CN106991409A - A kind of Mental imagery EEG feature extraction and categorizing system and method - Google Patents
A kind of Mental imagery EEG feature extraction and categorizing system and method Download PDFInfo
- Publication number
- CN106991409A CN106991409A CN201710244075.8A CN201710244075A CN106991409A CN 106991409 A CN106991409 A CN 106991409A CN 201710244075 A CN201710244075 A CN 201710244075A CN 106991409 A CN106991409 A CN 106991409A
- Authority
- CN
- China
- Prior art keywords
- mental imagery
- classification
- module
- training
- 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.)
- Granted
Links
- 230000003340 mental effect Effects 0.000 title claims abstract description 112
- 238000000605 extraction Methods 0.000 title claims abstract description 65
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000012549 training Methods 0.000 claims abstract description 78
- 238000012360 testing method Methods 0.000 claims abstract description 75
- 210000004556 brain Anatomy 0.000 claims abstract description 38
- 230000009467 reduction Effects 0.000 claims abstract description 30
- 238000006243 chemical reaction Methods 0.000 claims abstract description 21
- 238000004364 calculation method Methods 0.000 claims abstract description 7
- 239000011159 matrix material Substances 0.000 claims description 45
- 239000000203 mixture Substances 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 3
- 230000000284 resting effect Effects 0.000 description 14
- 230000008859 change Effects 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 3
- 238000004070 electrodeposition Methods 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000002123 temporal effect Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007177 brain activity Effects 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000033764 rhythmic process Effects 0.000 description 2
- 208000012661 Dyskinesia Diseases 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 230000003920 cognitive function Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000001054 cortical effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000005562 fading Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000003205 muscle Anatomy 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 210000001428 peripheral nervous system Anatomy 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000035807 sensation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
- G06F3/015—Input arrangements based on nervous system activity detection, e.g. brain waves [EEG] detection, electromyograms [EMG] detection, electrodermal response detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Neurology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Neurosurgery (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Dermatology (AREA)
- Human Computer Interaction (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Signal Processing (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present invention relates to a kind of Mental imagery EEG feature extraction and categorizing system and method, this method is based on a kind of Mental imagery EEG feature extraction and categorizing system, and the specific steps of this method include：Training to Mental imagery EEG signals：The Mental imagery EEG signals that brain machine interface system is gathered enter row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, dictionary learning and classification histogram calculation as training data, obtain the classification histogram of training dictionary and training data；Test to Mental imagery EEG signals：The Mental imagery EEG signals that brain machine interface system is gathered enter row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction as test data, the sparse expression of test data is obtained using the training dictionary in training module, obtain the classification histogram of test data, and the histogrammic comparing result progress test data classification of classification histogram and the classification of test data according to training data.
Description
Technical field
The invention belongs to the technical field of EEG Processing, more particularly to a kind of Mental imagery EEG feature extraction
With categorizing system and method.
Background technology
At present, brain-computer interface technology, as a kind of brand-new communication and control technology, can be to have a normal thinking but have serious
The patient of dyskinesia provides communication and environmental Kuznets Curves means.Wherein, brain-computer interface be defined as making one independent of
Peripheral nervous system and muscle and with external world's equipment for being communicated or being controlled.Brain-computer interface technology is not only applicable to non-patient
Communication and environmental Kuznets Curves are provided, automatically controlling, the scientific domain such as military cognitive also have potential application value.In view of its
Huge application prospect, brain-computer interface has caused the great attention of international scientific circle, referred to as brain science, rehabilitation project, biology
Engineering in medicine and a study hotspot in Human-machine Control field.
In the signal of all reflection brain activities that can be monitored to, due to EEG signals
(Electroencephalogram, EEG) has preferable temporal resolution, and monitoring instrument is simpler, the advantages of non-intrusive, quilt
Most of brain machine interface system is adopted.
In all brain machine interface systems based on EEG signal, can be used as the brain activity signal of control signal has：
VEP, P300 event related potentials, Mental imagery, cortical slow potential etc. are several.Wherein, believe for Mental imagery
Number, the theoretical foundation of the brain-computer interface based on Mental imagery is：When people carries out conscious activity in certain, corticocerebral correspondence area
Domain is active, and the μ rhythm and the low frequency part of β ripples that these regions are produced will appear from amplitude fading, i.e. event correlation is gone
Synchronous (event-related desynchronization, ERD)；At a time, corticocerebral regional area not by
To the excitation of conscious pattern, the EEG signals EEG local parts in the region will appear from the enhancing of amplitude, i.e. event-related design
(event-related synchronization, ERS).Brain machine interface system based on ERD/ERS is mainly discrimination motion and thought
The EEG signal of image thought operation, such as imagine left hand, the right hand, pin, the motion of tongue, so as to produce different control commands.
At present, the research based on Mental imagery brain machine interface system is one of study hotspot of current brain-computer interface.Mainly
Reason is that the Physiological Bases and mathematical modeling of ERS/ERD phenomenons are studied and proved by many scientific research institutions, and as spy at present
Beg for discriminative sensations, motion and the most commonly used method of cognitive function under normal and pathological state.In addition, the μ rhythm in the phenomenon
Do not need environmental stimuli to induce with the change of β ripples, be easy to subject to train and control, and then study most popular as brain-computer interface
Realization means.
But, being currently based on the brain-computer interface of Mental imagery, there are the following problems in terms of feature extraction and classifying：
(1) classification species is few：Feature extraction and classifying based on different motion imaginative thinking operation is very difficult, current energy
The most Mental imagery Mental tasks enough distinguished are six kinds.Classification species limits the application of brain-computer interface less；
(2) nicety of grading is not high：With the increase of Mental imagery species, nicety of grading declines therewith；
(3) work and resting state can not be distinguished at any time：Brain-computer interface is to user's long periods of wear, it is necessary to continue work
Make, then corresponding brain machine interface system is required to distinguish the Mental imagery state and resting state of user, when user can not possibly be long
Between be in Mental imagery state of a control.Current brain machine interface system does not possess the function also.
In summary, species of classifying for the brain-computer interface based on Mental imagery in terms of feature extraction and classifying is few, divide
The problem of class precision is low and can not distinguish work and resting state at any time, still lacks effective solution.
The content of the invention
The present invention in order to solve the above problems, overcome in the prior art the brain-computer interface based on Mental imagery in feature extraction
With classification in terms of exist classification species it is few, nicety of grading is low and can not distinguish work and resting state at any time the problem of there is provided
A kind of Mental imagery EEG feature extraction and categorizing system and method.
To achieve these goals, the present invention is adopted the following technical scheme that：
A kind of Mental imagery EEG feature extraction and categorizing system, the system include training module and test module；
The Mental imagery EEG signals that the training module is configured as gathering brain machine interface system are used as training data
Enter row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, dictionary learning and classification histogram calculation, obtain training word
Allusion quotation and the histogrammic module of the classification of training data；
The Mental imagery EEG signals that the test module is configured as gathering brain machine interface system are used as test data
Enter row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, obtain testing number using the training dictionary in training module
According to sparse expression, obtain the classification histogram of test data, and classification histogram according to training data and test data
The histogrammic comparing result of classification carries out the module of test data classification.
Further, the training module includes the first EEG signals memory module, and first EEG signals store mould
The Mental imagery EEG signals that block gathers brain machine interface system are stored as training data；
The test module includes the second EEG signals memory module, and the second EEG signals memory module connects brain machine
The Mental imagery EEG signals of port system collection are stored as test data.
The present invention in order to solve the above problems, overcome in the prior art the brain-computer interface based on Mental imagery in feature extraction
With classification in terms of exist classification species it is few, nicety of grading is low and can not distinguish work and resting state at any time the problem of there is provided
A kind of Mental imagery EEG feature extraction and categorizing system and method.
To achieve these goals, the present invention is adopted the following technical scheme that：
A kind of Mental imagery EEG feature extraction and sorting technique, this method are based on a kind of Mental imagery EEG signals
Feature extraction and classifying system, the specific steps of this method include：
(1) to the training of Mental imagery EEG signals：The Mental imagery EEG signals that brain machine interface system is gathered as
Training data enters row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, dictionary learning and classification histogram calculation, obtains
To training dictionary and the classification histogram of training data；
(2) to the test of Mental imagery EEG signals：The Mental imagery EEG signals that brain machine interface system is gathered as
Test data enters row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, is obtained using the training dictionary in training module
To the sparse expression of test data, the classification histogram of test data, and the classification histogram according to training data and survey are obtained
The histogrammic comparing result of classification for trying data carries out test data classification.
Further, enter row format conversion to training data in the step (1) to concretely comprise the following steps：
The Mental imagery EEG signals of collection are converted into three dimensional signal by 2D signal：First EEG signals are stored into mould
The Mental imagery EEG signals X ∈ IR of the brain machine interface system collection stored in blockn×m×TBuild two dimensional matrix.
In the analyzing and training to Mental imagery EEG signals, foundation can describe the Two-Dimensional Moment of EEG signals spatial information
The Mental imagery EEG signals of collection are converted to three dimensional signal by battle array by 2D signal so that during EEG signals description, no
Temporal information is only included, spatial information can be also included.
Further, feature point extraction and characteristic vector calculating are concretely comprised the following steps in the step (1)：
(1-1) critical point detection：The motion of the brain machine interface system collection stored in first EEG signals memory module is thought
As EEG signals X ∈ IRn×m×TIt is expressed as X=[X1,X2,...,XT]；
Key point is detected using Harris Keypoint detectors for each value in X, for all keys detected
Point extracts the module that size is (η × η × T)；
(1-2) characteristics extraction：Average is gone for the module all values in step (1-1), for each inside module
Individual pixel calculates its variance M along time orientation2, skewness M3With kurtosis M4, constitute time matrix M corresponding with moduler, r=
{ 2,3,4 }, wherein Mr=[mij], i, j=1,2 ..., η
Wherein, vijtIt is t-th timeslice of the position at { i, j }；
(1-3) characteristic vector is calculated：Each time matrix in step (1-2) is converted into a length for η2To
Amount, time matrix Mr, r three value composition characteristic vector m ∈ IR in r={ 2,3,4 }d, wherein d=3 η2For crucial point module
Characteristic vector length,
Further, dimensionality reduction carries out dimensionality reduction using accidental projection (random projection) method in the step (1)
Processing, is concretely comprised the following steps：
The matrix that note describes Mental imagery characteristic vector m is D ∈ IRd×p, wherein, d is the characteristic vector of crucial point module
Length, p is the quantity of crucial point module；
In the subspace that the vector projection for being d by dimension is n to a dimension, wherein n ＜ ＜ d；
It is multiplied by a random matrix R to realize the dimension-reduction treatment of eigenvectors matrix by matrix D, by eigenvectors matrix
Dimension is reduced to n × p：
Y=RD
Wherein, Y ∈ IRn×pFor the eigenvectors matrix after dimensionality reduction, R is accidental projection matrix, R ∈ IRn×d, average is 0,
Variance is 1.
Further, the dictionary learning in the step (1) is passed according to the eigenvectors matrix Y after dimensionality reduction by compressing
Sense obtains Y sparse expression, concretely comprises the following steps：
Using K-SVD algorithms, the dictionary Φ ∈ IR of following equation are metn×m(m ＞ n) and sparse expression formulaxi∈IRm, wherein, xiInclude k (k<<N) individual or less nonzero element：
Wherein, | | | |FIt is Frobenius norms, | | | |0It is l0Semi-norm, calculates the non-zero entry included in vector
Element.
Further, in the step (2) test data sparse expression calculate concretely comprise the following steps：
By test data in step (2) respectively through form conversion, feature point extraction, characteristic vector are calculated, dimensionality reduction is obtained
Dimensionality reduction after eigenvectors matrix Q, with reference to the dictionary Φ obtained in step (1),
According to：
Obtain the sparse expression X of test dataQ。
Further, the test data in the classification histogram and the step (2) of the training data in the step (1)
Classification histogram basis respectively：
Obtain, the classification histogram of the training data in the step (1) is hi, the training data in the step (2)
Classification histogram be hQ；
What test data was classified in the step (2) concretely comprises the following steps：
According to：
Determine the classification of the test data.
Beneficial effects of the present invention：
1. classification species is improved：A kind of Mental imagery EEG feature extraction and the categorizing system of the present invention and
Method, in the training to Mental imagery EEG signals of step (1), different Mental tasks can set up a common word
Allusion quotation Φ, in the test to Mental imagery EEG signals of step (2), as long as separating existing Mental task in training department, is being surveyed
The category can be just detected during examination, therefore, the classification species of such a algorithm is improved；
2. Mental imagery can be detected with resting state：A kind of Mental imagery EEG feature extraction of the present invention
With categorizing system and method, resting state is regarded into a kind of Mental task, the state is taken into account when setting up dictionary Φ, then exists
In follow-up test, when there is resting state, it can also detect that user is currently in resting state；
3. classification speed is very fast：A kind of Mental imagery EEG feature extraction of the present invention and categorizing system and method,
Different Mental tasks sets up a common dictionary, and different Mental tasks can disposably obtain classification knot in testing
Really.
Brief description of the drawings
Fig. 1 is flow chart of the method for the present invention；
Fig. 2 is the international 10-20 electrode positions schematic diagram of the present invention.
Embodiment：
It is noted that described further below is all exemplary, it is intended to provide further instruction to the application.Unless another
Indicate, all technologies used herein and scientific terminology are with usual with the application person of an ordinary skill in the technical field
The identical meanings of understanding.
It should be noted that term used herein above is merely to describe embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singulative
It is also intended to include plural form, additionally, it should be understood that, when in this manual using term "comprising" and/or " bag
Include " when, it indicates existing characteristics, step, operation, device, component and/or combinations thereof.
In the case where not conflicting, the feature in embodiment and embodiment in the application can be mutually combined.Tie below
Closing accompanying drawing, the invention will be further described with embodiment.
Embodiment 1：
As background technology is introduced, the brain-computer interface based on Mental imagery is in feature extraction and classifying in the prior art
Classification species that aspect is present is few, transport there is provided a kind of the problem of nicety of grading is low and can not distinguish work and resting state at any time
Dynamic imagination EEG feature extraction and categorizing system and method.
In a kind of Mental imagery EEG feature extraction of the application and the embodiment of categorizing system,
To achieve these goals, the present invention is adopted the following technical scheme that：
A kind of Mental imagery EEG feature extraction and categorizing system, the system include training module and test module；
The Mental imagery EEG signals that the training module is configured as gathering brain machine interface system are used as training data
Enter row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, dictionary learning and classification histogram calculation, obtain training word
Allusion quotation and the histogrammic module of the classification of training data；
The Mental imagery EEG signals that the test module is configured as gathering brain machine interface system are used as test data
Enter row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, obtain testing number using the training dictionary in training module
According to sparse expression, obtain the classification histogram of test data, and classification histogram according to training data and test data
The histogrammic comparing result of classification carries out the module of test data classification.
The training module includes the first EEG signals memory module, and the first EEG signals memory module connects brain machine
The Mental imagery EEG signals of port system collection are stored as training data；
The test module includes the second EEG signals memory module, and the second EEG signals memory module connects brain machine
The Mental imagery EEG signals of port system collection are stored as test data.
To achieve these goals, the present invention is adopted the following technical scheme that：
A kind of Mental imagery EEG feature extraction and sorting technique, this method are based on a kind of Mental imagery EEG signals
Feature extraction and classifying system, as shown in figure 1, the specific steps of this method include：
(1) to the training of Mental imagery EEG signals：The Mental imagery EEG signals that brain machine interface system is gathered as
Training data enters row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, dictionary learning and classification histogram calculation, obtains
To training dictionary and the classification histogram of training data；
In the step (1) brain machine interface system gather Mental imagery EEG signals as training data specific steps
For：
User wears the Mental imagery that EEG electrode caps carry out different mental operation, for example, imagine left hand motion, right hand fortune
Dynamic, it is X=[X to gather corresponding EEG signalsL:XR] and be stored in the first EEG signals memory module.
Enter row format conversion to training data in the step (1) to concretely comprise the following steps：
The Mental imagery EEG signals of collection are converted into three dimensional signal by 2D signal：First EEG signals are stored into mould
The Mental imagery EEG signals X ∈ IR of the brain machine interface system collection stored in blockn×m×TBuild two dimensional matrix.
In order to embody the spatial information of EEG signal, the EEG signal collected is stored as X ∈ IRn×m×T, wherein M=n × m
It is the two dimensional matrix built to embody EEG spatial informations, its foundation is international 10-20 electrode positions figure, such as Fig. 2
It is shown.If the 10-20 electrode position figures of standard, then the two dimensional matrix M built is 5*5 two-dimensional matrix, accordingly
Value on position is the EEG signals magnitude of voltage measured on the electrode in different time points, and the position for being 0 then represents that the point does not have
There is measurement signal, the value of the point is always 0.T is the time point of collection, if acquisition time is 1 second, frequency acquisition is 128, then T
=128.
The three dimensional signal for being then by such a data conversion, the EEG signals measured, can also regard the image of one as
Change over time and the video recording changed.
In the analyzing and training to Mental imagery EEG signals, foundation can describe the Two-Dimensional Moment of EEG signals spatial information
The Mental imagery EEG signals of collection are converted to three dimensional signal by battle array by 2D signal so that during EEG signals description, no
Temporal information is only included, spatial information can be also included.
When brain carries out different Mental imageries, the EEG signals on Different electrodes position can enter over time
Row change.This change is similar caused by same Mental imagery task, and the task of feature extraction is aiming at a certain kind
Mental imagery task, which is found, can describe the thing of this similitude.
Feature point extraction and characteristic vector calculating are concretely comprised the following steps in the step (1)：
(1-1) critical point detection：The motion of the brain machine interface system collection stored in first EEG signals memory module is thought
As EEG signals X ∈ IRn×m×TIt is expressed as X=[X1,X2,...,XT]；
For X1, X is detected using Harris Keypoint detectors1ρ key point.The task of feature extraction is exactly to see
Examine and how to change on (T-1) the individual time point of this ρ key point below.For this ρ key point, extraction size is (η
× η × T) module.
(1-2) characteristics extraction：Average is gone for the module all values in step (1-1), for each inside module
Individual pixel calculates its variance M along time orientation2, skewness M3With kurtosis M4, constitute time matrix M corresponding with moduler, r=
{ 2,3,4 }, wherein Mr=[mij], i, j=1,2 ..., η
Wherein, vijtIt is t-th timeslice of the position at { i, j }；
(1-3) characteristic vector is calculated：Each time matrix in step (1-2) is converted into a length for η2To
Amount, time matrix Mr, r three value composition characteristic vector m ∈ IR in r={ 2,3,4 }d, wherein d=3 η2For crucial point module
Characteristic vector length,
In the present embodiment, if one of module is v ∈ IRη×η×T, average is gone to the module all values first, then
For each pixel inside module its variance M is calculated along time orientation2, skewness M3With kurtosis M4。
Assuming that the time matrix related to module v is Mr, r={ 2,3,4 }, wherein Mr=[mij], i, j=1,2 ..., η
Wherein
Wherein vijtIt is t-th timeslice of the position at { i, j }.Each time matrix can be converted into a length
For η2Vector, therefore three of r values can be with composition of vector m ∈ IRd, wherein d=3 η2
The characteristic vector that the above method is obtained is typically all high dimension, very unfavorable to follow-up calculating.
In the present embodiment, in the case where not losing signal characteristic, dimensionality reduction uses accidental projection in the step (1)
(random projection) method carries out dimension-reduction treatment, concretely comprises the following steps：
Training data EEG signals have obtained p crucial point module, and the characteristic vector length of each crucial point module is
D,
The matrix that note describes Mental imagery characteristic vector m is D ∈ IRd×p, wherein, d is the characteristic vector of crucial point module
Length, p is the quantity of crucial point module；
In the subspace that the vector projection for being d by dimension is n to a dimension, wherein n ＜ ＜ d；
It is multiplied by a random matrix R to realize the dimension-reduction treatment of eigenvectors matrix by matrix D, by eigenvectors matrix
Dimension is reduced to n × p：
Y=RD
Wherein, Y ∈ IRn×pFor the eigenvectors matrix after dimensionality reduction, R is accidental projection matrix, R ∈ IRn×d, average is 0,
Variance is 1.
Dictionary learning in the step (1) obtains Y's according to the eigenvectors matrix Y after dimensionality reduction by compressing sensing
Sparse expression, is concretely comprised the following steps：
Using K-SVD algorithms, the dictionary Φ ∈ IR of following equation are metn×m(m ＞ n) and sparse expression formulaxi∈IRm, wherein, xiInclude k (k<<N) individual or less nonzero element：
Wherein, | | | |FIt is Frobenius norms, | | | |0It is l0Semi-norm, calculates the non-zero entry included in vector
Element.
K-SVD is a kind of classical dictionary training algorithm, and according to error minimum principle, SVD decomposition, choosing are carried out to error term
The dictionary atom and corresponding atomic that make the decomposition item of error minimum as renewal are selected, by continuous iteration so as to obtain
The solution of optimization.
Compressed sensing technology is applied in the present embodiment, after the eigenvectors matrix Y compressions after dimensionality reduction, obtained
To the dictionary of sparse expression, follow-up EEG signals set up the sparse expression of oneself using the dictionary, and utilize：If follow-up brain
Electric signal belongs to a certain Mental task existed during training dictionary, then the sparse expression coefficient in the Mental task part should
Comprising minimum nonzero value, and set up the classification that sparse expression histogram carries out Mental task.Different types of EEG signals are built
A common dictionary is found, once dictionary is set up, the assorting process of subsequent step (2) will save the plenty of time.
If in dictionary training process, it is believed that the state of user's rest is a kind of Mental task, then in subsequent classification process
In, the state that also user can be rested at any time is distinguished, therefore is solved current most of brain machine interface systems and can not be solved
The problem of, realize user and brain-computer interface is used for a long time, brain machine interface system can recognize the working condition and rest shape of user
State.
Training data classification histogram h in the step (1)i=[h1,h2,...,hK] obtained according to following formula calculating：
In the present embodiment, step (1) training part is terminated, and obtains two results：1) study dictionary Φ；2) classification Nogata
Scheme hi=[h1,h2,...,hK]。
(2) to the test of Mental imagery EEG signals：The Mental imagery EEG signals that brain machine interface system is gathered as
The form that test data enters test data in row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, step (2) turns
Change, feature point extraction, characteristic vector calculate with step (1) use method formula it is completely the same, just do not repeat here.
The sparse expression of test data is obtained using the training dictionary in training module, the classification Nogata of test data is obtained
Figure.
What the sparse expression of test data was calculated in the step (2) concretely comprises the following steps：
By test data in step (2) respectively through form conversion, feature point extraction, characteristic vector are calculated, dimensionality reduction is obtained
Dimensionality reduction after eigenvectors matrix Q, with reference to the dictionary Φ obtained in step (1),
According to：
Obtain the sparse expression X of test dataQ。
The classification histogram of test data in the step (2) basis respectively：
Obtain, the classification histogram of the test data in the step (2) is hQ；
Test data is carried out according to the histogrammic comparing result of the classification of the classification histogram of training data and test data
Classification.
What test data was classified in the step (2) concretely comprises the following steps：
According to：
Determine the classification of the test data.
Beneficial effects of the present invention：
1. classification species is improved：A kind of Mental imagery EEG feature extraction and the categorizing system of the present invention and
Method, in the training to Mental imagery EEG signals of step (1), different Mental tasks can set up a common word
Allusion quotation Φ, in the test to Mental imagery EEG signals of step (2), as long as separating existing Mental task in training department, is being surveyed
The category can be just detected during examination, therefore, the classification species of such a algorithm is improved；
2. Mental imagery can be detected with resting state：A kind of Mental imagery EEG feature extraction of the present invention
With categorizing system and method, resting state is regarded into a kind of Mental task, the state is taken into account when setting up dictionary Φ, then exists
In follow-up test, when there is resting state, it can also detect that user is currently in resting state；
3. classification speed is very fast：A kind of Mental imagery EEG feature extraction of the present invention and categorizing system and method,
Different Mental tasks sets up a common dictionary, and different Mental tasks can disposably obtain classification knot in testing
Really.
The preferred embodiment of the application is the foregoing is only, the application is not limited to, for the skill of this area
For art personnel, the application can have various modifications and variations.It is all within spirit herein and principle, made any repair
Change, equivalent substitution, improvement etc., should be included within the protection domain of the application.
Claims (10)
1. a kind of Mental imagery EEG feature extraction and categorizing system, the system include training module and test module；Its
It is characterized in：
The Mental imagery EEG signals that the training module is configured as gathering brain machine interface system are carried out as training data
Form conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, dictionary learning and classification histogram calculation, obtain training dictionary and
The histogrammic module of classification of training data；
The Mental imagery EEG signals that the test module is configured as gathering brain machine interface system are carried out as test data
Form conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, test data is obtained using the training dictionary in training module
Sparse expression, obtains the classification histogram of test data, and according to the classification histogram and the classification of test data of training data
Histogrammic comparing result carries out the module of test data classification.
2. a kind of Mental imagery EEG feature extraction and categorizing system as claimed in claim 1, it is characterized in that：The instruction
Practicing module includes the first EEG signals memory module, the fortune that the first EEG signals memory module gathers brain machine interface system
Dynamic imagination EEG signals are stored as training data；
The test module includes the second EEG signals memory module, and the second EEG signals memory module is by brain-computer interface system
The Mental imagery EEG signals of system collection are stored as test data.
3. a kind of Mental imagery EEG feature extraction and sorting technique, it is special that this method is based on a kind of Mental imagery EEG signals
Extraction and categorizing system are levied, it is characterized in that：The specific steps of this method include：
(1) to the training of Mental imagery EEG signals：The Mental imagery EEG signals that brain machine interface system is gathered are used as training
Data enter row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, dictionary learning and classification histogram calculation, are instructed
The classification histogram of allusion quotation of practising handwriting and training data；
(2) to the test of Mental imagery EEG signals：The Mental imagery EEG signals that brain machine interface system is gathered are used as test
Data enter row format conversion, feature point extraction, characteristic vector calculating, dimensionality reduction, are surveyed using the training dictionary in training module
The sparse expression of data is tried, the classification histogram of test data, and classification histogram and test number according to training data is obtained
According to the histogrammic comparing result of classification carry out test data classification.
4. a kind of Mental imagery EEG feature extraction and sorting technique as claimed in claim 3, it is characterized in that：The step
Suddenly enter row format conversion in (1) to training data to concretely comprise the following steps：
The Mental imagery EEG signals of collection are converted into three dimensional signal by 2D signal：By in the first EEG signals memory module
The Mental imagery EEG signals X ∈ IR of the brain machine interface system collection of storagen×m×TBuild two dimensional matrix.
5. a kind of Mental imagery EEG feature extraction and sorting technique as claimed in claim 3, it is characterized in that：The step
Suddenly feature point extraction and characteristic vector calculating are concretely comprised the following steps in (1)：
(1-1) critical point detection：The Mental imagery brain of the brain machine interface system collection stored in first EEG signals memory module
Electric signal X ∈ IRn×m×TIt is expressed as X=[X1,X2,...,XT]；
Key point is detected using Harris Keypoint detectors for each value in X, is carried for all key points detected
Take the module that size is (η × η × T)；
(1-2) characteristics extraction：Average is gone for the module all values in step (1-1), for each picture inside module
Element calculates its variance M along time orientation2, skewness M3With kurtosis M4, constitute time matrix M corresponding with moduler, r=2,3,
4 }, wherein Mr=[mij], i, j=1,2 ..., η
Wherein, vijtIt is t-th timeslice of the position at { i, j }；
(1-3) characteristic vector is calculated：Each time matrix in step (1-2) is converted into a length for η2Vector, when
Between matrix Mr, r three value composition characteristic vector m ∈ IR in r={ 2,3,4 }d, wherein d=3 η2For the feature of crucial point module
Vector length,
6. a kind of Mental imagery EEG feature extraction and sorting technique as claimed in claim 5, it is characterized in that：The step
Suddenly dimensionality reduction carries out dimension-reduction treatment using accidental projection (random projection) method in (1), concretely comprises the following steps：
The matrix that note describes Mental imagery characteristic vector m is D ∈ IRd×p, wherein, d is that the characteristic vector of crucial point module is long
Degree, p is the quantity of crucial point module；
In the subspace that the vector projection for being d by dimension is n to a dimension, wherein n ＜ ＜ d；
It is multiplied by a random matrix R to realize the dimension-reduction treatment of eigenvectors matrix by matrix D, by eigenvectors matrix dimension
It is reduced to n × p：
Y=RD
Wherein, Y ∈ IRn×pFor the eigenvectors matrix after dimensionality reduction, R is accidental projection matrix, R ∈ IRn×d, average is 0, and variance is
1。
7. a kind of Mental imagery EEG feature extraction and sorting technique as claimed in claim 6, it is characterized in that：The step
Suddenly the dictionary learning in (1) obtains Y sparse expression, specific step by compressing sensing according to the eigenvectors matrix Y after dimensionality reduction
Suddenly it is：
Using K-SVD algorithms, the dictionary Φ ∈ IR of following equation are metn×m(m ＞ n) and sparse expression formulaWherein, xiInclude k (k<<N) individual or less nonzero element：
Wherein, | | | |FIt is Frobenius norms, | | | |0It is l0Semi-norm, calculates the nonzero element included in vector.
8. a kind of Mental imagery EEG feature extraction and sorting technique as claimed in claim 3, it is characterized in that：The step
Suddenly in (2) test data sparse expression calculate concretely comprise the following steps：
By test data in step (2) respectively through form conversion, feature point extraction, the drop that characteristic vector is calculated, dimensionality reduction is obtained
Eigenvectors matrix Q after dimension, with reference to the dictionary Φ obtained in step (1),
According to：
Obtain the sparse expression X of test dataQ。
9. a kind of Mental imagery EEG feature extraction and sorting technique as claimed in claim 6, it is characterized in that：The step
Suddenly the classification histogram of the training data in (1) and the classification histogram of the test data in the step (2) basis respectively：
Obtain, the classification histogram of the training data in the step (1) is hi, the classification of the training data in the step (2)
Histogram is hQ。
10. a kind of Mental imagery EEG feature extraction and sorting technique as claimed in claim 9, it is characterized in that：It is described
What test data was classified in step (2) concretely comprises the following steps：
According to：
Determine the classification of the test data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710244075.8A CN106991409B (en) | 2017-04-14 | 2017-04-14 | Motor imagery electroencephalogram signal feature extraction and classification system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710244075.8A CN106991409B (en) | 2017-04-14 | 2017-04-14 | Motor imagery electroencephalogram signal feature extraction and classification system and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106991409A true CN106991409A (en) | 2017-07-28 |
| CN106991409B CN106991409B (en) | 2020-03-27 |
Family
ID=59416328
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710244075.8A Expired - Fee Related CN106991409B (en) | 2017-04-14 | 2017-04-14 | Motor imagery electroencephalogram signal feature extraction and classification system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106991409B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109558863A (en) * | 2019-01-07 | 2019-04-02 | 哈尔滨工业大学（深圳） | EEG signals character representation and extracting method, device and storage medium |
| CN109784211A (en) * | 2018-12-26 | 2019-05-21 | 西安交通大学 | A kind of Mental imagery Method of EEG signals classification based on deep learning |
| CN111695500A (en) * | 2020-06-11 | 2020-09-22 | 齐鲁工业大学 | Method and system for recognizing motor imagery task of stroke patient based on transfer learning |
| CN112075932A (en) * | 2020-10-15 | 2020-12-15 | 中国医学科学院生物医学工程研究所 | High-resolution time-frequency analysis method for evoked potential signals |
| CN114167983A (en) * | 2021-11-24 | 2022-03-11 | 哈尔滨工业大学（深圳） | Brain-computer interface system based on common amplitude and phase measurement |
| CN115079012A (en) * | 2022-07-22 | 2022-09-20 | 羿动新能源科技有限公司 | Characteristic data extraction system and method for big data prediction battery SOC |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101558997A (en) * | 2009-05-27 | 2009-10-21 | 江西蓝天学院 | Electroencephalograph signal identification method based on second-order blind identification |
| CN105468138A (en) * | 2015-07-15 | 2016-04-06 | 武汉理工大学 | Intelligent vehicle obstacle avoidance and navigation method based on brain-computer interface technology and lidar |
| CN105956624A (en) * | 2016-05-06 | 2016-09-21 | 东南大学 | Motor imagery electroencephalogram classification method based on space-time-frequency optimization feature sparse representation |
| CN106388814A (en) * | 2016-10-11 | 2017-02-15 | 天津大学 | Epilepsy electroencephalogram signal identification method based on optimal kernel time-frequency distribution visibility graph |
| CN106446762A (en) * | 2016-07-07 | 2017-02-22 | 哈尔滨工程大学 | Motion imagination EEG classification processing method based on sparse representation classification algorithm |
-
2017
- 2017-04-14 CN CN201710244075.8A patent/CN106991409B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101558997A (en) * | 2009-05-27 | 2009-10-21 | 江西蓝天学院 | Electroencephalograph signal identification method based on second-order blind identification |
| CN105468138A (en) * | 2015-07-15 | 2016-04-06 | 武汉理工大学 | Intelligent vehicle obstacle avoidance and navigation method based on brain-computer interface technology and lidar |
| CN105956624A (en) * | 2016-05-06 | 2016-09-21 | 东南大学 | Motor imagery electroencephalogram classification method based on space-time-frequency optimization feature sparse representation |
| CN106446762A (en) * | 2016-07-07 | 2017-02-22 | 哈尔滨工程大学 | Motion imagination EEG classification processing method based on sparse representation classification algorithm |
| CN106388814A (en) * | 2016-10-11 | 2017-02-15 | 天津大学 | Epilepsy electroencephalogram signal identification method based on optimal kernel time-frequency distribution visibility graph |
Non-Patent Citations (7)
| Title |
|---|
| 丁军等: ""一种联合阴影和目标区域图像的SAR目标识别方法"", 《电子与信息学报》 * |
| 刘桂红等: ""结合Gabor特征和自适应加权Fisher准则的人脸识别"", 《计算机工程与应用》 * |
| 徐宝国等: ""在线脑机接口中脑电信号的特征提取与分类方法"", 《电子学报》 * |
| 江友谊等: ""基于ICM的局部不变特征提取方法"", 《红外技术》 * |
| 王小玉等: ""基于内容的双字典学习及稀疏表示的图像重构"", 《仪器仪表学报》 * |
| 范影乐等: ""基于排列组合熵的脑电意识任务识别方法的研究"", 《传感技术学报》 * |
| 蒋静芳等: ""基于脑电信号的情绪评估演技综述"", 《信息工程大学学报》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109784211A (en) * | 2018-12-26 | 2019-05-21 | 西安交通大学 | A kind of Mental imagery Method of EEG signals classification based on deep learning |
| CN109558863A (en) * | 2019-01-07 | 2019-04-02 | 哈尔滨工业大学（深圳） | EEG signals character representation and extracting method, device and storage medium |
| CN111695500A (en) * | 2020-06-11 | 2020-09-22 | 齐鲁工业大学 | Method and system for recognizing motor imagery task of stroke patient based on transfer learning |
| CN112075932A (en) * | 2020-10-15 | 2020-12-15 | 中国医学科学院生物医学工程研究所 | High-resolution time-frequency analysis method for evoked potential signals |
| CN112075932B (en) * | 2020-10-15 | 2023-12-05 | 中国医学科学院生物医学工程研究所 | High-resolution time-frequency analysis method for evoked potential signals |
| CN114167983A (en) * | 2021-11-24 | 2022-03-11 | 哈尔滨工业大学（深圳） | Brain-computer interface system based on common amplitude and phase measurement |
| CN115079012A (en) * | 2022-07-22 | 2022-09-20 | 羿动新能源科技有限公司 | Characteristic data extraction system and method for big data prediction battery SOC |
| CN115079012B (en) * | 2022-07-22 | 2022-11-29 | 羿动新能源科技有限公司 | Characteristic data extraction system and method for big data prediction battery SOC |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106991409B (en) | 2020-03-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106991409A (en) | A kind of Mental imagery EEG feature extraction and categorizing system and method | |
| Jiao et al. | Deep convolutional neural networks for mental load classification based on EEG data | |
| Esfahani et al. | Classification of primitive shapes using brain–computer interfaces | |
| Kumar et al. | Envisioned speech recognition using EEG sensors | |
| Yang et al. | A recurrence quantification analysis-based channel-frequency convolutional neural network for emotion recognition from EEG | |
| Gao et al. | A recurrence network-based convolutional neural network for fatigue driving detection from EEG | |
| Kundu et al. | P300 detection with brain–computer interface application using PCA and ensemble of weighted SVMs | |
| Chuang et al. | Independent component ensemble of EEG for brain–computer interface | |
| Brodu et al. | Exploring two novel features for EEG-based brain–computer interfaces: Multifractal cumulants and predictive complexity | |
| Wang et al. | Local temporal common spatial patterns for robust single-trial EEG classification | |
| CN106529476B (en) | A kind of EEG feature extraction and classification method stacking network based on deep layer | |
| Vézard et al. | EEG classification for the detection of mental states | |
| Mohammadpour et al. | Classification of EEG-based emotion for BCI applications | |
| Ju et al. | Tensor-cspnet: A novel geometric deep learning framework for motor imagery classification | |
| CN111265212A (en) | Motor imagery electroencephalogram signal classification method and closed-loop training test interaction system | |
| Wang et al. | A new method for EEG-based concealed information test | |
| CN106943140A (en) | A kind of Mental imagery EEG feature extraction method based on RandomSelect RCSP | |
| Chen et al. | Fusing frequency-domain features and brain connectivity features for cross-subject emotion recognition | |
| Jiajie et al. | Clinical decision support system for alcoholism detection using the analysis of EEG signals | |
| Rejer | EEG feature selection for BCI based on motor imaginary task | |
| Li et al. | A parallel multi-scale time-frequency block convolutional neural network based on channel attention module for motor imagery classification | |
| Yoon et al. | Multiple kernel learning based on three discriminant features for a P300 speller BCI | |
| Wiechert et al. | Identifying users and activities with cognitive signal processing from a wearable headband | |
| Jinliang et al. | EEG emotion recognition based on granger causality and capsnet neural network | |
| Lamba et al. | Multimodal human eye blink recognition method using feature level fusion for exigency detection |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20200616 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: Feng Ming Licheng District Lingang Development Zone, 250101 Ji'nan Road, Shandong province No. 1000 Patentee before: SHANDONG JIANZHU University |
|
| TR01 | Transfer of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200327 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |