CN104536572B - It is a kind of based on event related potential across the universal brain-machine interface method of individual - Google Patents
It is a kind of based on event related potential across the universal brain-machine interface method of individual Download PDFInfo
- Publication number
- CN104536572B CN104536572B CN201410840127.4A CN201410840127A CN104536572B CN 104536572 B CN104536572 B CN 104536572B CN 201410840127 A CN201410840127 A CN 201410840127A CN 104536572 B CN104536572 B CN 104536572B
- Authority
- CN
- China
- Prior art keywords
- mrow
- event related
- related potential
- msub
- individual
- 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
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000000694 effects Effects 0.000 claims abstract description 29
- 239000000284 extract Substances 0.000 claims abstract description 7
- 230000004438 eyesight Effects 0.000 claims abstract description 7
- 210000004556 brain Anatomy 0.000 claims description 19
- 230000000638 stimulation Effects 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 11
- 210000001595 mastoid Anatomy 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 6
- 230000003014 reinforcing effect Effects 0.000 claims description 3
- 238000011160 research Methods 0.000 abstract description 6
- 238000012356 Product development Methods 0.000 abstract description 2
- 238000012546 transfer Methods 0.000 abstract description 2
- 239000000523 sample Substances 0.000 description 16
- 238000010586 diagram Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 230000004936 stimulating effect Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 206010002026 amyotrophic lateral sclerosis Diseases 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000011534 incubation Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000001960 triggered effect Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004070 electrodeposition Methods 0.000 description 1
- 230000000763 evoking effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002045 lasting effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 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
- 210000003205 muscle Anatomy 0.000 description 1
- 230000007433 nerve pathway Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 210000000578 peripheral nerve Anatomy 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 208000020431 spinal cord injury Diseases 0.000 description 1
- 230000002269 spontaneous effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000004304 visual acuity Effects 0.000 description 1
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses it is a kind of based on event related potential across the universal brain-machine interface method of individual, the described method comprises the following steps：Using the vision induced event related potential signal of Farwell experimental paradigms, and collection event related potential signal；Event related potential signal is pre-processed, extracts P300 signals；Using integrated linear classifier group to being predicted across individual character identification rate, and analyze its recognition effect.The integrated classifier has preferably across individual identification effect, can obtain ideal prediction effect to unbred subject data, and confirmed in the great amount of samples of 55 subjects.Optimum implementation is intended using patent transfer, technological cooperation or product development.This invention can be used for controlling the fields such as external equipment, electronic entertainment, have important Research Significance and commercial value.
Description
Technical field
The present invention relates to brain-computer interface field, more particularly to it is a kind of based on event related potential across the universal brain machine of individual
Interface method.
Background technology
Brain-computer interface (brain computer interface, BCI) is one kind independent of peripheral nerve and muscle groups
The information channel of conventional brain output channel such as knit.The research original intention of brain-computer interface technology is in order to intact to those brains but just
Normal nerve pathway be damaged (such as spinal cord injury SCI, amyotrophic lateral sclerosis ALS) patient provide it is a kind of again with outside
The approach of boundary's communication exchange, these sufferers are helped to improve the quality of living from many aspects, such as：With extraneous communication, use
Artificial intelligence prosthese, is controlled to external equipment.With the fast development of electronic information technology, brain-computer interface technology is answered
With having obtained very big expansion, potential application value is also shown in terms of other outside neural rehabilitation field, is such as lived
The fields such as amusement, military affairs, traffic safety.
Traditional brain-computer interface is mostly individual specificity's brain-computer interface (subject-dependent BCI, SDBCI), newly
User is before use, needs gather the data of the user, be only suitable for the user for establishing through calibration process after a while
The specific model of itself, therefore system versatility is poor.The calibration process of long period can expend time and efforts, or even meeting
Reduce follow-up efficiency in actual use.
The content of the invention
The invention provides it is a kind of based on event related potential across the universal brain-machine interface method of individual, the present invention is with passing
Individual specificity's brain-computer interface of system is compared, it is possible to reduce or even new user is eliminated to the calibration process before system use, not
On the premise of excessive sacrifice accuracy, the universal performance of system is drastically increased, more conforming to the actual use of user needs
Ask, it is described below：
It is a kind of based on event related potential across the universal brain-machine interface method of individual, the described method comprises the following steps：
Using the vision induced event related potential signal of Farwell experimental paradigms, and collection event related potential signal；
Event related potential signal is pre-processed, extracts P300 signals；
Using integrated linear classifier group to being predicted across individual character identification rate, and analyze its recognition effect.
The step of vision induced event related potential signal using Farwell experimental paradigms is specially：
Interface is stimulated to include several conventional characters, the character on screen defers to row or column flash pattern, and character, which is lighted, to be held
Start row or column after continuous several seconds to flash at random；Each row or column flicker continues for some time, and subject goes out in each target character
Within write from memory now and count its number being lit.
The step of collection event related potential signal is specially：
Eeg amplifier, brain wave acquisition software Scan4.5, sample rate 1000Hz, with a width of 0.05- are led using 40
100Hz, the impedance of each lead is maintained at below 5k Ω；The signal of six leads is gathered altogether, it is right using left mastoid process as reference
Side mastoid process ground connection.
It is described that event related potential signal is pre-processed, extract P300 signals the step of be specially：
Event related potential signal, P300 signals and noise by pretreatment represent as follows：
xi(n)=s (n)+ni(n), i=1,2 ... N
Wherein, i represents different stimulation examinations time, and n represents n-th of sampled value in this time record, record N number of sampling altogether every time
Value, xi(n) the event related potential signal collected for this time, ni(n) it is background brain noise, s (n) indicates a desire to what is obtained
True P300 signals；
By xi(n) M superposed average is passed through, the estimation of P300 signals represents as follows：
It is described to utilize integrated linear classifier group to being specially the step of individual character identification rate is predicted：
The integrated classifier of an effect reinforcing, model f (x can be obtained with reference to all sub-classifiersk) write as：
Wherein R1SD spaces are represented, i.e. training sample comes from same subject；R2SI spaces are represented, i.e. training sample is distinguished
From different subjects, sgn is to take sign function；
Test data is predicted using model, the target character predicted is：
Cr|c=argmaxr|c[f(xk r|c)]
Wherein, argmax represents the value of consult volume that searching makes f (x) reach maximum, r | c is to make f (xk) corresponding to maximization
Row/column, the crossover location of row and column predicts the position of obtained target character, will prediction character and known target character
It is corresponding, judge the correctness of grader test result.
The beneficial effect of technical scheme provided by the invention is：The present invention devises a kind of novel grader foundation side
Method, the electric electricity with non-targeted Stimulation of The Brain of the goal stimulus brain from different subjects is intersected and establishes sub-classifier, strengthen target and
The non-targeted effect to character recognition, while the otherness of individual is weakened, multiple sub-classifiers are integrated into stronger point of performance
Class device group, the integrated classifier have preferably across individual identification effect, and unbred subject data can be obtained more
Preferable prediction effect, and confirmed in the great amount of samples of 55 subjects.
Brief description of the drawings
Fig. 1 is a kind of flow chart across the universal brain-machine interface method of individual based on event related potential；
Fig. 2 is the schematic diagram for stimulating interface；
Wherein, a is prompting target character；B is the line flicker containing target character；C is blank stage.
Fig. 3 is stimulation timing diagram；
Fig. 4 is electrode placement positions；
Fig. 5 is grader schematic diagram；
Wherein, a is SI model classifiers schematic diagrames；B is SD model classifiers schematic diagrames.
Fig. 6 is SI and SD character recognition accuracy contrast schematic diagrams；
Fig. 7 is SI accuracy distribution schematic diagrams；
Fig. 8 is SI and SD accuracy gap distribution schematic diagrams；
Fig. 9 is the change contrast schematic diagram that SI models and SD models add with subject data to be measured.
Embodiment
To make the object, technical solutions and advantages of the present invention clearer, embodiment of the present invention is made below further
It is described in detail on ground.
Event related potential (event-related potentials, ERP) is a kind of special brain evoked potential, and it is logical
Crossing imparting stimulates with special psychological meaning, is triggered using a variety of or multiple stimulations, reflects the god in brain cognitive process
Through electrophysiological change.P300 is a kind of typical ERP endogenous components, and its amplitude is related to stimulating the predictability occurred, its
The individual difference of incubation period and amplitude is more obvious, and difficulty is brought for the foundation of general brain-computer interface.Due to based on P300
BCI systems be non-intrusion type, hardware device has portable and cheap advantage, in addition with the spelling based on other models
Device is compared, and P300 has the advantages that MIMD, accuracy is higher, is that currently the only one kind may apply to outside laboratory.Cause
This, has critically important application potential for the research across individual brain-computer interface based on ERP.
Across individual brain-computer interface (subject-independent BCI, SIBCI) new user can be made direct without calibration
System is used.For the brain computer interface application more and more developed towards normal person, suitable for a large number of users
General brain machine interface system be also more of practical significance and commercial value.
It is the system flow chart of the present invention as shown in Figure 1, stimulates interface with E-Prime software programmings, shown by computer
Device, which is presented, to stimulate, the ERP responses of vision induced subject.The ERP signals of offline collection subject, are stored in computer, pass through
In the stages such as pretreatment, feature extraction, pattern-recognition, obtain data results.
101：Using the vision induced event related potential signal of Farwell experimental paradigms (mainly P300 signals), and adopt
Collect event related potential signal；
Experimental section has raised 55 subject (ages altogether：23.2 ± 3.9, man 38, female 17), the normal visual acuity of all subjects
Or it is normal after correction, before without the use experience of P300 brain-computer interfaces.
Interface is stimulated to be presented by computer display screen, comprising 6*6 conventional characters, as shown in Figure 2.In experimentation, subject
It is sitting on the chair of one meter or so of range display, eye gaze screen simultaneously completes corresponding task according to prompting.The stimulation of experiment
Timing diagram is as shown in figure 3, the character on screen defers to traditional row or column flash pattern, in opening for each round (character)
Begin, target character is lit, and prompts the target to be watched attentively of this wheel of subject, as shown in Figure 2 a.After character lights lasting 5s
Start row or column to flash at random.The duration of each row or column flicker is 100ms, as shown in Figure 2 b, between flashing twice
A length of 75ms during blank, as shown in Figure 2 c.Need to complete 10 repetition flickers for each target character, flicker every time includes
6 rows dodge and 6 row dodge, therefore in the round of each character, target character occurs 20 times altogether.The task of subject exists
Within write from memory when each target character occurs and count its number being lit, the purpose of the task is to make subject's notice collection
In, to induce bigger P300 signals when goal stimulus occurs.Every subject need to complete the experiment of 20 characters, for
For different subjects, the order of this 20 characters is fixed identical.
Instrument used in collection event related potential signal leads eeg amplifier for the 40 of the production of Neuroscan companies
NuAmps, coordinate brain wave acquisition software Scan4.5, sample rate 1000Hz, with a width of 0.05-100Hz, resistance is led by real time more
The impedance of each lead is maintained at below 5k Ω by anti-measurement display, to reduce the interference noise in event related potential signal.
Experiment gathers the signal of this six leads of Fz, Cz, Pz, PO7, PO8 and Oz altogether, using left mastoid process as reference, right mastoid process ground connection,
8 electrodes (six leads of Fz, Cz, Pz, PO7, PO8, Oz and left and right mastoid process totally eight electrodes) are fixed on homemade electrode
On cap, electrode position is as shown in Figure 4.
102：Event related potential signal is pre-processed, extracts P300 signals；
It is related to the event collected first before feature extraction and Classification and Identification is carried out to event related potential signal
Current potential primary signal carries out some pretreatments, its purpose is to reduce the artefact in event related potential signal, removes event
Noise jamming in related potential signal, strengthen signal to noise ratio.
In preprocessing process, first to event related potential signal carry out 40Hz LPF, then it is down-sampled extremely
200Hz, 1 bandpass filtering for arriving 10Hz is next carried out to it again, according to Nyquist's theorem, most sample rate is down at last
20Hz.During this, LPF can remove useless High-frequency Interference information, and bandpass filtering can remove baseline drift
Move.And data volume can be reduced while information integrity is ensured by reducing sample rate, so as to improve follow-up operation efficiency.
The process entirely pre-processed substantially increases the signal to noise ratio of event related potential signal.
P300 signals are a kind of event related potential compositions with strict lock when property, and the analysis to it is mainly from time domain ripple
Shape is set out in itself, and coherence average algorithm is that one of the most frequently used method of small-signal is extracted under strong noise background, utilizes this side
Method can effectively extract P300 signals.
Event related potential signal, P300 signals and noise by pretreatment is represented as follows：
xi(n)=s (n)+ni(n), i=1,2 ... N (1)
Wherein, i represents different stimulation examinations time, and n represents n-th of sampled value in this time record, record N number of sampling altogether every time
Value, xi(n) the event related potential signal (i.e. primary signal) collected for this time, ni(n) it is background brain noise, s (n) tables
Show the true P300 signals for wanting to obtain.The event related potential signal that record obtains is passed through into M superposed average, P300 signals
Estimation can be represented with equation below：
Assuming that for each subject each time record P300 signals be held essentially constant, and spontaneous brain electricity and other
Noise is random presentation, therefore average can be considered zero, i.e. Section 2 in formula (2) can be 0
Therefore, the signal to noise ratio of the P300 signals after coherence average is handled improvesTimes.So, repeatedly stimulating
Additive process in, random background noise is cancelled out each other, with stimulate have fixing lock when relation P300 signals gradually strengthen, so as to
Extract more obvious P300 signals.It is mainly that speed is fast using the advantages of coherent averaging technique, method is easily achieved.
103：Using integrated linear classifier group to being predicted across individual character identification rate, and analyze its identification effect
Fruit.
Linear discriminant analysis (Linear Discriminant Analysis, LDA) is that one kind is applied to statistics, pattern
Identification and the classic algorithm in machine learning field.Its basic thought be for itself being difficult to two separated classes or multiclass sample,
A best projection direction is found, the vector can be such that sample i.e. to be sorted passes through after the linear transformation of the projection vector,
Class inherited is maximum, and difference is minimum in class.After so conversion, the dimension of initial data feature can be not only reduced, also
The classifying quality to initial data can be improved.
LDA has quite varied application in the brain-computer interface field based on P300, and has reached gratifying effect
Fruit, algorithm create weight coefficient according to the importance of feature, even if effect of the bigger feature of signal intensity in classification is heavier
Will.
In research before, LDA graders are all that height relies on individual, training sample and test sample all from
The same subject, the presence of individual difference cause these graders to have very strong specificity, and this just significantly limit base
In the universal performance of P300 brain-computer interface.And in the present invention, by integrated study technology, general ERP information is closed
And establish an integrated classifier with stronger versatility based on LDA sub-classifiers.
Integrated Algorithm combines multiple models, to obtain the prediction effect all better than any one Component Model.It is different from
Traditional data with the same subject are trained, when establishing submodel, by the goal stimulus ERP from different subjects
(tERPs) intersected modeling with non-targeted stimulation ERP (nERPs), i.e., all tERPs and nERPs for training pattern are
Respectively from different subjects, this avoid in traditional modeling method the problem of individual specificity.Although Different Individual
ERP signals can on incubation period and amplitude difference, but the difference between tERPs and nERPs is consistent, and in this method
Each submodel be all based on tERPs and nERPs difference and establish, this difference be exactly character recognition it is crucial according to
According to therefore, this modeling method more focuses on the difference between tERPs and nERPs and weakens the ERP signal differences between individual
It is different, it significantly more efficient can evade the individual specificity of model, the weaker sub-classifier of these classification performances is integrated,
Test data is identified simultaneously, using the unified result of most of sub-classifiers as the prediction result of integrated model, significantly
Improve the universal performance across individual body Model.
Specifically, it is assumed that share N name subjects, the sub-classifier between i-th subject and jth name subject is determined
Plan function can be write as
dij(xk)=ωijxk+biji；J, k, ＜ N (3)
Wherein, ωijBe by the goal stimulus sample of i-th subject and the non-targeted stimulation sample of jth name subject according to
The projection vector for training to obtain according to LDA algorithm, bijIt is bias term, xkIt is the test sample from kth name subject.Therefore for
For all samples of all subjects, N*N sub-classifier is established altogether.
The integrated classifier of an effect reinforcing, model f (x can be obtained with reference to all sub-classifiersk) can write
Into
Wherein R1SD spaces are represented, i.e. training sample comes from same subject；R2SI spaces are represented, i.e. training sample is distinguished
From different subjects, sgn is to take sign function；
Just as introduced earlier, the foundation across people's model is realized in this way, so as to weaken individual difference
Influence.
Generally, for the stimulation presentation process for each target character, every 6 line flickers and every 6 row dodge
The generation of P300 signals can be once triggered in bright respectively, test data is predicted using model accordingly, predicted
Target character is：
Cr|c=argmaxr|c[f(xk r|c)] (5)
Wherein, argmax represents the value of consult volume that searching makes f (x) reach maximum；R/c is to make f (xk) corresponding to maximization
Row/column, the crossover location of row and column predicts the position of obtained target character, will prediction character and known target character
It is corresponding, judge the correctness of grader test result.Using 10 folding cross validations with ensure the stability of classification accuracy rate and
Reliability.Fig. 5 is grader schematic diagram, and figure (a) is SD models, and the model is built by the off-line data of oneself the last week of subject to be measured
Vertical to form, to the subject, the real time data of oneself is predicted in online experiment.Figure (b) is SI models, and the model is by except treating
The off-line data foundation for surveying remaining subject the last week beyond subject forms, and sub-classifier is respectively by the target from different subjects
ERP and non-targeted ERP are established, the integrated LDA graders of these sub-classifiers composition, are not included in SI models during on-line testing to be measured
It is tested the information of itself..
Tested below with specific to verify the feasibility of this method, it is described below：
Overall across the individual identification effect of 55 subjects is analyzed, Fig. 6 shows 55 subject SD model (subjects
The data of person oneself are trained, and the data of oneself are tested) (data of remaining subject are trained, and subject is certainly with SI models
Oneself data are tested) recognition effect contrast.Wherein dotted line represents the Mean accurate rate of recognition of SI models, light shaded areas
For the standard deviation of Different Individual, solid line represents the Mean accurate rate of recognition of SD models, and dark-shaded represents the mark of Different Individual
Quasi- deviation.As can be seen from Fig. 6, all with number of repetition is stimulated, (every 6 row 6 arranges to have dodged at random to be referred to as the prediction accuracy of two kinds of models
Once repeat) increase and rise.It is consistent with the result of anticipation, utilize the prediction result of the personal data institute established model of subject
All the time the model built better than other several subject's data, it is respectively 1 time, 5 times and 10 when stimulating number of repetition specifically
When secondary, SD result is respectively 40.27%, 86.45% and 94.64%, and SI result is 34.91% respectively, 73.64% He
86.64%.ERP signals existing otherness between individuals is reflected across the decline of individual identification accuracy.On the other hand, to the greatest extent
The accuracy that pipe is identified across people not as good as individual specificity's model prediction effect, but more than 86% average accuracy it is some should
In be enough it is received, it means that the universal brain-computer interface based on ERP has certain feasibility and research valency
Value.
Although across the average accuracy of individual identification be only 86.64%, in order to probe into the specific manifestation of subject, further
Analyze the result of every subject and count the distribution of its accuracy.Fig. 7 result is the subject accuracy distribution map of SI models, can
To see, under 10 repetitive stimulations, subject of the accuracy higher than 90% has accounted for the 65.45% of total number of persons, and 20% people is just
Although true rate is less than 90% but also more than 70%, and only 14.55% (8 people or so) of result less than 70%.This result
The specific classifying quality for having reacted SI models, it is seen that most of subject can obtain very high recognition correct rate, only very
At least part of subject's accuracy is relatively low, this further demonstrates not calibrated for most people, directly carries out across people
Identification can reaches gratifying result, and bigger possibility is provided for universal brain-computer interface.
The SD and SI of different subjects effect difference distribution confirm this conclusion from another point of view, as shown in Figure 8.Should
As a result the accuracy for subtracting SI by SD accuracy obtains, and difference is smaller, represents that SI effect and SD are closer.Stimulation is repeated 10 times
When, it is only poor less than 10% between 74.55% SI results and SD in subject, wherein also including indivedual subject SI better than SD's
Situation.Gap accounts for the 10.90% of subject sum in 10-20% crowd, and the difference of two kinds of model results is more than 20% subject
There are 8 people, only account for the 14.55% of total number of persons.It can be seen that most of subject can obtain the result suitable with SD effects using SI, this
A part of crowd can omit in calibration process the step of building individual special purpose model, and directly application is basic to realize across individual body Model
The function of universal brain-computer interface.
In order to inquire into effect of the SI models for quickening calibration procedure, the character recognition of SD models and SI models is calculated respectively
Accuracy and the relation for training character number, as a result as shown in figure 9, abscissa represents the data volume institute for training in figure
Corresponding character number is, it is necessary to which explanation is a little that SD models completely by subject to be measured, established by the data of itself, now horizontal seat
Target numeral represents the character number corresponding to for the data of the subject of training pattern, and beyond SI models are by subject to be measured
Subject data established, now the numeral of abscissa represents to add the word corresponding to subject to be measured data itself into SI models
Accord with number.Solid line in figure represents the average result of SI models, and dark-shaded represents the standard deviation of Different Individual, and dotted line represents
The average result of SD models, the standard deviation of light shadow representation Different Individual.It can be seen that ought not yet it add to be measured
During subject's data in itself, the character identification rate of SI models is 86.55%, and now SD models are obviously 0, are adding 5
Before the data of character of subject to be measured itself, the character recognition effects of SI models is better than SD models, after 5 characters, SD moulds
The effect of type is just gradually slightly above SI models, in fact, its recognition effect difference and unobvious of the latter two.The result shows,
When no subject's data to be measured add, SI models can also have preferable recognition effect, with subject's information to be measured
Add, the recognition effect of SI models can faster rise to highest level than SD model, therefore compared with SD models, and SI models can be with
Save the calibration time used, two kinds of models of corresponding SI and SD, statistics discrimination reaches the subject to be measured itself needed when 90%
Character quantity corresponding to data, when being calibrated by saving of [N (SD)-N (SI)]/N (SD) calculating SI models compared to SD models
Between percentage, wherein N (SD) represent SD Model Identification rates is reached character number corresponding to the data volume of 90% needs, N (SI)
Expression make SI Model Identification rates reach the subject its data amount to be measured of 90% needs corresponding to character number, obtain 93.93%
± 13.17% result, this result, which reflects SI, can significantly reduce the calibration process before use, for brain-computer interface system
For the new user of system, SI model realization classifications of task can be directly utilized, and the effect can be with user's its data
Online adaptive and get a promotion rapidly.
In summary, the present invention propose it is a kind of based on event related potential across the universal brain-machine interface method of individual,
It can be established using this method in brain machine interface system across individual body Model, with traditional individual specificity's brain machine interface system phase
Than greatling save the prover time before new user's use, saving energy, strengthen practical effect.Optimum implementation
Plan uses patent transfer, technological cooperation or product development.This invention can be used for controlling the neck such as external equipment, electronic entertainment
Domain, there is important Research Significance and commercial value.
To the model of each device in addition to specified otherwise is done, the model of other devices is not limited the embodiment of the present invention,
As long as the device of above-mentioned function can be completed.
It will be appreciated by those skilled in the art that accompanying drawing is the schematic diagram of a preferred embodiment, the embodiments of the present invention
Sequence number is for illustration only, does not represent the quality of embodiment.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc., it should be included in the scope of the protection.
Claims (4)
1. it is a kind of based on event related potential across the universal brain-machine interface method of individual, it is characterised in that methods described includes
Following steps：
Using the vision induced event related potential signal of Farwell experimental paradigms, and collection event related potential signal；
Event related potential signal is pre-processed, extracts P300 signals；
Interception, which stimulates, to be started signal in rear 700ms and is used as feature, and end to end form of the feature of six leads is used to classify
Characteristic vector, the vector dimension is 6*0.7*20=84；
Using integrated linear classifier group to being predicted across individual character identification rate, and analyze its recognition effect；
Wherein, it is described to utilize integrated linear classifier group to being specially the step of individual character identification rate is predicted：
The integrated classifier of an effect reinforcing, model f (x can be obtained with reference to all sub-classifiersk) write as：
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>N</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<mi>s</mi>
<mi>g</mi>
<mi>n</mi>
<mo>&lsqb;</mo>
<msub>
<mi>d</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>&rsqb;</mo>
<mo>;</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>,</mo>
<mi>k</mi>
<mo>&Element;</mo>
<mfenced open = "{" close = "}">
<mtable>
<mtr>
<mtd>
<mrow>
<msub>
<mi>R</mi>
<mn>1</mn>
</msub>
<mo>:</mo>
<mi>i</mi>
<mo>=</mo>
<mi>j</mi>
<mo>=</mo>
<mi>k</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<mi>R</mi>
<mn>2</mn>
</msub>
<mo>:</mo>
<mi>i</mi>
<mo>&NotEqual;</mo>
<mi>j</mi>
<mo>&NotEqual;</mo>
<mi>k</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein, R1SD spaces are represented, i.e. training sample comes from same subject；R2Represent SI spaces, i.e., training sample respectively from
Different subjects, sgn are to take sign function；
Test data is predicted using model, the target character predicted is：
Cr|c=argmaxr|c[f(xk r|c)]
Wherein, argmax represents the value of consult volume that searching makes f (x) reach maximum；R | c is to make f (xk) the corresponding row of maximization/
Row, the crossover location of row and column predict the position of obtained target character, and prediction character is corresponding with known target character,
Judge the correctness of grader test result；
Wherein, dij(xk)=ωijxk+bij；I, j, k ＜ N
ωijIt is according to linear discriminant by the goal stimulus sample of i-th subject and the non-targeted stimulation sample of jth name subject
Parser trains obtained projection vector, bijIt is bias term, xkIt is the test sample from kth name subject, N is subject
Sum.
2. it is according to claim 1 it is a kind of based on event related potential across the universal brain-machine interface method of individual, it is special
Sign is, is specially the step of the event related potential signal vision induced using Farwell experimental paradigms：
Interface is stimulated to include several conventional characters, the character on screen defers to row or column flash pattern, and character is lighted persistently several
Start row or column after second to flash at random；Each row or column flicker continues for some time, and subject is when each target character occurs
Within write from memory and count its number being lit.
3. it is according to claim 1 it is a kind of based on event related potential across the universal brain-machine interface method of individual, it is special
Sign is, is specially the step of the collection event related potential signal：
Eeg amplifier is led using 40, brain wave acquisition software Scan4.5, sample rate 1000Hz, will with a width of 0.05-100Hz
The impedance of each lead is maintained at below 5k Ω；The signal of six leads is gathered altogether, is connect by reference, right mastoid process of left mastoid process
Ground.
4. it is according to claim 1 it is a kind of based on event related potential across the universal brain-machine interface method of individual, it is special
Sign is, described that event related potential signal is pre-processed, extract P300 signals the step of be specially：
Event related potential signal, P300 signals and noise by pretreatment represent as follows：
xi(n)=s (n)+ni(n), i=1,2 ... N
Wherein, the stimulation examination that i represents different is secondary, and n represents n-th of sampled value in once recording, every time the common N number of sampled value of record, xi
(n) the event related potential signal collected for this time, ni(n) it is background brain noise, s (n) indicates a desire to obtain true
P300 signals；
By xi(n) M superposed average is passed through, the estimation of P300 signals represents as follows：
<mrow>
<mover>
<mi>s</mi>
<mo>^</mo>
</mover>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>M</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>M</mi>
</munderover>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>s</mi>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mfrac>
<mn>1</mn>
<mi>M</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>M</mi>
</munderover>
<msub>
<mi>n</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
<mo>.</mo>
</mrow>
2
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410840127.4A CN104536572B (en) | 2014-12-30 | 2014-12-30 | It is a kind of based on event related potential across the universal brain-machine interface method of individual |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410840127.4A CN104536572B (en) | 2014-12-30 | 2014-12-30 | It is a kind of based on event related potential across the universal brain-machine interface method of individual |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104536572A CN104536572A (en) | 2015-04-22 |
| CN104536572B true CN104536572B (en) | 2017-12-05 |
Family
ID=52852112
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410840127.4A Expired - Fee Related CN104536572B (en) | 2014-12-30 | 2014-12-30 | It is a kind of based on event related potential across the universal brain-machine interface method of individual |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104536572B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104978035B (en) * | 2015-07-27 | 2018-04-10 | 中国医学科学院生物医学工程研究所 | Brain machine interface system and its implementation based on body-sensing electric stimulus inducing P300 |
| ITUB20153680A1 (en) * | 2015-09-16 | 2017-03-16 | Liquidweb Srl | Assistive technology control system and related method |
| CN107463250B (en) * | 2017-07-11 | 2019-11-22 | 天津大学 | The method for improving P300 spelling device using effect under Mental Workload state |
| CN108108763B (en) * | 2017-12-25 | 2021-07-23 | 中国科学院深圳先进技术研究院 | Electroencephalogram classification model generation method and device and electronic equipment |
| CN111012343A (en) * | 2019-11-28 | 2020-04-17 | 天津大学 | Cross-individual brain-computer interface based on Pearson correlation coefficient weighting |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102509103A (en) * | 2011-09-29 | 2012-06-20 | 天津大学 | Identifying method based on visual evoked P3 potential |
| CN102793540A (en) * | 2012-06-14 | 2012-11-28 | 天津大学 | Method for optimizing audio-visual cognitive event-related potential experimental paradigm |
| CN103699226A (en) * | 2013-12-18 | 2014-04-02 | 天津大学 | Tri-modal serial brain-computer interface method based on multi-information fusion |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050085744A1 (en) * | 2003-10-20 | 2005-04-21 | Stmicroelectronics S.R.I. | Man-machine interfaces system and method, for instance applications in the area of rehabilitation |
| US9058473B2 (en) * | 2007-08-29 | 2015-06-16 | International Business Machines Corporation | User authentication via evoked potential in electroencephalographic signals |
| US8239030B1 (en) * | 2010-01-06 | 2012-08-07 | DJ Technologies | Transcranial stimulation device and method based on electrophysiological testing |
-
2014
- 2014-12-30 CN CN201410840127.4A patent/CN104536572B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102509103A (en) * | 2011-09-29 | 2012-06-20 | 天津大学 | Identifying method based on visual evoked P3 potential |
| CN102793540A (en) * | 2012-06-14 | 2012-11-28 | 天津大学 | Method for optimizing audio-visual cognitive event-related potential experimental paradigm |
| CN103699226A (en) * | 2013-12-18 | 2014-04-02 | 天津大学 | Tri-modal serial brain-computer interface method based on multi-information fusion |
Non-Patent Citations (2)
| Title |
|---|
| P300 Speller中基于AdaBoost SVM的导联筛选研究;綦宏志等;《仪器仪表学报》;20120531;第33卷(第5期);第985-990页，论文第1页第1栏-第5页第2栏 * |
| 运动想象脑电信号的识别方法及应用;刘净瑜;《中国优秀硕士学位论文全文数据库信息科技辑》;20090915(第09期);I137-17，中文摘要、正文第16、33-48页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104536572A (en) | 2015-04-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104536572B (en) | It is a kind of based on event related potential across the universal brain-machine interface method of individual | |
| US10384064B2 (en) | Biomimetic multichannel neurostimulation | |
| CN103472922A (en) | Destination selecting system based on P300 and SSVEP (Steady State Visual Evoked Potential) hybrid brain-computer interface | |
| CN104503580B (en) | A kind of recognition methods to Steady State Visual Evoked Potential brain-computer interface target | |
| CN109299751B (en) | EMD data enhancement-based SSVEP electroencephalogram classification method of convolutional neural model | |
| CN107168524B (en) | Steady-state visual evoked potential classification method based on deep learning hybrid model | |
| CN105956624B (en) | Mental imagery brain electricity classification method based on empty time-frequency optimization feature rarefaction representation | |
| CN111783942B (en) | Brain cognitive process simulation method based on convolutional recurrent neural network | |
| CN111714118B (en) | Brain cognition model fusion method based on ensemble learning | |
| CN105877766A (en) | Mental state detection system and method based on multiple physiological signal fusion | |
| CN103150023A (en) | System and method for cursor control based on brain-computer interface | |
| CN107981997B (en) | A kind of method for controlling intelligent wheelchair and system based on human brain motion intention | |
| CN103955270B (en) | Character high-speed input method of brain-computer interface system based on P300 | |
| CN105852885A (en) | N400 evoked potential lie detection method based on improved extreme learning machine | |
| CN108960182A (en) | A kind of P300 event related potential classifying identification method based on deep learning | |
| CN107577343A (en) | It is a kind of based on the notice of haptic device and electroencephalogramsignal signal analyzing training and evaluating apparatus | |
| CN109645994A (en) | A method of based on brain-computer interface system aided assessment vision positioning | |
| CN101828921A (en) | Identity identification method based on visual evoked potential (VEP) | |
| CN113205074B (en) | Gesture recognition method fusing multi-mode signals of myoelectricity and micro-inertia measurement unit | |
| CN105496404A (en) | Assessment method of auxiliary CRS-R scale auditory shock based on brain-computer interface | |
| CN106569604A (en) | Audiovisual dual-mode semantic matching and semantic mismatch co-stimulus brain-computer interface paradigm | |
| CN103019382B (en) | A kind of brain-machine interface method utilizing Evoked ptential to reflect brain subjective motivation signal | |
| CN104461007A (en) | Driver-car interactive system assisting driver based on electroencephalograms | |
| CN103300852A (en) | Training method for controlling remotely-controlled trolley in electrocerebral way based on motor imagery | |
| Ko et al. | Enhancing the hybrid BCI performance with the common frequency pattern in dual-channel EEG |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 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 |
Granted publication date: 20171205 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |