CN105677043B - The two stages adaptive training method of Mental imagery brain-computer interface - Google Patents
The two stages adaptive training method of Mental imagery brain-computer interface Download PDFInfo
- Publication number
- CN105677043B CN105677043B CN201610107996.5A CN201610107996A CN105677043B CN 105677043 B CN105677043 B CN 105677043B CN 201610107996 A CN201610107996 A CN 201610107996A CN 105677043 B CN105677043 B CN 105677043B
- Authority
- CN
- China
- Prior art keywords
- trial
- matrix
- classifier
- mental imagery
- feature
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Neurology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Computer Interaction (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Dermatology (AREA)
- Biomedical Technology (AREA)
- Neurosurgery (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
本发明涉及一种运动想象脑机接口的两阶段自适应训练方法，包括先经历的单信任阶段和再进入的互信任阶段。单信任指，系统信任数据，用数据更新分类器。在该阶段，系统先根据用户开展的多个trial进行初步的训练，得到一个初步可行的分类器，再采用增量学习的方法不断更新分类器。互信任指，系统同时信任数据与分类器，让两者相互适应。在该阶段，系统先采用单信任阶段得到的分类器来识别，当用户开展一定量的trial后，根据反馈结果，配合以SVM寻求支持向量的方法优选数据，用增量学习的方法不断更新分类器，对之后的trial用新的分类器进行识别和反馈，反复此过程直至训练结束。本发明提供的方法，能够增强分类器与用户的相互适应，并且耗时短，准确率高。
The invention relates to a two-stage self-adaptive training method of a motor imagery brain-computer interface, which includes a single trust stage experienced first and a mutual trust stage entered again. Single trust means that the system trusts the data and updates the classifier with the data. At this stage, the system first conducts preliminary training based on multiple trials conducted by users to obtain a preliminary feasible classifier, and then uses incremental learning to continuously update the classifier. Mutual trust means that the system trusts the data and the classifier at the same time, so that the two can adapt to each other. At this stage, the system first uses the classifier obtained in the single-trust stage to identify. After the user conducts a certain amount of trials, according to the feedback results, the data is optimized by using SVM to find support vectors, and the classification is continuously updated by incremental learning. The classifier is used to identify and feed back the subsequent trial with a new classifier, and this process is repeated until the end of the training. The method provided by the invention can enhance the mutual adaptation between the classifier and the user, and has short time consumption and high accuracy.
Description
Technical field
It is mutually adapted problem the present invention relates to the man-machine training process of brain-computer interface, is connect for Mental imagery type brain machine
A kind of adaptive training method of mouth.
Background technique
There is be mutually adapted between the user and brain machine interface system of brain-computer interface.In man-machine training process
In, the characteristics of brain machine interface system constantly collects data and adapts to user by the method for machine learning, and user passes through observation
The feedback of brain machine interface system adjusts the activity in brain actively also to adapt to the working method of brain-computer interface.Brain-computer interface user
Being mutually adapted between brain machine interface system is a dynamic process.User feedback is not considered when collecting training data, only
Carry out machine learning on a static training set, it is not consistent with this characteristic of brain-computer interface, it is unfavorable for improving brain machine
The performance of interface.
Summary of the invention
In view of this, the purpose of the present invention is the brain-computer interface for Mental imagery type provides, one kind is man-machine to be mutually adapted
Training method.The man-machine training process of Mental imagery brain-computer interface is divided into two stages by the present invention, and previous stage is single letter
Appoint the stage, the latter half is mutual trusted phase.In single trusted phase, brain machine interface system trust data, using incremental learning
Method constantly updates classifier with new data.In mutual trusted phase, brain machine interface system trust data and classifier simultaneously allow two
Person is mutually adapted.
The present invention is realized using under type: a kind of two stages adaptive training method of Mental imagery brain-computer interface, including
Following steps:
Step S1: this step is in single trusted phase, and Mental imagery brain machine interface system trust data is updated with data and divided
Class device；When initial, user carries out the Mental imagery of multiple trial, and system online acquisition sample obtains one using LDA/QR algorithm
A preliminary transfer matrix G, and classification center vector set C is calculated, form preliminary classification device；User carries on Mental imagery,
System obtains new samples with the Mental imagery of classifier online recognition user and to user feedback；Whenever completion one
Trial, system updates transfer matrix G using ILDA/QR algorithm, and calculates classification center vector set C, forms new classifier simultaneously
It is used for the identification and feedback of next trial, is terminated until the stage；
Step S2: this step is in mutual trusted phase, Mental imagery brain machine interface system while trust data and classifier,
The two is allowed to be mutually adapted；When initial, system uses the Mental imagery of the finally obtained classifier online recognition user of step S1 simultaneously
To user feedback；After user carries out a certain number of trial, system is used using optimum seeking method screening new samples are mutually trusted
LDA/QR algorithm or ILDA/QR algorithm update transfer matrix G, and calculate classification center vector set C, form new classifier and incite somebody to action
For it is next identification and feedback, until man-machine training terminates.
Further, the step S1 specifically includes the following steps:
Step S11: user carries out the Mental imagery of multiple trial, system online acquisition signal, every a time interval
A segment signal is intercepted, m dimensional feature vector is converted into through feature extraction, is denoted as x；
Step S12: all feature vectors obtained in the step S11 construct data matrix A, and according to classification number
K structural matrix E；
Step S13: LDA/QR algorithm is executed, obtains optimal transfer matrix G, and class center vector collection C, shape is calculated
At preliminary classification device；
Step S14: user carries out the Mental imagery of a trial；System online acquisition EEG signals, every a time
Interval one segment signal of interception is converted to feature vector x through feature extraction, identifies x with classifier, controlled according to recognition result movement
Object is to user feedback；
Step S15: executing ILDA/QR algorithm, updates transfer matrix G, then updates data matrix A, and calculate class center
Vector set C forms new classifier；
Step S16: this stage terminates；Otherwise, new classifier is used for the identification and feedback of next trial, returns to step
Rapid S14.
Further, the step S2 specifically includes the following steps:
Step S21: setting mark flag is 0, and construction container is for storing k type games imagination sample；
Step S22: user carries out the Mental imagery of a trial；System online acquisition EEG signals, every a time
Interval one segment signal of interception is converted to feature vector x through feature extraction, identifies x with classifier, controlled according to recognition result movement
Object is to user feedback；
Step S23: at the end of current trial, if controll plant hits the mark, all samples of the trial is stored in and are held
Device, and record the trial and hit the mark the spent time；
Step S24: if trial number is not up to adaptive updates condition, return step S22；Otherwise, system uses mutual trust
Optimum seeking method is appointed to screen new samples；
Step S25: if flag is equal to 0, constructing data matrix A, according to classification number k structural matrix E, executes LDA/QR and calculates
Method obtains new transfer matrix G, and setting flag is 1；Otherwise, it executes ILDA/QR algorithm and updates transfer matrix G, and more new data
Matrix A；
Step S26: being calculated new class center vector collection C according to transfer matrix G and data matrix A, forms new point
Class device, and empty container；
Step S27: man-machine training terminates；Otherwise, return step S22.
Further, mutual trust optimum seeking method described in step S24 comprises the steps of:
Step S41: according to hitting the mark, the spent time screens trial；With the corresponding consuming of trial all in container
Time one preliminary set of composition, finds out the minimum consuming time, other consuming times exist using the minimum consuming time as symmetry axis
Its symmetric position generates a virtual time, and all virtual times are added to preliminary set and form reference set, calculates reference set
The standard deviation of conjunction is chosen and expends the time less than the minimum trial for expending the sum of time and standard deviation；
Step S42: with the method Screening Samples of supporting vector；All sample standard deviations corresponding to the trial that step S41 chooses
Candidate samples collection is added, the training on candidate samples collection, chooses those samples for being confirmed as supporting vector with SVM；
Wherein, the trial indicates an experimental considerations unit in the man-machine training process of Mental imagery brain-computer interface；trial
When beginning, brain machine interface system gives a target at random；User's imagination of launching a campaign is tried hard to the mobile controll plant of target, brain
Machine interface system determines the direction of practical movement controll plant by identifying the type of Mental imagery；When controll plant hits mesh
Mark or time-out, trial terminate.
Further, the method for construction described in step S12 and S25 or update data matrix A are as follows:
Data matrix A=[x1,x2,…,xn]=[A1,…,Ak]∈Rm×n, xi∈Rm×1I=1 ..., n indicates a m dimension
Sample point, n sample point is shared in A；M is determined by feature extracting method, is fixed and invariable in system operation；n
It is that dynamic increases in system operation；Each sample block matrixWhat is indicated is the i-th class
The set of all sample points shares k classification, niI=1 ..., k indicate the number of samples of the i-th class,
Further, according to classification number k structural matrix E's described in step S12 and S25 method particularly includes:
MatrixWherein
Further, the input of the LDA/QR algorithm is data matrix A ∈ Rm×nWith E ∈ Rn×k, export as transfer matrix
G∈Rm×k；
Specifically includes the following steps:
Step S31: the economic QR for calculating A is decomposed, A=QR, Q ∈ Rm×n, R ∈ Rn×n, wherein Q is to arrange orthogonal, R right and wrong
Unusual；
Step S32: one lower triangular linear system R of solutionTH=E, solution obtain H；
Step S33: G, G=QH is calculated.
Further, the input of the ILDA/QR algorithm is column orthogonal matrix Q, optimal transfer matrix G and new sample
Eigen vector x and its classification l export column orthogonal matrix Q and transfer matrix G for update；
Specifically includes the following steps:
Step S41: r=Q is calculatedTX,Q=[Q (x-Qr)/α]；
Step S42: r=-G is calculatedTX, r (l)=r (l)+1, r=r/ α, G=G+Q (:, n+1) rT。
Further, the calculation method of the classification center vector set C are as follows: the data matrix A of existing known class and
Transfer matrix G calculates P=ATG,Wherein pj∈R1×kJ=1 ..., n indicates every in data matrix A
The corresponding projection result of one sample point, Pi∈R1×kI=1 ..., that k is indicated is all p for belonging to the i-th classjSet；It calculates
The center C of each classi∈R1×k,Obtaining class center vector collection is
Further, the specific classification method of the classifier are as follows: sample to be determined is characterized vector x, calculates di=| |
xTG-Ci||2I=1 ..., k select the smallest di, feature vector x is determined as corresponding classification.
The user of brain-computer interface both needs to allow before using brain machine interface system phase by a personal-machine training process
Coadaptation.The state of user is not static constant in this process, feedback procedure of the user in observation brain machine interface system
In adjust the state of itself constantly to adapt to the working method of brain-computer interface.The matter of the collected data set of brain machine interface system
Amount is the variation with User Status and changes.System is allowed to be capable of the data of preferred high quality automatically in man-machine training process
It is very important.Therefore, compared with prior art, the invention has the following advantages that
1. the present invention can allow brain machine interface system automatic preferred sample in man-machine training process, so that brain-computer interface is used
The kilter at family can highlight in new samples set.
2. present invention employs the methods of incremental learning.When there are new samples, it is not necessarily to re -training, brain machine interface system
The information that new samples are contained can quickly be absorbed.
3. present invention employs be concisely and efficiently algorithm.Brain machine interface system can in man-machine training process it is online self
It updates, so that brain machine interface system and being mutually adapted for user can be dynamically completed in man-machine training process.
To sum up, method provided by the invention can enhance brain machine interface system and the mutual of user in man-machine training process and fit
It should be able to power.
Detailed description of the invention
Fig. 1 is main-process stream schematic diagram of the invention.
Fig. 2 is the flow diagram of step S1 of the invention.
Fig. 3 is the flow diagram of step S2 of the invention.
Fig. 4 is the schematic diagram of a trial of Mental imagery of the present invention.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and embodiments.
This implementation provides a kind of two stages adaptive training method of Mental imagery brain-computer interface, as shown in Figure 1, include with
Lower step:
Step S1: this step is in single trusted phase, and Mental imagery brain machine interface system trust data is updated with data and divided
Class device；When initial, user carries out the Mental imagery of multiple trial, and system online acquisition sample obtains one using LDA/QR algorithm
A preliminary transfer matrix G, and classification center vector set C is calculated, form preliminary classification device；User carries on Mental imagery,
System obtains new samples with the Mental imagery of classifier online recognition user and to user feedback；Whenever completion one
Trial, system updates transfer matrix G using ILDA/QR algorithm, and calculates classification center vector set C, forms new classifier simultaneously
It is used for the identification and feedback of next trial, is terminated until the stage；
Step S2: this step is in mutual trusted phase, Mental imagery brain machine interface system while trust data and classifier,
The two is allowed to be mutually adapted；When initial, system uses the Mental imagery of the finally obtained classifier online recognition user of step S1 simultaneously
To user feedback；After user carries out a certain number of trial, system is used using optimum seeking method screening new samples are mutually trusted
LDA/QR algorithm or ILDA/QR algorithm update transfer matrix G, and calculate classification center vector set C, form new classifier and incite somebody to action
For it is next identification and feedback, until man-machine training terminates.
In the present embodiment, as shown in Fig. 2, the step S1 specifically includes the following steps:
Step S11: user carries out the Mental imagery of multiple trial, system online acquisition signal, every a time interval
A segment signal is intercepted, m dimensional feature vector is converted into through feature extraction, is denoted as x；
Step S12: all feature vectors obtained in the step S11 construct data matrix A, and according to classification number
K structural matrix E；
Step S13: LDA/QR algorithm is executed, obtains optimal transfer matrix G, and class center vector collection C, shape is calculated
At preliminary classification device；
Step S14: user carries out the Mental imagery of a trial；System online acquisition EEG signals, every a time
Interval one segment signal of interception is converted to feature vector x through feature extraction, identifies x with classifier, controlled according to recognition result movement
Object is to user feedback；
Step S15: executing ILDA/QR algorithm, updates transfer matrix G, then updates data matrix A, and calculate class center
Vector set C forms new classifier；
Step S16: this stage terminates；Otherwise, new classifier is used for the identification and feedback of next trial, returns to step
Rapid S14.
In the present embodiment, as shown in figure 3, the step S2 specifically includes the following steps:
Step S21: setting mark flag is 0, and construction container is for storing k type games imagination sample；
Step S22: user carries out the Mental imagery of a trial；System online acquisition EEG signals, every a time
Interval one segment signal of interception is converted to feature vector x through feature extraction, identifies x with classifier, controlled according to recognition result movement
Object is to user feedback；
Step S23: at the end of current trial, if controll plant hits the mark, all samples of the trial is stored in and are held
Device, and record the trial and hit the mark the spent time；
Step S24: if trial number is not up to adaptive updates condition, return step S22；Otherwise, system uses mutual trust
Optimum seeking method is appointed to screen new samples；
Step S25: if flag is equal to 0, constructing data matrix A, according to classification number k structural matrix E, executes LDA/QR and calculates
Method obtains new transfer matrix G, and setting flag is 1；Otherwise, it executes ILDA/QR algorithm and updates transfer matrix G, and more new data
Matrix A；
Step S26: being calculated new class center vector collection C according to transfer matrix G and data matrix A, forms new point
Class device, and empty container；
Step S27: man-machine training terminates；Otherwise, return step S22.
In the present embodiment, mutual trust optimum seeking method described in step S24 comprises the steps of:
Step S41: according to hitting the mark, the spent time screens trial；With the corresponding consuming of trial all in container
Time one preliminary set of composition, finds out the minimum consuming time, other consuming times exist using the minimum consuming time as symmetry axis
Its symmetric position generates a virtual time, and all virtual times are added to preliminary set and form reference set, calculates reference set
The standard deviation of conjunction is chosen and expends the time less than the minimum trial for expending the sum of time and standard deviation；
Step S42: with the method Screening Samples of supporting vector；All sample standard deviations corresponding to the trial that step S41 chooses
Candidate samples collection is added, the training on candidate samples collection, chooses those samples for being confirmed as supporting vector with SVM；
Wherein, the trial indicates an experimental considerations unit in the man-machine training process of Mental imagery brain-computer interface；trial
When beginning, brain machine interface system gives a target at random；User's imagination of launching a campaign is tried hard to the mobile controll plant of target, brain
Machine interface system determines the direction of practical movement controll plant by identifying the type of Mental imagery；When controll plant hits mesh
Mark or time-out, trial terminate.
In the present embodiment, the method for construction described in step S12 and S25 or update data matrix A are as follows:
Data matrix A=[x1,x2,…,xn]=[A1,…,Ak]∈Rm×n, xi∈Rm×1I=1 ..., n indicates a m dimension
Sample point, n sample point is shared in A；M is determined by feature extracting method, is fixed and invariable in system operation；n
It is that dynamic increases in system operation；Each sample block matrixWhat is indicated is the i-th class
The set of all sample points shares k classification, niI=1 ..., k indicate the number of samples of the i-th class,
In the present embodiment, according to classification number k structural matrix E's described in step S12 and S25 method particularly includes:
MatrixWherein
In the present embodiment, the input of the LDA/QR algorithm is data matrix A ∈ Rm×nWith E ∈ Rn×k, export as transfer
Matrix G ∈ Rm×k；
Specifically includes the following steps:
Step S31: the economic QR for calculating A is decomposed, A=QR, Q ∈ Rm×n, R ∈ Rn×n, wherein Q is to arrange orthogonal, R right and wrong
Unusual；
Step S32: one lower triangular linear system R of solutionTH=E, solution obtain H；
Step S33: G, G=QH is calculated.
In the present embodiment, the input of the ILDA/QR algorithm is column orthogonal matrix Q, optimal transfer matrix G and newly arrives
Sampling feature vectors x and its classification l, export the column orthogonal matrix Q and transfer matrix G for update；
Specifically includes the following steps:
Step S41: r=Q is calculatedTX,Q=[Q (x-Qr)/α]；
Step S42: r=-G is calculatedTX, r (l)=r (l)+1, r=r/ α, G=G+Q (:, n+1) rT。
In the present embodiment, the calculation method of the classification center vector set C are as follows: the data matrix of existing known class
A and transfer matrix G calculates P=ATG,Wherein pj∈R1×kJ=1 ..., n indicates data matrix A
In the corresponding projection result of each sample point, Pi∈R1×kI=1 ..., that k is indicated is all p for belonging to the i-th classjSet；
Calculate the center C of each classi∈R1×k,Obtaining class center vector collection is
In the present embodiment, the specific classification method of the classifier are as follows: sample to be determined is characterized vector x, calculates di
=| | xTG-Ci||2I=1 ..., k select the smallest di, feature vector x is determined as corresponding classification.
In the present embodiment, the Mental imagery experiment of user is to be made of multiple run, and each run is by multiple
Trial composition, each trial is made of the Windowslength of multiple regular lengths for having overlapping.Such as Fig. 4 institute
Show, trial indicates an experimental considerations unit in the man-machine training process of Mental imagery brain-computer interface, and the unit is from any given
Target start, hit the mark to controll plant or time-out.Specifically: occur Target Board generation first in 0s, on screen
Table wishes the target that user hits, and user can start to imagine corresponding movement at this time.The controll plant indicated after 2s with bead
The centre for appearing in screen, between 2s-10s later, bead differentiates that the imagination result of user carries out accordingly according to classifier
Movement.If bead hits the mark plate earlier than 10s, the trial is terminated in advance, into the rest period.When 10s without
Whether hit the mark plate by bead, bead stop motions, trial terminates and enters the rest period.In rest period bead and
Target Board disappears, duration 2s, starts a new trial again later.In the present embodiment, as shown in figure 4,
Using the data segment of 2s as the length of a fixed length Windows length, take a Windows length's long every 0.1s
Data are input in system using each collected Windows length as online signal data section x.
The foregoing is merely presently preferred embodiments of the present invention, all equivalent changes done according to scope of the present invention patent with
Modification, is all covered by the present invention.
Claims (8)
1. a kind of two stages adaptive training method of Mental imagery brain-computer interface, which comprises the following steps:
Step S1: this step is in single trusted phase, and Mental imagery brain machine interface system trust data is updated with data and classified
Device；When initial, user carries out the Mental imagery of multiple trial, and system online acquisition sample obtains one using LDA/QR algorithm
Preliminary transfer matrix G, and class center vector collection C is calculated, form preliminary classification device；User carries on Mental imagery, system
With the Mental imagery of preliminary classification device online recognition user and to user feedback, while obtaining new samples；Whenever completion one
Trial, system updates transfer matrix G using ILDA/QR algorithm, and calculates class center vector collection C, forms new classifier and incites somebody to action
The identification and feedback for next trial, terminate until the stage；
Step S2: this step is in mutual trusted phase, Mental imagery brain machine interface system trust data and classifier simultaneously, allows two
Person is mutually adapted；When initial, system using the finally obtained classifier online recognition user of step S1 Mental imagery and to
Family feedback；After user carries out a certain number of trial, system is using optimum seeking method screening new samples are mutually trusted, using LDA/
QR algorithm or ILDA/QR algorithm update transfer matrix G, and calculate class center vector collection C, form new classifier and are used for it
Next identification and feedback, until man-machine training terminates；
The step S2 specifically includes the following steps:
Step S21: setting mark flag is 0, and construction container is for storing k type games imagination sample；
Step S22: user carries out the Mental imagery of a trial；System online acquisition EEG signals, every a time interval
It intercepts a segment signal and is converted to feature vector x through feature extraction, identify x with classifier, according to the mobile controll plant of recognition result
To user feedback；
Step S23: at the end of current trial, if controll plant hits the mark, being stored in container for all samples of the trial, and
The trial is recorded to hit the mark the spent time；
Step S24: if trial number is not up to adaptive updates condition, return step S22；Otherwise, system is excellent using mutually trusting
Method is selected to screen new samples；
Step S25: if flag is equal to 0, constructing data matrix A, according to classification number k structural matrix E, executes LDA/QR algorithm and obtains
To new transfer matrix G, it is 1 that flag, which is arranged,；Otherwise, it executes ILDA/QR algorithm and updates transfer matrix G, and update data matrix
A；
Step S26: being calculated new class center vector collection C according to transfer matrix G and data matrix A, form new classifier,
And empty container；
Step S27: man-machine training terminates；Otherwise, return step S22；
Mutual trust optimum seeking method described in step S24 comprises the steps of:
Step S241: according to hitting the mark, the spent time screens trial；When consuming corresponding with trial all in container
Between constitute a preliminary set, find out it is minimum expend the time, other expend times, and to expend the time using minimum be symmetry axis at it
Symmetric position generates a virtual time, and all virtual times are added to preliminary set and form reference set, calculates reference set
Standard deviation, choose and expend the time and be less than the minimum trial for expending the sum of time and standard deviation；
Step S242: with the method Screening Samples of supporting vector；All sample standard deviations corresponding to the trial that step S241 chooses add
Enter candidate samples collection, the training on candidate samples collection, chooses those samples for being confirmed as supporting vector with SVM；
Wherein, the trial indicates an experimental considerations unit in the man-machine training process of Mental imagery brain-computer interface；Trial starts
When, brain machine interface system gives a target at random；User's imagination of launching a campaign is tried hard to the mobile controll plant of target, and brain machine connects
Port system determines the direction of practical movement controll plant by identifying the type of Mental imagery；Hit the mark when controll plant or
Time-out, trial terminate.
2. a kind of two stages adaptive training method of Mental imagery brain-computer interface according to claim 1, feature exist
In, the step S1 specifically includes the following steps:
Step S11: user carries out the Mental imagery of multiple trial, and system online acquisition signal is intercepted every a time interval
One segment signal is converted into m dimensional feature vector through feature extraction, is denoted as x；
Step S12: all feature vectors obtained in the step S11 construct data matrix A, and according to classification number k structure
Make matrix E；
Step S13: executing LDA/QR algorithm, obtain optimal transfer matrix G, and class center vector collection C is calculated, and is formed just
Beginning classifier；
Step S14: user carries out the Mental imagery of a trial；System online acquisition EEG signals, every a time interval
It intercepts a segment signal and is converted to feature vector x through feature extraction, identify x with classifier, according to the mobile controll plant of recognition result
To user feedback；
Step S15: executing ILDA/QR algorithm, updates transfer matrix G, then updates data matrix A, and calculate class center vector
Collect C, forms new classifier；
Step S16: this stage terminates；Otherwise, new classifier is used for the identification and feedback of next trial, return step
S14。
3. a kind of two stages adaptive training method of Mental imagery brain-computer interface according to claim 1 or 2, feature
It is, the method for the construction or update data matrix A are as follows:
Data matrix A=[x1,x2,…,xn]=[A1,…,Ak]∈Rm×n, xiIndicate the sample point of m dimension, xi∈Rm×1(i=
1 ..., n), n sample point is shared in A；M is determined by feature extracting method, is fixed and invariable in system operation；N exists
It is that dynamic increases in system operation；Each sample block matrix AiWhat is indicated is the set of all sample points of the i-th class,
K classification is shared,niIndicate the number of samples of the i-th class, i=1 ..., k,
4. a kind of two stages adaptive training method of Mental imagery brain-computer interface according to claim 1 or 2, feature
It is, it is described according to classification number k structural matrix E's method particularly includes:
MatrixWhereinN indicates data matrix A
In sample point number, niIndicate the number of samples of the i-th class.
5. a kind of two stages adaptive training method of Mental imagery brain-computer interface according to claim 1 or 2, feature
It is, the input of the LDA/QR algorithm is data matrix A ∈ Rm×nWith E ∈ Rn×k, export as transfer matrix G ∈ Rm×k, m expression
The dimension of feature vector x, n indicate the sample point number in data matrix A；
Specifically includes the following steps:
(1) QR for calculating A is decomposed, A=QR, Q ∈ Rm×n, R ∈ Rn×n, wherein Q is that column are orthogonal, and R is nonsingular；
(2) a lower triangular linear system R is solvedTH=E, solution obtain H；
(3) G, G=QH is calculated.
6. a kind of two stages adaptive training method of Mental imagery brain-computer interface according to claim 1 or 2, feature
It is, the input of the ILDA/QR algorithm is the feature vector x of column orthogonal matrix Q, optimal transfer matrix G and sample of newly arriving
And its classification l, export the column orthogonal matrix Q and transfer matrix G for update；
Specifically includes the following steps:
(1) r=Q is calculatedTX,Q=[Q (x-Qr)/α]；
(2) r=-G is calculatedTX, r (l)=r (l)+1, r=r/ α, G=G+Q (:, n+1) rT。
7. a kind of two stages adaptive training method of Mental imagery brain-computer interface according to claim 1 or 2, feature
It is,
The calculation method of the class center vector collection C are as follows: the data matrix A and transfer matrix G of existing known class calculate P
=ATG,Wherein pjIndicate the corresponding projection result of each sample point in data matrix A, pj∈
R1×k(j=1 ..., n)；PiThat indicate is all p for belonging to the i-th classjSet,It calculates each
The center C of classi∈R1×k,Obtaining class center vector collection isN indicates the sample in data matrix A
Point number, niIndicate the number of samples of the i-th class.
8. a kind of two stages adaptive training method of Mental imagery brain-computer interface according to claim 7, feature exist
In the specific classification method of the classifier are as follows: sample to be determined is characterized vector x, calculates di=| | xTG-Ci||2, i=
1 ..., k select the smallest di, feature vector x is determined as corresponding classification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610107996.5A CN105677043B (en) | 2016-02-26 | 2016-02-26 | The two stages adaptive training method of Mental imagery brain-computer interface |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610107996.5A CN105677043B (en) | 2016-02-26 | 2016-02-26 | The two stages adaptive training method of Mental imagery brain-computer interface |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105677043A CN105677043A (en) | 2016-06-15 |
| CN105677043B true CN105677043B (en) | 2018-12-25 |
Family
ID=56305206
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610107996.5A Expired - Fee Related CN105677043B (en) | 2016-02-26 | 2016-02-26 | The two stages adaptive training method of Mental imagery brain-computer interface |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105677043B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106371590B (en) * | 2016-08-29 | 2019-06-18 | 华南理工大学 | High-performance online brain-computer interface system for motor imagery based on OpenVIBE |
| CN110123313B (en) * | 2019-04-17 | 2022-02-08 | 中国科学院深圳先进技术研究院 | Self-training brain-computer interface system and related training method |
| CN113180695B (en) * | 2021-04-20 | 2024-04-05 | 西安交通大学 | Brain-computer interface signal classification method, system, equipment and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2007138598A2 (en) * | 2006-06-01 | 2007-12-06 | Tylerton International Inc. | Brain stimulation and rehabilitation |
| WO2008097200A1 (en) * | 2007-02-09 | 2008-08-14 | Agency For Science, Technology And Research | A system and method for classifying brain signals in a bci system |
| CN102629156A (en) * | 2012-03-06 | 2012-08-08 | 上海大学 | Method for achieving motor imagery brain computer interface based on Matlab and digital signal processor (DSP) |
| CN103429145A (en) * | 2010-03-31 | 2013-12-04 | 新加坡科技研究局 | A method and system for motor rehabilitation |
| CN103488297A (en) * | 2013-09-30 | 2014-01-01 | 华南理工大学 | Online semi-supervising character input system and method based on brain-computer interface |
| CN104182042A (en) * | 2014-08-14 | 2014-12-03 | 华中科技大学 | BCI (brain-computer interface) method for multi-modal signals |
-
2016
- 2016-02-26 CN CN201610107996.5A patent/CN105677043B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2007138598A2 (en) * | 2006-06-01 | 2007-12-06 | Tylerton International Inc. | Brain stimulation and rehabilitation |
| WO2008097200A1 (en) * | 2007-02-09 | 2008-08-14 | Agency For Science, Technology And Research | A system and method for classifying brain signals in a bci system |
| CN103429145A (en) * | 2010-03-31 | 2013-12-04 | 新加坡科技研究局 | A method and system for motor rehabilitation |
| CN102629156A (en) * | 2012-03-06 | 2012-08-08 | 上海大学 | Method for achieving motor imagery brain computer interface based on Matlab and digital signal processor (DSP) |
| CN103488297A (en) * | 2013-09-30 | 2014-01-01 | 华南理工大学 | Online semi-supervising character input system and method based on brain-computer interface |
| CN104182042A (en) * | 2014-08-14 | 2014-12-03 | 华中科技大学 | BCI (brain-computer interface) method for multi-modal signals |
Non-Patent Citations (5)
| Title |
|---|
| P300脑机接口的在线半监督学习算法与系统研究;张锦涛;《中国优秀硕士学位论文全文数据库 医药卫生科技辑》;20151215;全文 * |
| 基于LDA主题模型的标签传递算法;刘培奇等;《计算机应用》;20120201;全文 * |
| 基于QR分解和支持向量的伪逆LDA;吴秀清;《聊城大学学报(自然科学版)》;20111225;全文 * |
| 基于想象左右手运动的在线自适应BCI系统研究;佟晓丽;《中国优秀硕士学位论文全文数据库 信息科技辑》;20141115;全文 * |
| 脑机接口的MapReduce计算模型;黄志华;《福州大学学报（自然科学版）》;20131231;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105677043A (en) | 2016-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Long et al. | Stand-alone inter-frame attention in video models | |
| KR102663375B1 (en) | Apparatus and method for automatically focusing the audio and the video | |
| US20160343147A1 (en) | Image search system, image search apparatus, and image search method | |
| CN105095882B (en) | Gesture recognition method and device | |
| CN103294194B (en) | A kind of interpretation method based on eye tracking and system | |
| CN105677043B (en) | The two stages adaptive training method of Mental imagery brain-computer interface | |
| KR20180111859A (en) | Intelligent robot wake-up method and intelligent robot | |
| CN109740113A (en) | Hyper parameter threshold range determines method, apparatus, storage medium and electronic equipment | |
| Huang et al. | Sap: Self-adaptive proposal model for temporal action detection based on reinforcement learning | |
| CN103914149A (en) | Gesture interaction method and gesture interaction system for interactive television | |
| CN111027403A (en) | Gesture estimation method, device, equipment and computer readable storage medium | |
| WO2021129144A1 (en) | Multi-screen display system and cursor switching control method therefor | |
| CN107194216A (en) | A kind of mobile identity identifying method and system of the custom that swiped based on user | |
| CN105653037A (en) | Interactive system and method based on behavior analysis | |
| CN105630325A (en) | Fingerprint identification based gesture operation method and apparatus as well as terminal | |
| CN104731307A (en) | Somatic action identifying method and man-machine interaction device | |
| CN103793056A (en) | Mid-air gesture roaming control method based on distance vector | |
| Hu et al. | LF-ViT: Reducing spatial redundancy in vision transformer for efficient image recognition | |
| CN110084235A (en) | Information collecting method, device and acquisition equipment based on point of interest | |
| CN105847252B (en) | A kind of method and device of more account switchings | |
| CN113326829A (en) | Method and device for recognizing gesture in video, readable storage medium and electronic equipment | |
| CN103632668A (en) | Method and apparatus for training English voice model based on Chinese voice information | |
| CN113268143B (en) | Multimodal man-machine interaction method based on reinforcement learning | |
| CN107346207B (en) | Dynamic gesture segmentation recognition method based on hidden Markov model | |
| Zhao et al. | Gaussian processes for flow modeling and prediction of positioned trajectories evaluated with sports data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| 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: 20181225 |