CN103300852A - Training method for controlling remotely-controlled trolley in electrocerebral way based on motor imagery - Google Patents

Abstract

The invention discloses a training method for controlling a remotely-controlled trolley in an electrocerebral way based on motor imagery. The training method includes an offline mode training method and an online feedback mode training method. The training method comprises the following specific steps: trolley motion, signal acquisition, data preprocessing, acquisition of CSP (Chip Scale Package) parameters, acquisition of classifier parameters, inspection of parameters, real-time signal acquisition, real-time data preprocessing, character extraction, obtaining of a classifying result, obtaining of a control command, control of the trolley motion and feedback stimulation. Due to the adoption of the training method, the influences of various negative emotions on a subject in the conventional training process are eliminated, and the importance of positive and active participation of the subject into an experimental process is considered fully; the training method has the advantages of good training effect and high quality of acquired data; and a practicable method is provided for human-computer interaction.

Description

The training method of based on motion imagination brain electric control remote operated vehicle

Technical field

The invention belongs to areas of information technology, more a step relates to the training method of the based on motion imagination brain electrical remote control dolly of using brain-computer interface (Brain-Computer Interface, BCI) system in life science.Imagine the training method of brain electric control remote operated vehicle among the present invention about motion.Can allow better tested being dissolved in the whole motion imagination training process, advance the quality that once promotes the EEG signals that gathers, thereby improve the classification performance of whole BCI system, realize better to external world the ground controls such as equipment such as mouse, wheelchair, switch, finally reach the purpose of utilizing the motion imagination to carry out rehabilitation training and more effectively improve muscle and neurological disorder person's life.

Background technology

When preparing and carry out the one-sided finger motion imagination, the corticocerebral functional connection of people changes, thereby cause the EEG signals energy of its offside brain motor sensory area mu and the beta rhythm and pace of moving things to weaken, and the EEG signals energy of its homonymy brain motor sensory area mu and the beta rhythm and pace of moving things strengthen.The energy variation of specific brain regions district characteristic frequency EEG signals during the one-sided finger motion of this imagination is called as the relevant phenomenon that desynchronizes of event.This phenomenon is to differentiate the most basic feature of left and right sides finger motion imagination EEG signals.Therefore, the direction that the experimenter moves and imagines is differentiated in the analysis of EEG signals when moving the imagination by the experimenter, thereby realizes the to external world control of device.At present, bring out the experimenter move the imagination method mainly contain visual stimulus and auditory stimulus.

Visual stimulus mainly is to provide stimulation to the experimenter by clear and definite direction signs, allows the experimenter according to the corresponding direction imagination of moving.Auditory stimulus generally then is used for aid prompting, and the prompting experimenter tests beginning or end etc.

Jiangxi Lantian College is at its patent application document " based on the game auxiliary control method of imagination brain electricity " (application number 200810107345.1, a kind of control method of coming assist control that stimulates by game has been proposed applying date 2008.11.05, publication number 101430600).The method provides stimulation by the side-to-side movement of simulation dolly in the game, has effectively improved the positive and initiative that the experimenter participates in testing, and has improved to a certain extent the quality of the EEG signals that collects.But, the deficiency that the method exists is, the game householder method that provides be in the simple utilization game dolly stimulate the experimenter, do not have last classification results Real-time Feedback to the experimenter, the experimenter was regulated in real time to motion imagination time and degree, real the realization do not allow the experimenter initiatively participate in the whole experiment, thereby can produce certain impact to whole experiment effect.

And for existing other training method, be that the simple left and right sides arrow that passes through on the display to provide stimulation to the experimenter, there is the easily shortcoming such as absent-minded, absent minded and doze of uninteresting hard to bear, the experimenter of long, process of training time, do not allow the experimenter participate in the whole experiment positively, whole experiment effect is had certain impact.

Summary of the invention

The object of the invention is to overcome the deficiency that above-mentioned existing based on motion is imagined the training method of brain electricity, propose a kind of training method of based on motion imagination brain electric control remote operated vehicle.The method takes into full account the impact of emotions such as easily being felt sleepy, be weary of in experimenter's training process, allows the experimenter have the initiative in hands in whole training process, is conducive to promote the quality of the EEG signals that collects, thereby obtains higher classification accuracy rate.

The main thought that realizes the inventive method is: when the user will control the remote operated vehicle move left and right, need not carry out any body language or action, only need the motion of imagination left hand or the right hand, at this moment, will produce specific ERD/ERS phenomenon on μ rhythm and the beta response in the EEG signals that corresponding sensation of movement region electrode records on the scalp, after computer obtains EEG signals, the spatial distribution characteristic of the ERD/ERS phenomenon by analyzing eeg data, can identify the user is in the motion of imagination left hand or right hand motion on earth, thereby control command is passed to the remote operated vehicle control section, the motion of control dolly, and feed back to the experimenter, allow the experimenter further adjust imagination process, thereby improve classification accuracy rate.

According to above-mentioned main thought, the inventive method mainly comprises two kinds of training methodes, off-line mode training method and online feedback pattern drill method;

The concrete steps of described off-line mode training method are as follows:

(1) moving of car:

By the side-to-side movement of remote operated vehicle remote controller control dolly, provide visual stimulus to the experimenter, cause experimenter's the one-sided finger motion of the imagination.

(2) signals collecting:

The electrode cap that the eeg signal acquisition system wears by the experimenter gathers the EEG signals that the experimenter imagines that one-sided finger motion produces.

(3) data pretreatment:

3a) space filtering: adopt the method for common average reference, the EEG signals of each electrode collection on experimenter's electrode cap is deducted the average of the EEG signals that all electrodes gather, obtain the EEG signals behind the common average reference space filtering;

3b) baseline correction: the EEG signals behind the common average reference space filtering is deducted baseline, obtain the EEG signals after the baseline correction;

3c) bandpass filtering: utilize finite impulse response filter, the EEG signals after the baseline correction is carried out bandpass filtering, obtain the EEG signals that frequency band is 4-40Hz;

3d) intercept signal section: utilize EEGLAB software, the intercepting experimenter imagines the EEG signals section in the one-sided finger motion process in the EEG signals behind bandpass filtering, obtains pretreated EEG signals.

(4) obtain the CSP parameter:

Utilize cospace mode method CSP that pretreated EEG signals is processed, obtain EEG signals characteristic parameter and CSP parameter.

(5) obtain classifier parameters:

Utilize the LDA classifier methods that the EEG signals characteristic parameter is processed, obtain classification accuracy rate value and LDA classifier parameters.

(6) inspection parameter:

Whether the classification accuracy rate value that obtains in the determining step (5) is more than 90%, if then parameter is qualified, carry out the concrete steps of online feedback pattern drill method, otherwise parameter is defective, carries out the step (1) of the concrete steps of off-line mode training method.

The concrete steps of described online feedback pattern drill method are as follows:

(7) live signal collection:

The electrode cap that the eeg signal acquisition system wears by the experimenter gathers the EEG signals that the experimenter imagines that one-sided finger motion produces.

(8) real time data pretreatment:

8a) space filtering: adopt the method for common average reference, the EEG signals of each electrode collection on experimenter's electrode cap is deducted the average of the EEG signals that all electrodes gather, obtain the EEG signals behind the common average reference space filtering;

8b) baseline correction: the EEG signals behind the common average reference space filtering is deducted baseline, obtain the EEG signals after the baseline correction;

8c) bandpass filtering: utilize finite impulse response filter, the EEG signals after the baseline correction is carried out bandpass filtering, obtain the EEG signals that frequency band is 4-40Hz;

8d) intercept signal section: utilize EEGLAB software, the intercepting experimenter imagines the EEG signals section in the one-sided finger motion process in the EEG signals behind bandpass filtering, obtains pretreated EEG signals.

(9) feature extraction:

Utilize the cospace mode method that pretreated EEG signals in the step (8) is processed, obtain the feature for classification.

(10) obtain classification results:

Utilize the LDA classifier methods that the feature that obtains in the step (9) is classified, obtain the intention that the experimenter moves and imagines, namely the experimenter is carrying out the left hand motion imagination or the right hand motion imagination.

(11) obtain control command:

The classification results that obtains in the step (10) is converted to left and right sides control command, and classification results is that the control command of the right hand motion imagination is 1, and classification results is that the control command of the left hand motion imagination is-1.

(12) control moving of car:

Controlled order in the step (11) is transferred to the wifi hardware module by wifi and based on ICP/IP protocol, the wifi hardware module is passed to control command the dolly remote controller again, control command is that 1 control dolly moves right, and control command is that-1 control dolly is to left movement.

(13) feedback stimulates:

With dolly motion result Real-time Feedback in the step (12) to the experimenter, the experimenter sees the course of dolly, and according to time and the degree of dolly course and self imagination time length to recently adjusting next imagery motion, the relatively short then experimenter of moving line increases the time of the motion imagination and strengthens the degree of the imagination, the relatively long degree that then suitably reduces the time of the motion imagination and weaken the imagination of moving line.

The present invention compared with prior art has following advantage:

First, the present invention is owing to adopting remote operated vehicle as visual stimulus, can transfer the enthusiasm that the experimenter participates in testing largelyr, effectively overcome in the prior art training process emotion such as tired, barren on tested impact, so that the present invention has advantages of the quality that can effectively improve the EEG signals that gathers and then the classification accuracy rate that improves EEG signals.

Second, the present invention is owing to adopting remote operated vehicle to carry out the Real-time Feedback stimulation, having overcome experimenter in the prior art training process can't be according to the classification results in real time effectively time of the adjustment movement imagination and the shortcoming of degree, so that the present invention has and can transfer greatly that the experimenter initiatively participates in the experimentation, and in real time adjustment movement imagination time and degree, thereby improve the advantage of the classification accuracy rate of EEG signals.

The 3rd, the present invention is owing to adopting the wifi wireless technology to realize by the control of EEG signals to remote operated vehicle, overcome training and control in the prior art and only rested on shortcoming in theoretical and the game, so that the present invention has advantages of that the muscle of being and neurological disorder person exchange a kind of effective and feasible mode that provides with the external world.

Description of drawings

Fig. 1 is training flow chart of the present invention.

The specific embodiment

The present invention will be further described below in conjunction with accompanying drawing 1.

Step 1, moving of car:

By the side-to-side movement of remote operated vehicle remote controller control dolly, provide visual stimulus to the experimenter, cause experimenter's the one-sided finger motion of the imagination.

Step 2, signals collecting:

The electrode cap that the eeg signal acquisition system wears by the experimenter gathers the EEG signals that the experimenter imagines that one-sided finger motion produces.EEG signals is obtained by the electrode cap that is worn on experimenter's head, and amplifies and the A/D converter conversion by eeg amplifier, and the input computer is with form storage and the demonstration of signal voltage amplitude.

In an embodiment of the present invention, the experimenter wears electrode cap, and puts on earphone, is sitting in the dolly of overlooking on the chair apart from about its 5m.The sample frequency of eeg signal acquisition system is 250Hz, and test electrode is respectively C3, Cz, and C4, the fluctuation codomain of EEG signals is ± 100 μ V.In image data step of the present invention, prompting mode has three kinds, and by acoustical signal prompting experimenter preparing experiment, the side-to-side movement by dolly comes prompting user to carry out right-hand man's imagination of moving.

In the embodiment of the invention, image data is to test beginning by the voice suggestion experimenter in the time of 0 second, and the experimenter enters the Preparatory work of experiment state, and begins by the voice suggestion campaign imagination at the 3rd second.Subsequently, the experimenter moves to the left or to the right by the remote controller remote operated vehicle, and continues 1.25 seconds.At the 4th second, the experimenter began to imagine corresponding finger motion, and continued 3 seconds.Had a rest 1.5 seconds to 2.5 seconds behind each embodiment.Left and right sides finger motion thought experiment each 120 times, random alignment on the order.

Step 3, the data pretreatment:

Utilize the preprocessing function of EEGLAB software that the EEG signals that gathers is carried out pretreatment.

At first, carry out space filtering.Adopt the method for common average reference, the EEG signals of each electrode collection on experimenter's electrode cap is deducted the average of the EEG signals that all electrodes gather, obtain the EEG signals behind the common average reference space filtering; The common average reference method of above-mentioned employing refers to that the EEG signals with each passage that collects deducts the meansigma methods of all channel datas.

Secondly, carry out baseline correction.Imagine that take the experimenter one-sided finger motion 200ms EEG signals before is as baseline.EEG signals behind the common average reference space filtering is deducted baseline, obtain the EEG signals after the baseline correction.Described baseline is imagined one-sided finger motion 200ms EEG signals before for the experimenter.

Again, carry out bandpass filtering.Utilize finite impulse response filter, the EEG signals after the baseline correction is carried out bandpass filtering, obtain the EEG signals that frequency band is 4-40Hz.

At last, carry out the signal segment intercepting.Utilize EEGLAB software, the intercepting experimenter imagines the EEG signals section in the one-sided finger motion process in the EEG signals behind bandpass filtering, obtains pretreated EEG signals.

Step 4 obtains the CSP parameter:

If the common space mode method is so that the signal of a certain task has maximum variance, meanwhile the signal in another task has minimum variance.Its ultimate principle is the Simultaneous Diagonalization of Covariance Matrices to two kinds of task signals, extracts the main component that is used for distinguishing two kinds of task signals.The specific implementation step is as follows:

The first step utilizes following formula to estimate the covariance matrix of left hand and the right hand two type games imagination EEG signals;

${\mathrm{\Σ}}_{\mathrm{\ω}}=\frac{1}{n_{\mathrm{\ω}}}\underset{i=1}{\overset{n_{\mathrm{\ω}}}{\mathrm{\Σ}}}{S}_{\mathrm{\ω}}^i{S}_{\mathrm{\ω}}^{\mathrm{iT}}$

Wherein, ∑ _ωBe the covariance matrix of class ω, class ω is the set of right-hand man's two type games imagination, when ω is the left hand motion imagination, and n _ωBe the number of left hand motion imagination EEG signals,

Be i the EEG signals behind the left hand motion imagination space filtering, i=1,2 ... n _ωWhen ω is the right hand motion imagination, n _ωBe the number of right hand motion imagination EEG signals,

Be i EEG signals that belongs to after space filtering is imagined in right hand motion, i=1,2 ... n _ω, T is the transposition symbol.

Second step, utilize following formula with about two Simultaneous Diagonalization of Covariance Matrices, obtain its common generalized eigenvector; With the mapping matrix of this generalized eigenvector as the common space pattern;

W＝diag([∑ ₁，∑ _r])

Wherein, W is the generalized eigenvector matrix that obtains after the diagonalization, ∑ ₁Be left hand covariance matrix, ∑ _rBe right hand covariance matrix, the diagonal matrix of diag () for taking advantage of the square matrix diagonal element to consist of;

The 3rd step is with front m column vector and rear m alternative mapping matrix of column vector formation of the mapping matrix of common space pattern

1≤m＜n/2 wherein, n is the number of the mapping matrix column vector of common space pattern;

The 4th goes on foot, and extracts the characteristic vector of the EEG signals of distinguishing the two type games imagination according to following formula;

$f=\log (\mathrm{diag}\left({\stackrel{\‾}{W}}^T{\mathrm{SS}}^T\stackrel{\‾}{W}\right)/\mathrm{tr}\left({\stackrel{\‾}{W}}^T{\mathrm{SS}}^T\stackrel{\‾}{W}\right))$

Wherein, f is the characteristic vector expression of two type games imagination EEG signals; Log () is logarithmic function; The diagonal matrix of diag () for taking advantage of the square matrix diagonal element to consist of;

Alternative mapping matrix for the common space pattern; T represents the transposition symbol; S is the subsignal behind the Time Domain Piecewise; Tr () for take advantage of the square matrix diagonal element and.

Step 5 obtains classifier parameters:

The optimal characteristics vector of the EEG signals that selection step 4 obtains is as test data, and the optimal characteristics vector of all the other EEG signals is as training data; The LDA grader utilizes training data to set up disaggregated model, and test data substitution disaggregated model is obtained class categories, and match stop classification and actual task classification are correctly classified or the result of misclassification; Successively with the optimal characteristics vector of each EEG signals as a test data, add up the classification results of all test datas, obtain classification accuracy rate.The specific implementation step is as follows:

The first step is divided into groups to n the characteristic vector that the cospace pattern obtains, and as test data, a remaining n-1 feature is as training data with a feature.

Second step utilizes following formula to obtain LDA classifier parameters and classification results;

Wherein, k is the number of classification, and value is 2, y _k(x) be the discriminant classification value, x works as y for needing the data characteristics of classification _k(x)＞0 o'clock, x belongs to left hand imagination class, works as y _kO'clock (x)＜0, x belongs to right hand imagination class,

Be the slope coefficient on classification plane, w _K0Offset parameter for the classification plane.

The 3rd step, obtain classification accuracy rate according to classification results obtained above, namely obtain divided by total number with correct number.

Step 6, whether inspection parameter is qualified:

Whether the classification accuracy rate value that obtains in the determining step 5 if then parameter qualified, carry out the concrete steps of online feedback pattern drill method, otherwise parameter defective more than 90%, carries out the step 1 of the concrete steps of off-line mode training method.

The concrete steps of described online feedback pattern drill method are as follows:

Step 7, the live signal collection:

The electrode cap that the eeg signal acquisition system wears by the experimenter gathers the EEG signals that the experimenter imagines that one-sided finger motion produces.

Step 8, the real time data pretreatment:

At first, carry out space filtering.Adopt the method for common average reference, the EEG signals of each electrode collection on experimenter's electrode cap is deducted the average of the EEG signals that all electrodes gather, obtain the EEG signals behind the common average reference space filtering.Described baseline is imagined one-sided finger motion 200ms EEG signals before for the experimenter.

Secondly, carry out baseline correction.EEG signals behind the common average reference space filtering is deducted baseline, obtain the EEG signals after the baseline correction.

Again, carry out bandpass filtering.Utilize finite impulse response filter, the EEG signals after the baseline correction is carried out bandpass filtering, obtain the EEG signals that frequency band is 4-40Hz.Described baseline is imagined one-sided finger motion 200ms EEG signals before for the experimenter.

At last, carry out the signal segment intercepting.Utilize EEGLAB software, the intercepting experimenter imagines the EEG signals section in the one-sided finger motion process in the EEG signals behind bandpass filtering, obtains pretreated EEG signals.

Step 9, feature extraction:

Adopt the cospace mode method, pretreated EEG signals is carried out feature extraction, obtain the EEG signals characteristic vector of the one-sided finger motion of the imagination.

The first step utilizes following formula to estimate the covariance matrix of left hand and the right hand two type games imagination EEG signals;

${\mathrm{\Σ}}_{\mathrm{\ω}}=\frac{1}{n_{\mathrm{\ω}}}\underset{i=1}{\overset{n_{\mathrm{\ω}}}{\mathrm{\Σ}}}{S}_{\mathrm{\ω}}^i{S}_{\mathrm{\ω}}^{\mathrm{iT}}$

Wherein, ∑ _ωBe the covariance matrix of class ω, class ω is the set of right-hand man's two type games imagination, when ω is the left hand motion imagination, and n _ωBe the number of left hand motion imagination EEG signals,

Be i the EEG signals behind the left hand motion imagination space filtering, i=1,2 ... n _ωWhen ω is the right hand motion imagination, n _ωBe the number of right hand motion imagination EEG signals,

Be i EEG signals that belongs to after space filtering is imagined in right hand motion, i=1,2 ... n _ω, T is the transposition symbol.

Second step, utilize following formula with about two Simultaneous Diagonalization of Covariance Matrices, obtain its common generalized eigenvector; With the mapping matrix of this generalized eigenvector as the common space pattern;

W＝diag([∑ _l，∑ _r])

Wherein, W is the generalized eigenvector matrix that obtains after the diagonalization, ∑ _lBe left hand covariance matrix, ∑ _rBe right hand covariance matrix, the diagonal matrix of diag () for taking advantage of the square matrix diagonal element to consist of;

The 3rd step is with front m column vector and rear m alternative mapping matrix of column vector formation of the mapping matrix of common space pattern

1≤m＜n/2 wherein, n is the number of the mapping matrix column vector of common space pattern;

The 4th goes on foot, and extracts n characteristic vector of the EEG signals of distinguishing the two type games imagination according to following formula;

$f=\log (\mathrm{diag}\left({\stackrel{\‾}{W}}^T{\mathrm{SS}}^T\stackrel{\‾}{W}\right)/\mathrm{tr}\left({\stackrel{\‾}{W}}^T{\mathrm{SS}}^T\stackrel{\‾}{W}\right))$

Wherein, f is the characteristic vector expression of two type games imagination EEG signals; Log () is logarithmic function; The diagonal matrix of diag () for taking advantage of the square matrix diagonal element to consist of;

Alternative mapping matrix for the common space pattern; T represents the transposition symbol; S is the subsignal behind the Time Domain Piecewise; Tr () for take advantage of the square matrix diagonal element and.

Step 10 obtains classification results:

Utilize the LDA classifier methods that n the characteristic vector that obtains in the step 9 classified, obtain classification results;

The first step is divided into groups to n the characteristic vector that the cospace pattern obtains, and as test data, a remaining n-1 feature is as training data with a feature.

Second step utilizes following formula to obtain LDA classifier parameters and classification results.

$y_k\left(x\right)=w_k^{T}x+w_{k0}$

Wherein, k is the number of classification, and value is 2, y _k(x) be the discriminant classification value, x works as y for needing the data characteristics of classification _k(x)＞0 o'clock, x belongs to left hand imagination class, works as y _kO'clock (x)＜0, x belongs to right hand imagination class,

Be the slope coefficient on classification plane, w _K0Offset parameter for the classification plane.

The 3rd step, obtain classification accuracy rate according to classification results obtained above, namely obtain divided by total number with correct number.

Step 11 obtains control command:

The classification results that obtains in the step 10 is converted to left and right sides control command, and classification results is that the control command of the right hand motion imagination is 1, and classification results is that the control command of the left hand motion imagination is-1.

Step 12, the control moving of car:

Controlled order in the step 11 is transferred to the wifi hardware module by wifi and based on ICP/IP protocol, the wifi hardware module is passed to control command the dolly remote controller again, control command is that 1 control dolly moves right, and control command is that-1 control dolly is to left movement.

Step 13, feedback stimulates:

With dolly motion result Real-time Feedback in the step 12 to the experimenter, the experimenter sees the course of dolly, and according to time and the degree of dolly course and self imagination time length to recently adjusting next imagery motion, the relatively short then experimenter of moving line increases the time of the motion imagination and strengthens the degree of the imagination, the relatively long degree that then suitably reduces the time of the motion imagination and weaken the imagination of moving line.

Claims (5)

1. the training method of based on motion imagination brain electric control remote operated vehicle mainly comprises two kinds of training methodes, off-line mode training method and online feedback pattern drill method;

The concrete steps of described off-line mode training method are as follows:

(1) moving of car:

By the side-to-side movement of remote operated vehicle remote controller control dolly, provide visual stimulus to the experimenter, cause experimenter's the one-sided finger motion of the imagination;

(2) signals collecting:

The electrode cap that the eeg signal acquisition system wears by the experimenter gathers the EEG signals that the experimenter imagines that one-sided finger motion produces;

(3) data pretreatment:

3a) space filtering: adopt the method for common average reference, the EEG signals of each electrode collection on experimenter's electrode cap is deducted the average of the EEG signals that all electrodes gather, obtain the EEG signals behind the common average reference space filtering;

3b) baseline correction: the EEG signals behind the common average reference space filtering is deducted baseline, obtain the EEG signals after the baseline correction;

3c) bandpass filtering: utilize finite impulse response filter, the EEG signals after the baseline correction is carried out bandpass filtering, obtain the EEG signals that frequency band is 4-40Hz;

3d) intercept signal section: utilize EEGLAB software, the intercepting experimenter imagines the EEG signals section in the one-sided finger motion process in the EEG signals behind bandpass filtering, obtains pretreated EEG signals;

(4) obtain the CSP parameter:

Utilize cospace mode method CSP that pretreated EEG signals is processed, obtain EEG signals characteristic parameter and CSP parameter;

(5) obtain classifier parameters:

Utilize the LDA classifier methods that the EEG signals characteristic parameter is processed, obtain classification accuracy rate value and LDA classifier parameters;

(6) inspection parameter:

Whether the classification accuracy rate value that obtains in the determining step (5) is more than 90%, if then parameter is qualified, carry out the concrete steps of online feedback pattern drill method, otherwise parameter is defective, carries out the step (1) of the concrete steps of off-line mode training method;

The concrete steps of described online feedback pattern drill method are as follows:

(7) live signal collection:

The electrode cap that the eeg signal acquisition system wears by the experimenter gathers the EEG signals that the experimenter imagines that one-sided finger motion produces;

(8) real time data pretreatment:

8a) space filtering: adopt the method for common average reference, the EEG signals of each electrode collection on experimenter's electrode cap is deducted the average of the EEG signals that all electrodes gather, obtain the EEG signals behind the common average reference space filtering;

8b) baseline correction: the EEG signals behind the common average reference space filtering is deducted baseline, obtain the EEG signals after the baseline correction;

8c) bandpass filtering: utilize finite impulse response filter, the EEG signals after the baseline correction is carried out bandpass filtering, obtain the EEG signals that frequency band is 4-40Hz;

8d) intercept signal section: utilize EEGLAB software, the intercepting experimenter imagines the EEG signals section in the one-sided finger motion process in the EEG signals behind bandpass filtering, obtains pretreated EEG signals;

(9) feature extraction:

Utilize the cospace mode method that pretreated EEG signals in the step (8) is processed, obtain the feature for classification;

(10) obtain classification results:

Utilize the LDA classifier methods that the feature that obtains in the step (9) is classified, obtain the intention that the experimenter moves and imagines, namely the experimenter is carrying out the left hand motion imagination or the right hand motion imagination;

(11) obtain control command:

The classification results that obtains in the step (10) is converted to left and right sides control command, and classification results is that the control command of the right hand motion imagination is 1, and classification results is that the control command of the left hand motion imagination is-1;

(12) control moving of car:

Controlled order in the step (11) is transferred to the wifi hardware module by wifi and based on ICP/IP protocol, the wifi hardware module is passed to control command the dolly remote controller again, control command is that 1 control dolly moves right, and control command is that-1 control dolly is to left movement;

(13) feedback stimulates:

With dolly motion result Real-time Feedback in the step (12) to the experimenter, the experimenter sees the course of dolly, and according to time and the degree of dolly course and self imagination time length to recently adjusting next imagery motion, the relatively short then experimenter of moving line increases the time of the motion imagination and strengthens the degree of the imagination, the relatively long degree that then suitably reduces the time of the motion imagination and weaken the imagination of moving line.

2. the training method of based on motion according to claim 1 imagination brain electric control remote operated vehicle is characterized in that: step 3a) and step 8a) described common average reference method refers to that the EEG signals with each passage that collects deducts the meansigma methods of all channel datas.

3. the training method of based on motion according to claim 1 imagination brain electric control remote operated vehicle is characterized in that: step 3b) and step 8b) described baseline imagines one-sided finger motion 200ms EEG signals before for the experimenter.

4. based on motion according to claim 1 is imagined the training method of brain electric control remote operated vehicle, and it is characterized in that: the common space mode method described in step (4) and the step (9) is:

The first step utilizes following formula to estimate the covariance matrix of left hand and the right hand two type games imagination EEG signals;

${\mathrm{\Σ}}_{\mathrm{\ω}}=\frac{1}{n_{\mathrm{\ω}}}\underset{i=1}{\overset{n_{\mathrm{\ω}}}{\mathrm{\Σ}}}{S}_{\mathrm{\ω}}^i{S}_{\mathrm{\ω}}^{\mathrm{iT}}$

Wherein, ∑ _ωBe the covariance matrix of class ω, class ω is the set of right-hand man's two type games imagination, when ω is the left hand motion imagination, and n _ωBe the number of left hand motion imagination EEG signals,

Be i the EEG signals behind the left hand motion imagination space filtering, i=1,2 ... n _ωWhen ω is the right hand motion imagination, n _ωBe the number of right hand motion imagination EEG signals, Be i EEG signals that belongs to after space filtering is imagined in right hand motion, i=1,2 ... n _ω, T is the transposition symbol;

Second step, utilize following formula with about two Simultaneous Diagonalization of Covariance Matrices, obtain its common generalized eigenvector; With the mapping matrix of this generalized eigenvector as the common space pattern;

W＝diag([∑ _l，∑ _r])

Wherein, W is the generalized eigenvector matrix that obtains after the diagonalization, ∑ _lBe left hand covariance matrix, ∑ _rBe right hand covariance matrix, the diagonal matrix of diag () for taking advantage of the square matrix diagonal element to consist of;

The 3rd step is with front m column vector and rear m alternative mapping matrix of column vector formation of the mapping matrix of common space pattern

1≤m＜n/2 wherein, n is the number of the mapping matrix column vector of common space pattern;

The 4th goes on foot, and extracts the characteristic vector of the EEG signals of distinguishing the two type games imagination according to following formula;

$f=\log (\mathrm{diag}\left({\stackrel{\‾}{W}}^T{\mathrm{SS}}^T\stackrel{\‾}{W}\right)/\mathrm{tr}\left({\stackrel{\‾}{W}}^T{\mathrm{SS}}^T\stackrel{\‾}{W}\right))$

Wherein, f is the characteristic vector expression of two type games imagination EEG signals; Log () is logarithmic function; The diagonal matrix of diag () for taking advantage of the square matrix diagonal element to consist of;

Alternative mapping matrix for the common space pattern; T represents the transposition symbol; S is the subsignal behind the Time Domain Piecewise; Tr () for take advantage of the square matrix diagonal element and.

5. based on motion according to claim 1 is imagined the training method of brain electric control remote operated vehicle, and it is characterized in that: the LDA classifier methods described in step (5) and the step (10) is:

The first step is divided into groups to n the characteristic vector that the cospace pattern obtains, and as test data, a remaining n-1 feature is as training data with a feature;

Second step utilizes following formula to obtain the LDA classifier parameters;

$y_k\left(x\right)=w_k^{T}x+w_{k0}$

Wherein, k is the number of classification, and value is 2, y _k(x) be the discriminant classification value, x works as y for needing the data characteristics of classification _k(x)＞0 o'clock, x belongs to left hand imagination class, works as y _kO'clock (x)＜0, x belongs to right hand imagination class,

Be the slope coefficient on classification plane, w _K0Offset parameter for the classification plane;

The 3rd step, obtain classification accuracy rate according to classification results obtained above, namely obtain divided by total number with correct number.

Images