CN105929966A - Peripheral device brain wave control method capable of learning adaptively - Google Patents

Abstract

The invention discloses a peripheral device brain wave control method capable of learning adaptively. The peripheral device brain wave control method capable of learning adaptively comprises the steps of acquiring a brain wave signal through a brain wave device; carrying out independent component analysis for the acquired multi-lead brain wave data; removing extra interference such as electrooculogram, myoelectricity and power frequency in the brain wave signal; calculating multiple power spectrum densities in the brain wave after FFT (Fast Fourier Transform); calculating a degree of brain activity through the multiple power spectrum densities; comparing the current degree of brain activity with the range of the learnt degree of brain activity in a specific state; generating a peripheral device control signal once the condition is satisfied; and carrying out adaptive learning correction for the state. The peripheral device brain wave control method capable of learning adaptively can efficiently remove brain wave interference signals, extract multiple power spectrum densities exactly, and can realize supervisory control of the brain in the specific state and control of the peripheral device.

Description

A kind of can adaptive learning E.E.G control ancillary equipment method

Technical field

The present invention relates to δ ripple in brain-computer interface device, θ ripple, α ripple and the extraction of β wave energy, brain liveness self adaptation Learning method, and the method carrying out peripheral unit control based on the wavy state of specific brain regions.

Background technology

Correlational study shows, utilizes EEG signals (EEG) can directly reflect the active state of brain, and EEG has become brain The important evidence of evaluation central nervous system's change widely used in active research.EEG signals can be decomposed into 4 basic joints Rule, i.e. δ ripple, θ ripple, α ripple and β ripple, as shown in table 1.

Table 1

The energy of these 4 rhythm and pace of moving things can change along with the change of brain liveness.When α ripple and β ripple are dominant advantage, show The consciousness of people is clear-headed, and brain enlivens；And when θ ripple and δ ripple account for dominant advantage, then indicate the confusion of people even The appearance of slight sleep.And when people is in a kind of particular state, the liveness index of human brain is the most stable, is obtained by E.E.G Fetching puts acquisition brain wave signal, is given by wireless communication transmissions and receives processor, and brain wave signal is gone interference to process by processor After i.e. available brain wave signal accurately, calculate human brain liveness according to brain wave signal, and the wavy state of specific brain regions learnt enlivened Degree interval compares and then controls ancillary equipment, and brain liveness when driver such as enters fatigue learns, when driving When the person of sailing enters fatigue driving state, processor can carry out early warning by control ancillary equipment in time, and to fatigue driving state Liveness carries out self adaptation state study, improves early warning precision.

Summary of the invention

It is an object of the invention to provide and a kind of the E.E.G of adaptive learning can control ancillary equipment method, to realize in time by control Ancillary equipment processed carries out early warning, and improves early warning precision.

In order to solve above technical problem, the concrete technical scheme that the present invention uses is as follows:

A kind of can adaptive learning E.E.G control ancillary equipment method, it is characterised in that comprise the following steps:

Step 1: utilize the brain wave signal of E.E.G device collection people

Step 2: the brain wave signal collected is carried out data prediction, removes obvious drift data, obtains observation signal and sets For x, carrying out independent component analysis, remove the clutter unrelated with EEG signals, detailed process is as follows:

It is transformed to matrix: x by linear for described observation signal x^~=Mx so that E [x^~x^~']=I, have after conversion:

x^~=Mx=MAS=BS

Wherein, M is whitening matrix, and B=MA is an orthogonal matrix, in order to solve orthogonal solution hybrid matrix w^Λ, make output Each component be separate；Described output y:

Y=(w^Λ)^TMx^~=(w^Λ)^TMx^~=(w^Λ)^TMAs

The kurtosis of the stochastic variable V of zero-mean is:

Kurt (t)=E [V⁴]-3(E[V²])²

By maximizing kurtosis, being separated one by one by source signal, its recurrence formula is as follows:

W (k)=E [x (w (k-1)^T)³]-3w(k-1)

Wherein, w=w_i(for a line of w), and | | w | |=1.

The clutter unrelated with EEG signals can be removed by above process, source signal is separated one by one.

Step 3: brain wave signal is carried out Fourier transformation, detailed process is as follows:

First the brain wave signal collected is preserved according to δ ripple, θ ripple, α ripple and β ripple frequency range segmentation, through fast Fourier transform Obtain Fourier's composition；The discrete fourier (DFT) of N point finite length sequence x (n) is:

$x\left(k\right)=DFT\[x\left(n\right)\]={\mathrm{\Σ}}_{n=0}^{N-1}x\left(n\right)w_N^{kn},0\≤k\≤N-1$

X (k) is the Fourier's composition after conversion, w_NRepresent what orthogonal sequence was concentratedI.e.

Step 4: calculate multiple power spectrum density in E.E.G

Brain wave signal after removing eye electricity, myoelectricity interference is designated as x₁, utilize x₁The power spectrum tried to achieve after carrying out Fourier transformation Density, reacts the status relation that corresponding δ ripple, θ ripple, α ripple and β ripple are shared in current E.E.G well；At FFT spectrum In analysis, the resolution of system is by sample frequency f_sDetermine, that is: with sampling number N

$\mathrm{\Δ}f=\frac{f_s}{N}$

Actual samples frequency is that 1000Hz is after Fourier transformation, it is thus achieved that x (k), k=1,2,3 ..., N；

P (k)=x (k)²/N

The power spectral density that P (k) estimates for utilizing period map method；

Step 5: calculate brain liveness

Brain wave signal is decomposed into 4 basilic rhythms, i.e. δ ripple, θ ripple, α ripple and β ripple；Described 4 basilic rhythms Energy can change along with the change of brain liveness；When α ripple and β ripple are dominant advantage, show that the consciousness of people is clear-headed, Brain enlivens；When θ ripple and δ ripple account for dominant advantage, show the appearance that the confusion of people is the most slightly slept；Therefore, Brain wave after independent component analysis process eliminates eye electricity and myoelectricity interference is dominated relation be analyzed, and calculate corresponding Characteristic parameter, thus realize the judgement of brain liveness and assessment；Corresponding to described δ ripple, θ ripple, α ripple, β ripple Frequency range be 1～3.8Hz, 4～7.8Hz, 8～12.8Hz, 13～30Hz respectively, corresponding energy is shown below:

$\left\{\begin{array}{c}{E}_{\mathrm{\δ}}=\ΣP_i,1\≤f\left(i\right)\≤3.8\\ E_{\mathrm{\θ}}=\ΣP_i,4\≤f\left(i\right)\≤7.8\\ E_{\mathrm{\α}}=\ΣP_i,8\≤f\left(i\right)\≤12.8\\ E_{\mathrm{\β}}=\ΣP_i,13\≤f\left(i\right)\≤30\end{array}\right.$

Wherein,E_δ, E_θ, E_α, E_βRepresent that δ ripple, θ ripple, α ripple, β ripple are corresponding Energy

Brain liveness H is by calculating with following formula

$H=\frac{E_{\mathrm{\α}}+E_{\mathrm{\β}}}{E_{\mathrm{\δ}}+E_{\mathrm{\θ}}};$

Step 6: if carrying out the study of particular state brain liveness, then arranged for gathering E.E.G number of times i under particular state Secondary, preserve particular state liveness maximum, H_maxWith minima H_minInterval [H_min,H_max]；

If the monitoring of particular state brain liveness, then arrange for gathering E.E.G number of times i time under particular state, calculate active Whether degree is at interval [H_min,H_max]: if liveness is at interval [H_min,H_max], then trigger peripheral unit control signal, for Improve the accuracy to the monitoring of particular state interval, the liveness under this state is carried out adaptive learning, persistent period 0.2s； If liveness range limit H_upWith lower limit H_downBeyond described particular state liveness [H_min,H_max] interval, in order to ensure Adaptive learning is undistorted, for liveness range limit H_upWith lower limit H_downCarry out defined below:

If H_up-H_max< 2%* (H_max-H_min)/2, then H_max=H_up

If H_min-H_down< 2%* (H_max-H_min)/2, then H_min=H_down

If meeting conditions above, the liveness that self-adaptative adjustment has stored is interval, otherwise keeps original liveness interval constant.

Present invention also offers a kind of E.E.G control peripheral system carrying out adaptive learning based on E.E.G, including brain wave signal Monitoring device, brain wave signal processing means；

Brain wave signal monitoring device is substantially carried out monitoring in real time brain wave signal, and by signal real-time Transmission to brain wave signal processing means.

Brain wave signal processing means include brain wave signal receiver module, liveness computing module, designated state liveness study module, Liveness state matching module.

Brain wave signal receiver module receives brain wave signal；

Liveness computing module carries out independent component analysis to brain wave signal, obtains E.E.G power spectral density, according to power spectral density Calculate liveness；

Designated state liveness study module is that the brain liveness to designated state stores；

Liveness state matching module is whether monitoring current active degree is in designated state liveness interval, if being in appointment district Between then send peripheral unit control signal；

Adaptive learning modules, can adaptive learning designated state liveness.

The present invention has beneficial effect.The present invention, by the brain wave signal extracted under particular state, calculates E.E.G liveness index, root Carry out state coupling according to liveness, when liveness is in appointment interval, i.e. produce peripheral unit control signal, it is achieved that to E.E.G The extraction of signal, state study, adaptive correction, and control to ancillary equipment during particular state, for the product of E.E.G product Product provide the implementation of scientific and efficient.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of peripheral device control method one example carrying out adaptive learning based on E.E.G

Fig. 2 is the peripheral unit control device one embodiment schematic flow sheet carrying out adaptive learning based on E.E.G

Fig. 3 is the Peripheral unit control system one example structure schematic diagram carrying out adaptive learning based on E.E.G

Fig. 4 driver's brain liveness monitoring figure.

Fig. 5 is the adaptive learning early warning success rate contrast test of the present invention.

The realization of the object of the invention, functional characteristics and advantage will in conjunction with the embodiments, are described further referring to the drawings.

Detailed description of the invention

Step will be implemented according to the present invention below, in conjunction with explanatory diagram and specific embodiment, further illustrate this below in conjunction with the accompanying drawings Bright technical scheme.

Fig. 1 is the schematic flow sheet that the present invention carries out peripheral device control method one example of adaptive learning based on E.E.G, according to Schematic flow sheet, the method includes the steps of.

Step 1, by E.E.G device gather brain wave signal, as shown in Figure 2.

Brain wave signal can be gathered, then by the wireless transmission means such as bluetooth, wifi E.E.G by wearable E.E.G device Signal is transferred to brain wave signal processing terminal, advantage of this is that and allows E.E.G harvester separate with signal processing apparatus, both may be used To be simulated particular state collection brain wave signal voluntarily by user, it is also possible to carried out when not affecting user given state by other people Signals collecting, is greatly improved brain wave signal collecting efficiency.

Step 2, obvious drift data is removed in the brain wave data pretreatment that collects, then carry out independent component analysis, go Except the clutter unrelated with EEG signals.

The brain wave signal collected due to E.E.G harvester potentially includes eye electricity and myoelectricity interference signal, if directly using original Signal calculates and analyzes, and necessarily causes brain wave signal to have error, is substantially reduced band signal each for particular state brain and adopts The accuracy of collection, therefore, it is necessary to the brain wave signal collecting E.E.G harvester carries out independent component analysis process, removes The clutter unrelated with EEG signals.

Assume source signal s (t)=[s of one group of independence₁(t),s₂(t),…,s_m(t)]^T, mix through linear system A, Obtain observation signal:

X (t)=[x₁(t),x₂(t),…,x_m(t)]^T

That is:

X (t)=A (s)

Source signal s (t) and hybrid system A are all unknown, and the most mixed x (t) can be observed or measure, at n >=m Under conditions of, if s at most comprises only a Gaussian process, it is possible to obtain u (t)=wx (t) by solving mixed matrix w, Making vector, make vector u approach s, simply in u, ordering and the ratio scale of each component may be different from s.Therefore, If finding the mixed matrix w of solution to make the component of u as far as possible independently, so u is exactly the estimation to s.Due to by directly side It is the most relatively difficult that method makes signal independence as far as possible carrying out decompose, and maximizes w^TThe Gauss degree of x can obtain an isolated component.

Fast independent component analysis (Fast Independent Component Analysis, FastICA) proposes in recent years Very effective data analysis tool, it is mainly used to extract original independent signal from blended data.It divides as signal From a kind of effective ways and paid close attention to widely.It is a kind of to find the one of non-Gaussian system maximum based on fixed point iteration Plant independent component analysis algorithm.

Owing to kurtosis as the object function of non-gaussian degree (degree of independence), can make the maximum (or minimum) of kurtosis change W, has determined that an isolated component, that is:

w^TX=(b) x

B is mixed signal parameter matrix.

If using w as the weight vector inputting x in neural algorithm, and limiting | | w | |=1, so following formula is set up

kurt(w^TX)=E [(w^Tx)⁴]-3(E[w^Tx])²=E [(w^Tx)⁴]-3||w||⁴

By maximizing kurtosis, source signal can be separated singly, its recurrence formula is as follows:

W (k)=E [x (w (k-1)^T)³]-3w(k-1)

Wherein, w=w_i(for a line of w), and | | w | |=1, k be number of signals

Specific algorithm is accomplished by

(1) initializing w (0), making its mould is 1, puts k=1.

(2) w (k)=E [x (w (k-1) is made^T)³]-3w (k-1), it is desirable to value can be by the sampling of a large amount of brain wave signal x vectors Put and calculate (such as 1000 point).

(3) using | | w | | to remove w (k), | | w | |, will it be unitization divided by its norm.

(4) if | w (k) w (k-1) | is unsatisfactory for being substantial access to, then puts k=k+1, it is back to (2nd) step, the most defeated Go out vector.

In order to estimate m isolated component, need to run above-mentioned algorithm m time.Different in order to ensure estimating one every time Isolated component, needs to add a simple rectangular projection, i.e. the 3rd step in above-mentioned algorithm in this circulates and adds one Individual conversion:

$w\left(k\right)=w\left(k\right)-\underset{j=1}{\overset{k=1}{\&Sigma;}}{w}_k^T{w}_j{w}_j$

Compared with other algorithms, it is a cube convergence that FastICA algorithm has the following advantages it, and other fixed-point algorithm is the most wired Property convergence.Compared with algorithm based on gradient, it is not necessary to select Learning Step or other parameter so that this algorithm is more easy to use, More reliable.An isolated component rather than all of component is the most only extracted due to this algorithm.So, if as long as extracted Certain component, has again enough prioris, it is possible to soon it is extracted, thus reduces amount of calculation.Additionally, not Pipe have positive kurtosis or negative kurtosis component it can extract.

Step 3: carry out Fourier transformation.

First the EEG signals collected is preserved according to δ ripple, θ ripple, α ripple and β ripple frequency range segmentation, through fast Fourier transform Obtain Fourier's composition.The discrete fourier (DFT) of N point finite length sequence x (n) is:

$x\left(k\right)=DFT\&lsqb;x\left(n\right)\&rsqb;={\mathrm{\&Sigma;}}_{n=0}^{N-1}x\left(n\right)w_N^{kn},0\&le;k\&le;N-1$

X (k) is the Fourier's composition after conversion, and n represents the n-th EEG signals, w_NRepresent what orthogonal sequence was concentratedI.e.

Step 4: calculate multiple power spectrum density in E.E.G

EEG signals after removing eye electricity, myoelectricity interference is designated as x₁.Then to x₁The power spectrum tried to achieve after carrying out Fourier transformation Degree, can reflect the status relation that corresponding δ ripple, θ ripple, α ripple and β ripple are shared in current E.E.G well.At FFT In spectrum analysis, the resolution of system is by sample frequency f_sDetermine, that is: with sampling number N

$\mathrm{\&Delta;}f=\frac{f_s}{N}$

Actual samples frequency is that 1000Hz is after Fourier transformation, it is thus achieved that x (k) (k=1,2,3 ..., N)

P (k)=x (k)²/N

The power spectral density that P (k) estimates for utilizing period map method.

Step 5: calculate brain liveness

EEG signals can be decomposed into 4 basilic rhythms, i.e. δ ripple, θ ripple, α ripple and β ripple.

The brain wave processed via independent component analysis after eliminating eye electricity and myoelectricity interference is dominated relation be analyzed, and calculate Corresponding characteristic parameter, it is possible to achieve judgement and the assessment to brain liveness.Corresponding to δ ripple, θ ripple, α ripple, β ripple Frequency range be 1～3.8Hz, 4～7.8Hz, 8～12.8Hz, 13～30Hz respectively.

$\left\{\begin{array}{c}{E}_{\mathrm{\&delta;}}=\&Sigma;P_i,1\&le;f\left(i\right)\&le;3.8\\ E_{\mathrm{\&theta;}}=\&Sigma;P_i,4\&le;f\left(i\right)\&le;7.8\\ E_{\mathrm{\&alpha;}}=\&Sigma;P_i,8\&le;f\left(i\right)\&le;12.8\\ E_{\mathrm{\&beta;}}=\&Sigma;P_i,13\&le;f\left(i\right)\&le;30\end{array}\right.$

Wherein,E_δ,E_θ,E_α,E_βRepresent δ ripple, θ ripple, α ripple, energy that β ripple is corresponding Amount.

The energy of these 4 rhythm and pace of moving things can change along with the change of brain liveness.When α ripple and β ripple are dominant advantage, show The consciousness of people is clear-headed, and brain enlivens；And when θ ripple and δ ripple account for dominant advantage, then indicate the confusion of people even The appearance of slight sleep.Therefore, α ripple and β ripple signal reaction brain active degree, θ ripple and δ ripple have reacted the blunt of brain The ratio of degree, α ripple and β ripple sum and θ ripple and δ ripple sum can be with combined reaction brain active degree, and this ratio is the biggest Represent that brain active degree is the highest, otherwise, this ratio the least reaction brain active degree is the lowest.

Therefore, brain liveness H can be by calculating with following formula:

$H=\frac{E_{\mathrm{\&alpha;}}+E_{\mathrm{\&beta;}}}{E_{\mathrm{\&delta;}}+E_{\mathrm{\&theta;}}}$

Brain liveness H comparing result is as shown in table 2 when several testees are regained consciousness again and time tired.

With brain liveness H contrast time tired when more than 2 testee of table is regained consciousness again

From table 2 it can clearly be seen that work as testee waking state, liveness is substantially high than being worth under fatigue state, it addition, Different people has certain gap clear-headed with liveness under fatigue state, so, the particular state liveness for different people has The feature of relative independentability.

Respectively to driver A and driver B on the run, monitored a brain liveness every 6 minutes, altogether monitoring 24 times, result is as shown in Figure 4.By liveness data dynamic in figure, can with driver along with the passage of drive time, Brain liveness declines, and it is different that different drivers decline degree.

Step 6: if carrying out the study of particular state brain liveness, then arrange for gathering E.E.G number of times i time under particular state, Preserve the interval [H of particular state liveness_min,H_max].Note: particular state refer to user be in fatigue driving, sleep-walking, The state such as doze off, drunk, at this moment collection E.E.G number of times can be freely set according to the persistent period of particular state and stability and adopt Collection interval, it is ensured that gather brain wave signal very big with the dependency of particular state, the most just can be implemented in the standard of particular state monitoring Really property.

Step 7: if the monitoring of particular state brain liveness, then arrange for gathering E.E.G number of times i time under particular state, calculate Whether liveness is at interval [H_min,H_max], then sending triggering peripheral unit control by processor if present in this interval Signal, processor can be the programmable devices such as single-chip microcomputer, PC and Android mobile phone, ancillary equipment include audio effect processing, The devices such as motor machine, can be arranged according to needs when there is brain particular state, flexibly as it is shown on figure 3, such as can lead to Cross single-chip microcomputer obtain brain wave signal and calculate brain liveness, and connecting sound equipment device, when user is in fatigue driving state, Brain liveness is in fatigue driving state, and Single-chip Controlling PA-system sends alarm.

, there is certain error with the brain liveness of virtual condition in step 8: the study of particular state is likely to be at emulation mode, in order to Improve the accuracy to the monitoring of particular state interval, the liveness under this state can be learnt, persistent period 0.2s again, If liveness range limit H_upWith lower limit H_downBeyond original particular state liveness [H_min,H_max] interval, in order to ensure Adaptive learning is undistorted, for liveness range limit H_upWith lower limit H_downIt is defined, as follows,

H_up-H_max< 2%* (H_max-H_min)/2, then H_max=H_up

H_min-H_down< 2%* (H_max-H_min)/2, then H_min=H_down

If meeting conditions above, the liveness that self-adaptative adjustment has stored is interval, otherwise keeps original interval constant.

Such as it is in fatigue driving state as driver, processing terminal can be operated by the other staff in driver's cabin and carry out this shape Liveness under state learns again, improves the accuracy of status monitoring.Table 3 is for enter 200 driver tired driving states Row state learns, and is divided into use adaptive learning series 1 and does not use adaptive learning function series 2, to fatigue driving state Monitoring accuracy carries out record, as shown in Figure 5.

As can be seen from Fig. 5, do not use the series 1 early warning success rate of adaptive learning to remain at about 92%, and use certainly The series 2 early warning success rate of adaptive learning gradually steps up along with the number of times increase used, it was demonstrated that use adaptive learning function can To significantly improve the study of brain particular state and calibration efficiency.

Claims (5)

1. one kind can adaptive learning E.E.G control ancillary equipment method, it is characterised in that comprise the following steps:

Step 1: utilize the brain wave signal of E.E.G device collection people；

Step 2: the brain wave signal collected is carried out data prediction, removes obvious drift data, obtains observation signal and is set to x, carries out independent component analysis, removes the clutter unrelated with EEG signals；

Step 3: brain wave signal is carried out Fourier transformation；

Step 4: calculate multiple power spectrum density in E.E.G；

Step 5: calculate brain liveness；

Step 6: if carrying out the study of particular state brain liveness, then arrange for gathering E.E.G number of times i time under particular state, preserve particular state liveness maximum, H_maxWith minima H_minInterval [H_min,H_max]；

If the monitoring of particular state brain liveness, then arrange for gathering E.E.G number of times i time under particular state, whether calculate liveness at interval [H_min,H_max]: if liveness is at interval [H_min,H_max], then trigger peripheral unit control signal, in order to improve the accuracy to the monitoring of particular state interval, the liveness under this state is carried out adaptive learning, persistent period 0.2s；If liveness range limit H_upWith lower limit H_downBeyond described particular state liveness [H_min,H_max] interval, undistorted in order to ensure adaptive learning, for liveness range limit H_upWith lower limit H_downCarry out defined below:

If H_up-H_max< 2%* (H_max-H_min)/2, then H_max=H_up

If H_min-H_down< 2%* (H_max-H_min)/2, then H_min=H_down

If meeting conditions above, the liveness that self-adaptative adjustment has stored is interval, otherwise keeps original liveness interval constant.

The most according to claim 1 a kind of can adaptive learning E.E.G control ancillary equipment method, it is characterised in that the detailed process of described step 2 is as follows:

It is transformed to matrix: x by linear for described observation signal x^~=Mx so that E [x^~x^～']=I, has after conversion:

x^~=Mx=MAS=BS

Wherein, M is whitening matrix, and B=MA is an orthogonal matrix, in order to solve orthogonal solution hybrid matrix w^Λ, each component making output y is separate, described output y:

Y=(w^Λ)^TMx^~=(w^Λ)^TMx^~=(w^Λ)^TMAS

The kurtosis of the stochastic variable V of zero-mean is:

Kurt (t)=E [V⁴]-3(E[V²])²

By maximizing kurtosis, being separated one by one by source signal, its recurrence formula is as follows:

W (k)=E [x (w (k-1)^T)³]-3w(k-1)

Wherein, w=w_i(for a line of w), and | | w | |=1.

The most according to claim 1 a kind of can adaptive learning E.E.G control ancillary equipment method, it is characterised in that the detailed process of described step 3 is as follows:

First the brain wave signal collected is preserved according to δ ripple, θ ripple, α ripple and β ripple frequency range segmentation, obtain Fourier's composition through fast Fourier transform；The discrete fourier (DFT) of N point finite length sequence x (n) is:

0≤k≤N-1

X (k) is the Fourier's composition after conversion, w_NRepresent what orthogonal sequence was concentratedI.e.。

The most according to claim 1 a kind of can adaptive learning E.E.G control ancillary equipment method, it is characterised in that the detailed process of described step 4 is as follows:

Brain wave signal after removing eye electricity, myoelectricity interference is designated as x₁, utilize x₁The power spectral density tried to achieve after carrying out Fourier transformation, reacts the status relation that corresponding δ ripple, θ ripple, α ripple and β ripple are shared in current E.E.G well；In FFT spectrum is analyzed, the resolution of system is by sample frequency f_sDetermine, that is: with sampling number N

Actual samples frequency is that 1000Hz is after Fourier transformation, it is thus achieved that x (k), k=1,2,3 ..., N；

P (k)=x (k)²/N

The power spectral density that P (k) estimates for utilizing period map method.

The most according to claim 1 a kind of can adaptive learning E.E.G control ancillary equipment method, it is characterised in that the detailed process of described step 5 is as follows:

Brain wave signal is decomposed into 4 basilic rhythms, i.e. δ ripple, θ ripple, α ripple and β ripple；The energy of described 4 basilic rhythms can change along with the change of brain liveness；When α ripple and β ripple are dominant advantage, showing that the consciousness of people is clear-headed, brain enlivens；When θ ripple and δ ripple account for dominant advantage, show the appearance that the confusion of people is the most slightly slept；Therefore, the brain wave after independent component analysis process eliminates eye electricity and myoelectricity interference is dominated relation and is analyzed, and calculate corresponding characteristic parameter, thus realize the judgement to brain liveness and assessment；Frequency range corresponding to described δ ripple, θ ripple, α ripple, β ripple is 1～3.8Hz, 4～7.8Hz, 8～12.8Hz, 13～30Hz respectively, and corresponding energy is shown below:

Wherein,E_δ, E_θ, E_α, E_βRepresent δ ripple, θ ripple, α ripple, energy that β ripple is corresponding；

Brain liveness H by calculating with following formula,