrusion type brain-computer interface extraction phonation, to the brain telecommunications collecting
Number carries out pretreatment;
Step 2, based on diva model provide user interface channel parameters are configured, make diva model be simulated send out
Sound, obtains fmri data;
Fmri data input is counted in drawing instrument spm and is analyzed, the affine transformation using 12 parameters is entered to fmri data
Row normalized, then the image after processing and high-resolution structural images are carried out registering, and it is normalized to the space of mni,
Then do space smoothing process using three-dimensional Gaussian function pair through the data of above-mentioned normalized;
Step 3, the EEG signals data processing through step one is carried out with preliminary albefaction or spheroidising, remove each observation
Dependency between signal, is then processed to EEG signals data using ica method, draws the transformation matrix of optimum, enters
And recover signal source matrix, isolate effective composition from EEG signals data, each effective composition corresponds to a dipole
Son;
Step 4, combine head model and its coefficient of conductivity, optimum dipole is solved to the effective composition that each extracts
Configuration so that the scalp Electric Field Distribution being produced by above-mentioned dipole and the scalp Electric Field Distribution measuring under mean square meaning
Little, using the activation point position in fmri data as the seed point finding dipole position, limit the possibility in source in fmri data
Globally optimal solution is solved in space;
Step 5, by after fusion treatment result output.
2. as claimed in claim 1 a kind of brain-electrical signal processing method based on diva model it is characterised in that described step
In three, EEG signals data is carried out process specific as follows using ica method:
Set the Scalp Potential x of recordi, i=1,2 ..., as observation vector, m represents the quantity of electrode, observation vector x for .., mi
It is n time upper independent signal source sj, j=1,2 ..., n linear mixing, each sjIt is all statistical iteration and have it
Fixing spatial information weight aj, ajIt is the jth row of lead-field matrix a;
Under above-mentioned setting, directly find the transformation matrix w of optimum using ica method, and recover signal source matrix m;Using estimating
The signal y of meter rebuilds to observation signal x:
X=w-1y (1)
J-th time series of x can be expressed from the next:
Wherein,It is w-1M row n-th row element.
3. as claimed in claim 2 a kind of brain-electrical signal processing method based on diva model it is characterised in that described step
In three, using ica method, EEG signals data is processed, carries out following settings:
301st, the signal that independent signal source produces is statistical iteration;
302nd, the EEG signals data observing is instantaneous linear mixed signal;
303rd, the quantity of independent signal source is less than the quantity of electrode.
4. as claimed in claim 3 a kind of brain-electrical signal processing method based on diva model it is characterised in that described step
Four detailed process is as follows:
The electromagnetic field observation signal setting brain epidermis is approx linear with the source signal of the internal any position of brain, by
Following formula represents:
X=as+n (3)
Wherein, x is the observation signal of scalp electrode record, and a is referred to as lead-field matrix, and s is dipole vector, and n represents each electricity
The vector of the noise composition in the receipt signal of pole;
Lead-field matrix a is the nonlinear function of dipole position, brain geometry and the medium coefficient of conductivity, it each
Row expression contribution to left end observation signal x positioned at the source of the unit strength of a certain position;
Optimization problem be equivalent to solve following formula:
Wherein, c is signal to noise ratio normalization matrix, and λ is regularization parameter;
The spatial property of object function and solution is connected, objective function is:
In formula, p item is a kind of space constraint to solution, that is, combine the possible space position that fmri limits source, be defined by the formula:
Wherein, l is the quantity of dipole;riIt is the position vector of i-th dipole;siIt is the spatial information of fmri activation point;c,
diIt is constant, the position of dipole is related to the activation point position of fmri, the activation point position of fmri is searching dipole
The seed point of sub- position;
After setting up object function, the parameter of dipole is regarded as one of higher dimensional space node, calculated using simulated annealing
Method seeks globally optimal solution.
5. a kind of brain-electrical signal processing method based on diva model as described in claim 1 or 4 is it is characterised in that step
Preprocessing process described in one includes: remove nictation artefact, ocular movement artefact, low-pass filtering, bad electrode reset, average with
And baseline correction.
6. a kind of brain-electrical signal processing method based on diva model as described in claim 1 or 4 is it is characterised in that step
It is 12mm*12mm*24mm that Gaussian function described in two needs to meet its halfwidth.
7. a kind of brain-electrical signal processing method based on diva model as described in claim 1 or 4 is it is characterised in that step
Dipole arrangement described in four includes position, direction and size.
8. a kind of brain-electrical signal processing method based on diva model as described in claim 1 or 4 is it is characterised in that step
Exporting the result after fusion treatment described in five, is by being exported using fit workbox in matlab.
9. as claimed in claim 4 a kind of brain-electrical signal processing method based on diva model it is characterised in that described constant
diValue be 5mm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510050938.9A CN104605845B (en) | 2015-01-30 | 2015-01-30 | Electroencephalogram signal processing method based on DIVA model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510050938.9A CN104605845B (en) | 2015-01-30 | 2015-01-30 | Electroencephalogram signal processing method based on DIVA model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104605845A CN104605845A (en) | 2015-05-13 |
CN104605845B true CN104605845B (en) | 2017-01-25 |
Family
ID=53140705
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510050938.9A Expired - Fee Related CN104605845B (en) | 2015-01-30 | 2015-01-30 | Electroencephalogram signal processing method based on DIVA model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104605845B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR3038997A1 (en) * | 2015-07-13 | 2017-01-20 | Univ Pierre Et Marie Curie (Paris 6) | DATA PROCESSING DEVICE WITH VALUE REPRESENTATION THROUGH INTERVALS OF TIME BETWEEN EVENTS |
CN105054928A (en) * | 2015-07-17 | 2015-11-18 | 张洪振 | Emotion display equipment based on BCI (brain-computer interface) device electroencephalogram acquisition and analysis |
CN105266804B (en) * | 2015-11-12 | 2017-10-17 | 杭州电子科技大学 | A kind of brain-electrical signal processing method based on low-rank and sparse matrix decomposition |
CN107961007A (en) * | 2018-01-05 | 2018-04-27 | 重庆邮电大学 | A kind of electroencephalogramrecognition recognition method of combination convolutional neural networks and long memory network in short-term |
CN108742602A (en) * | 2018-03-26 | 2018-11-06 | 校宝在线(杭州)科技股份有限公司 | A kind of stammerer detection method based on brain wave |
CN111513711B (en) * | 2020-05-22 | 2021-06-04 | 电子科技大学 | Electroencephalogram bad lead interpolation method based on reference electrode |
CN111616704A (en) * | 2020-06-24 | 2020-09-04 | 天津大学 | Research method for brain dynamic function network interaction mode of sentence listening task |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP3735668B2 (en) * | 2002-11-27 | 2006-01-18 | 独立行政法人情報通信研究機構 | Brain function measuring device and brain function measuring method |
EP1758714A4 (en) * | 2004-06-14 | 2009-03-11 | Cephos Corp | Question and control paradigms for detecting deception by measuring brain activity |
CN102880906B (en) * | 2012-07-10 | 2015-01-28 | 南京邮电大学 | Chinese vowel pronunciation method based on DIVA nerve network model |
CN103310272B (en) * | 2013-07-02 | 2016-06-08 | 南京邮电大学 | Based on the DIVA neural network model manner of articulation that sound channel action knowledge base is improved |
-
2015
- 2015-01-30 CN CN201510050938.9A patent/CN104605845B/en not_active Expired - Fee Related
Also Published As
Publication number | Publication date |
---|---|
CN104605845A (en) | 2015-05-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104605845B (en) | Electroencephalogram signal processing method based on DIVA model | |
WO2023040588A1 (en) | Graph model-based brain function registration method | |
CN105956624B (en) | Mental imagery brain electricity classification method based on empty time-frequency optimization feature rarefaction representation | |
CN103345749B (en) | A kind of brain network function connectivity lateralization detection method based on modality fusion | |
CN105054928A (en) | Emotion display equipment based on BCI (brain-computer interface) device electroencephalogram acquisition and analysis | |
CN109614885A (en) | A kind of EEG signals Fast Classification recognition methods based on LSTM | |
CN109784023B (en) | Steady-state vision-evoked electroencephalogram identity recognition method and system based on deep learning | |
Marzetti et al. | The use of standardized infinity reference in EEG coherency studies | |
CN110090017A (en) | A kind of EEG signals source localization method based on LSTM | |
CN109344816A (en) | A method of based on brain electricity real-time detection face action | |
CN111728590A (en) | Individual cognitive ability prediction method and system based on dynamic function connection | |
CN109199376B (en) | Decoding method of motor imagery electroencephalogram signal based on OA-WMNE brain source imaging | |
Cui et al. | EEG source localization using spatio-temporal neural network | |
CN110013248A (en) | Brain electricity tensor mode identification technology and brain-machine interaction rehabilitation system | |
Wei et al. | Edge sparse basis network: a deep learning framework for EEG source localization | |
CN114052668A (en) | Brain function analysis method based on magnetoencephalogram data | |
WO2024083059A1 (en) | Working memory task magnetoencephalography classification system based on machine learning | |
Jinliang et al. | EEG emotion recognition based on granger causality and capsnet neural network | |
CN111539509B (en) | Virtual digital brain construction method and system and intelligent robot control system | |
CN102508184A (en) | Brain function active region detection method based on moving average time series models | |
CN101502413A (en) | Method for developing test for neurosychiatric disease | |
CN102743166B (en) | Source positioning method of event-related potential | |
CN104537711B (en) | A kind of white matter fiber parameterized model construction method based on healthy population | |
Tang et al. | Eye movement prediction based on adaptive BP neural network | |
CN115813409A (en) | Ultra-low-delay moving image electroencephalogram decoding method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20200512 Address after: 710000 25 / F, block D, Tsinghua Science Park, Keji 2nd Road, Zhangba Street office, hi tech Zone, Xi'an City, Shaanxi Province Patentee after: Xi'an Huinao Intelligent Technology Co.,Ltd. Address before: 210003 Gulou District, Jiangsu, Nanjing new model road, No. 66 Patentee before: NANJING UNIVERSITY OF POSTS AND TELECOMMUNICATIONS |
|
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170125 |