CN101716079A - Brainprint identity identification authentication method based on multi-characteristics algorithm - Google Patents

Abstract

The invention belongs to the technical field of biomedicalengineering and information, in particular to a brainprint identity identification authentication method based on a multi-characteristics algorithm. By fully utilizing characteristic fusion, different brain wave signals are generated by training a trainee with different stimulus patterns; the brain wave signals generated by the stimulus are analyzed by using a multi-characteristic extraction method, the extracted individual characteristic signals generate individual classifiers, target characteristics are extracted form the acquired brain wave signals when in identity identification or authentication, and are input to corresponding individual classifiers, output results are uniformized to judge the results, and judged results appear in probability distribution way. The identification rate can exceed 90 percent by selecting a suitable threshold value. The brainprint identity identification authentication technology can change biological characteristics, the trainee only imagines a certain motion, the motion can be used as a password for storage, keys, passwords or other tools are not needed when in application, and by only remembering the passwords, the operation can be completed.

Description

Brainprint identity identification authentication method based on multi-characteristics algorithm

Technical field

The invention belongs to biomedical engineering and areas of information technology, particularly relate to Brainprint identity identification authentication method based on multi-characteristics algorithm.

Technical background:

Identification and checking are the important prerequisites that guarantees national public safety and information security.In applications such as national security, public security, the administration of justice, ecommerce, E-Government, safety inspection, guard monitors, all need identification accurately and evaluation.The verification method of traditional identify label article (as key, certificate, bank card) etc.; Another kind of is the verification method that indicates knowledge (as user name, password etc.) based on identity.But marking object is lost easily or is palmed off, and sign knowledge is forgotten easily or decoded.Biometrics identification technology has brought the possibility that realizes for this hope.People may forget or lose their card or password, but can never forget or lose the biological characteristic of oneself, as people's face, fingerprint, iris, palmmprint, brain wave etc.Biometrics identification technology (Biometrics) is meant that high-tech means is close combines by computer and optics, acoustics, biosensor and biostatistics's principle etc., utilizes inherent physiological property of human body (as fingerprint, people's face, iris, brain wave, pulse etc.) or behavior characteristics (as person's handwriting, voice, gait etc.) to carry out the authentication of personal identification.Biometrics identification technology do not have can forget, be difficult for forge or stolen, carry and advantage such as available whenever and wherever possible, safer, secret than traditional identity identifying method, make things convenient for.At present, comparative maturity and several biometrics identification technologies of having application prospect most comprise fingerprint, people's face, people's face thermogram, iris, retina, hand-type, vocal print and signature etc.Wherein, iris identification and fingerprint recognition are acknowledged as the most reliable two kinds of biological identification technologies.

People's any physiology or behavior characteristics just can be used as biological characteristic in principle and are used for the identity discriminating as long as it satisfies following condition: (1) universality, and everyone has; (2) uniqueness, everyone is different; (3) stability is constant a certain period; (4) collection property, quantitative measurement easily.Certainly, only satisfying above condition may not be feasible, and actual system also should consider: (1) performance, i.e. Shi Bie accuracy, speed, robustness and for reaching the requirement resource needed; (2) acceptability, people are to the acceptance level of this bio-identification; (3) but fraudulence, can be by the out-trick complexity of system of the method for subjectivity swindle.

At present, there is such or such problem in biological identification technology commonly used, and for example recognition of face is powerless for twins; Application on Voiceprint Recognition is imitated easily; Fingerprint recognition can be subjected to the influence of finger injuries, also usurps easily simultaneously.Brain electricity (EEG) signal is not only a very useful clinical diagnosis instrument, and is a kind of living things feature recognition instrument that well is used for authentication.At first, it possesses universality, and everyone has brain wave; Secondly because causing, differences such as everyone brain characteristic, mode of thinking, memory have different EEG signal (brain stricture of vagina) between men; The 3rd, EEG also possesses certain stability, and within a certain period of time, the EEG signal can keep relative stability, and is last, and the EEG signal is convenient to gather.Biological recognition system based on the EEG signal can reach certain accuracy and fast speeds, and can not produce any injury to human body, and people also can accept.Because the EEG signal derives from the thinking activities of brain, be difficult to forge, the robustness of system is very high.The EEG signals of human brain researched and analysed show, Different Individual can produce different Nerve impulse reactions in different brain districts, according to this EEG signals difference, can extract individual EEG signals feature, utilize the sorting algorithm of setting, can make that EEG signals possesses the individual specificity.Based on above analysis, be a kind of new identity identification system that application prospect is arranged based on the biological identification system of brain electricity.

At present at the early-stage based on the identification and the Authentication Research of brain stricture of vagina, basically all in the laboratory research stage, but also be common signal processing and mode identification method, the subject matter that exists be that stability is not high, diverse ways specificity difference, that is to say, effective especially for some crowd someway, and may be with regard to poor effect to other crowds.What also might occur is, is applicable to someone's special time period someway, and may poor effect to the other times section.This also is a principal element of restriction Brainprint identity Verification System practical application.

Summary of the invention

The objective of the invention is, a kind of personal identification system of quick, many features is provided, employing comes the experimenter is carried out Brainprint identity identification based on multi-characteristics algorithm, the experimenter produces the motion imagination, only select for use in the calculating with the relevant EEG signals of the imagination of moving and analyze, the multiple signal processing method of imagining the classification of brain electricity with being used to move integrates, and realizes the identification and the authentication of identity by neutral net.

For achieving the above object, the present invention based on the Brainprint identity identification authentication method of multi-characteristics algorithm, adopts the input signal of EEG signals as identification by the following technical solutions, it is characterized in that, may further comprise the steps:

1) set stimulation programs in computer, the experimenter makes the different motion imaginations according to the requirement of experiment in the stimulation programs;

2) original EEG signals is gathered by the EEG amplifier;

3) EEG signals of gathering is amplified with A/D change;

4) with the EEG signals of gathering according to be integrated in the embedded chip or the computer software system in algorithm extract characteristic quantity, EEG signals is handled, comprise public average reference, filtering, extraction characteristic quantity, produce personalized grader, set up many tagsorts device, eeg data input neural network with the unknown, determine BP neural network structure and suitable experimenter's motion imagination type, discern or authenticate;

5) the threshold value variable is set, many tagsorts device output is judged, be higher than the authentication success that is considered as of threshold value.

Described personalized grader is that known eeg data input neural network is learnt, determine the BP neural network structure, every kind of characteristic quantity is applied to train the BP neutral net separately, and designs separately at every experimenter, each feature and every kind of motion imagination type.

Described many tagsorts device is to be gathered by the personalized grader of difference to form, and can arbitrarily change different characteristic quantity inputs, and the number of characteristic quantity can be regulated, and adjusts characteristic quantity automatically according to classifying quality.

Described identification or authentication are with the form of probability distribution the every class result who occurs to be selected.

The invention has the advantages that 1) adopt the input signal of EEG signals as identification, fingerprint different from the past, iris etc.; 2) use many tagsorts device, and every experimenter has been trained separately, set up personalized grader; When 3) every section EEG data being authenticated, no longer be simple 100% or 0%, but a probability distribution; 4) be provided with threshold value according to level of security; 5) adopted motion imagination EEG signal initiatively, made the brain stricture of vagina become subjective controlled password, also be applicable to other EEG signals (bringing out current potential etc.) certainly such as visual evoked potential, incident; 6) realized identification and authentication function, identification refers to judges it is whose EEG signals from some individuals' EEG signals, judge whether a certain EEG signals is target person's EEG signals and authenticate to refer to; 7), select to be fit to experimenter's motion imagination type automatically by different experimenters' characteristics.The present invention is fit to physical disabilities, and all kinds of crowds such as dysopia have the better suitability.

Description of drawings

The chosen position figure of Fig. 1 electrode of the present invention

Fig. 2 feature extraction flow chart of the present invention

Fig. 3 identification flow chart of the present invention

The probability distribution graph of Fig. 4 unknown sample U of the present invention (a)

The probability distribution graph of Fig. 5 unknown sample V of the present invention (b),

Fig. 6 different threshold values of the present invention are to the influence curve figure of classification rate

The specific embodiment

The invention will be further described below in conjunction with drawings and Examples, and referring to accompanying drawing 1-6, the Brainprint identity identification authentication method based on multi-characteristics algorithm may further comprise the steps:

1) set stimulation programs in computer, the experimenter makes the different motion imaginations according to the requirement of experiment in the stimulation programs;

2) original EEG signals is gathered by the EEG amplifier;

3) EEG signals of gathering is amplified with A/D change;

4) with the EEG signals of gathering according to be integrated in the embedded chip or the computer software system in algorithm extract characteristic quantity, EEG signals is handled, comprise public average reference, filtering, extraction characteristic quantity, produce personalized grader, set up many tagsorts device, eeg data input neural network with the unknown, determine BP neural network structure and suitable experimenter's motion imagination type, discern or authenticate;

5) the threshold value variable is set, many tagsorts device output is judged, be higher than the authentication success that is considered as of threshold value.

Described personalized grader is that known eeg data input neural network is learnt, determine the BP neural network structure, every kind of characteristic quantity is applied to train the BP neutral net separately, and designs separately at every experimenter, each feature and every kind of motion imagination type.

Described many tagsorts device is to be gathered by the personalized grader of difference to form, and can arbitrarily change different characteristic quantity inputs, and the number of characteristic quantity can be regulated, and adjusts characteristic quantity automatically according to classifying quality.

Described identification or authentication are with the form of probability distribution the every class result who occurs to be selected.

Specific embodiment:

1) experimenter is with the polar cap that powers on, original brain electricity (EEG) signal is to lead the EEG amplifier collection that meets 10/20 method that international electroencephalography can demarcate by 64, sample rate is 250Hz, is reference electrode with the left mastoid process, and the band filter passband is 1-50Hz, choose 6 electrodes and gather the EEG signals (C3 in 10/20 international standard that can demarcate of just international electroencephalography, C4, P3, P4, O1 and O2 be totally 6 electrode positions), gather experimenter's EEG signals that different motion is imagined process.

According to principle: people's brain is not when handling the sensation input or producing motion output, and it is the μ ripple that the EEG activity in the brain concentrates on motor cortex, concentrates on visual cortex and just shows as the β ripple.Experiment shows: imagery motion or warm-up all can be accompanied by reducing of μ ripple and β wave mode.This reduces to be called the relevant desynchronization (ERD) of incident; In contrast, when motion was finished and loosened, waveform just can increase, and this phenomenon is incident related synchronizationization (ERS).And ERD and ERS do not need to produce actual action, will produce in the motion imagination, and 6 electrodes that we select are all relevant with the motion imagination.

2) stimulation programs that basis configures on computer screen (the prompting experimenter begins imagery motion), the experimenter is according to requirement of experiment, make the motion imagination and comprise the limb motion imagination or organ movement's imagination, select the different motion of four classes to imagine (motion of imagination left hand, the right hand move, lower limb is moving and tongue moves), the experimenter is through training, be familiar with experimentation, can produce special brain.

3) EEG signals that collects is carried out input feature vector extraction system after the pretreatment.Pretreatment comprises carries out public average reference to signal, with the FIR wave filter of strict linear phase signal is carried out 8-30Hz filtering.

4) EEG feature extraction system extracts each experimenter's EEG signals feature.With the EEG signals of gathering according to according to be integrated in the embedded chip or the computer software system in following algorithm extract characteristic quantity.

The feature here can be any signal processing method, and the present invention is the example explanation with two category feature amounts.One category feature amount is the feature that single electrode possessed, for example autoregressive coefficient, linear complexity, energy spectral density, energy entropy; Another kind of characteristic quantity is two degrees of association between the electrode signal, for example phase place lock value, cross-correlation and mutual information.

Specific algorithm is as follows:

● autoregression (Auto Regression, AR) coefficient

The initial predicted error:

e _on＝b _on＝x _n n＝0，1，...，N-1 (1)

Reflection coefficient:

$a_{\mathrm{kk}}=\frac{-2\underset{n=k}{\overset{N-1}{\mathrm{\Σ}}}{b}_{k-1,n-1}^{*}{e}_{k-1,n}}{\underset{n=k}{\overset{N-1}{\mathrm{\Σ}}}|{b_{k-1,n-1}|}^2+{\left|e_{k-1,n}\right|}^2}k=\mathrm{1,2},...,p---\left(2\right)$

$a_{\mathrm{ki}}=a_{k-1,i}+a_{\mathrm{kk}}{a}_{k-1,k-i}^{*}i=\mathrm{1,2},...,k-1---\left(3\right)$

${\mathrm{\σ}}_k^2=(1-{\left|a_{\mathrm{kk}}\right|}^2){\mathrm{\σ}}_{k-1}^2$

Forecast error:

e _kn＝e _k-1，n+a _kkb _k-1，n-1

$b_{\mathrm{kn}}=b_{k-1,n-1}+a_{\mathrm{kk}}^{*}{e}_{k-1,n}---\left(4\right)$

Try to achieve a _P1, a _P2..., a _PpAnd σ _p ²After, be calculated as follows x _nPower spectral density

$P\left(f\right)=\frac{{\mathrm{\σ}}_p^2\mathrm{\Δt}}{{|1+\underset{k=1}{\overset{p}{\mathrm{\Σ}}}{a}_{\mathrm{pk}}{e}^{-j2\mathrm{\πfk\Δt}}|}^2}---\left(5\right)$

Coefficient a is as characteristic quantity.

● linear complexity (Linear Complexity, LC)

For certain L electrode lead signals, linear complexity is defined as so:

$\mathrm{\Ω}=\exp (-\underset{i=1}{\overset{L}{\mathrm{\Σ}}}{\mathrm{\ξ}}_i\log {\mathrm{\ξ}}_i)---\left(6\right)$

ξ wherein _i=λ _i/ ∑ _iλ _i, L * n matrix covariance matrix is:

$C=\frac{1}{n}\underset{k}{\mathrm{\Σ}}{u}_k{u}_k^{T}k=\mathrm{1,2},...L---\left(7\right)$

λ so _iIt is the eigenvalue of covariance matrix C.

To frequency band range 8-13Hz, 14-20Hz, 21-30Hz wave band calculate linear complexity as characteristic quantity.

● energy spectral density (energy spectrum density, ESD)

Energy spectral density is described signal energy, if x (t) is the finite energy signal, this signal spectra density Φ (ω) be signal Fourier transform value square, be calculated as follows:

$\mathrm{\Φ}\left(\mathrm{\ω}\right)={\left|\frac{1}{\sqrt{2\mathrm{\π}}}{\∫}_{-\∞}^{\∞}x\left(t\right)e^{-\mathrm{j\ωt}}\mathrm{dt}\right|}^2=\frac{X\left(\mathrm{\ω}\right)X^{*}\left(\mathrm{\ω}\right)}{2\mathrm{\π}}---\left(8\right)$

Wherein X (ω) is the Fourier transform of x (t), X ^*(ω) be its complex conjugate.

To frequency band range 8-13Hz, 14-20Hz, 21-30Hz wave band carry out energy spectral density and calculate characteristic quantity.

● the energy entropy (Energy Entropy, EE)

The energy entropy can reflect the Energy distribution information of EEG signals frequency space, has reflected the Energy distribution feature of EEG signals on time-domain and frequency-domain simultaneously.

If E ₁, E ₂... E _mBe the Energy distribution of signal x (t) on m frequency band, then the power spectrum on frequency domain can form a kind of division to signal energy.Signal gross energy E equals each component ENERGY E in window sometime _jSum, promptly

Di establishes for frequency j goes up the power spectrum value

P _j＝E _j/E (9)

∑ P then _j=1, so define corresponding energy entropy be:

$W_e=-\underset{j}{\mathrm{\Σ}}{P}_j\log P_j---\left(10\right)$

The calculating of energy entropy is defined within a certain regular length window ranges, is 2 to slide in the short time discrete Fourier transform signal by step-length, extracts characteristic quantity.

● the phase place lock value (Phase Locking Value, PLV)

Two synchronous measuring of signal are phase place lock value PLV, and the method is only considered the phase place of this signal.

PLV＝|<exp(j{Φ _i(t)-Φ _j(t)})>| (11)

Here, Φ _i(t), Φ _j(t) be electrode i, the instantaneous phase of j.

${\tilde{x}}_i\left(t\right)=\frac{1}{\mathrm{\&pi;}}\mathrm{PV}{\&Integral;}_{-\&infin;}^{\&infin;}\frac{x_i\left(\mathrm{\&tau;}\right)}{t-\mathrm{\&tau;}}\mathrm{d\&tau;}---\left(12\right)$

In the following formula definition,

Be time series x _i(t) Hilbert conversion, PV is meant Cauchy's principal value.

${\mathrm{\&Phi;}}_i\left(t\right)=\arctan \frac{{\tilde{x}}_i\left(t\right)}{x_i\left(t\right)}---\left(13\right)$

In the present invention, select six electrodes in the data description for use, left and right three electrodes match in twos, and each counter-electrodes is carried out the phase-locked value calculating of phase place, draw characteristic quantity.

● mutual information (Mutual Information, MI)

Mutual information is a kind of Useful Information tolerance in the theory of information, and it is meant two dependencys between the event sets.The mutual information of two incident x and y is defined as:

I(x，y)＝H(x)+H(y)-H(x，y) (14)

Wherein:

H(x，y)＝-∑p(x，y)logp(x，y) (15)

In the present invention, select six electrodes in the data description for use, left and right three electrodes match in twos, and each counter-electrodes are carried out mutual information calculate, and draw characteristic quantity.

● cross-correlation (Cross Correlation, CC)

In statistics, cross-correlation is used for representing two covariances between the random vector sometimes, is a tolerance that is used for representing similarity between two signals, usually by relatively being used for seeking the characteristic of unknown signaling with known signal.It is with respect to the function of time between two signals.Select six electrodes in the data description for use, left and right three electrodes match in twos, and each counter-electrodes is carried out cross-correlation calculation, draw characteristic quantity.

The present invention has also set up many tagsorts device, and be about to every kind of feature and be applied to train the BP neutral net separately, and at every experimenter, each feature and every kind of independent design personalized grader of motion imagination type.

Algorithm is as follows:

For a certain known EEG signals (known experimenter and known motion imagination type), (other characterization method are also passable, and process is similar to extract feature one by one according to previously described seven kinds of features.), as the input set of neutral net; According to every kind of motion imagination type, every independent learning training neutral net of experimenter, (motion imagination type)=(if any a plurality of experimenters, process is similar for 12 graders to have 3 (experimenter) * 4 for every kind of feature like this.); The present invention adopts neural network learning, and the input layer number is a number of features, and output layer node number is 1, and value is 1 during study.

When identification or when authenticating a certain unknown sample U, extract seven kinds of features earlier, in 12 graders that input has trained respectively, like this for every kind of feature, each grader all has output one by one, as following table:

Table: the grader output matrix of a certain unknown sample U.

In the table: A, B, C are three experimenters, and L, R, T, F represent imagination left hand, the right hand, tongue and lower limb motion respectively.

After seven kinds of characteristic quantities of this sample are exported by corresponding personalized grader, homogenization again after value is average.As can be seen from the table, the probability that this sample belongs to the motion of experimenter A imagination left hand is 40.4%, far moves above other experimenters and imagines the probability of type.

Each output result is averaged and homogenization after, simultaneously can obtain probability distribution graph, as Fig. 4.

According to probability distribution graph, and the threshold value that configures in advance, we just can judge whom this EEG signals belonged to.As Fig. 4, if threshold value is made as 25%, have only the output valve of A-L grader to surpass threshold value, then motion authenticates or identification as experimenter A imagination left hand with this sample in system.

Similarly, for another unknown sample V, probability distribution such as Fig. 5, given threshold gets 25% equally, the output valve of neither one grader surpasses threshold value, even with threshold value reduce to 15% also the output valve of neither one grader surpass threshold value, then sample V can not authenticate or be identified as any class of known 12 classes.

The present invention has introduced the notion of threshold value.According to different level of securitys, different threshold values can be set, as Fig. 4, given threshold is set in 50%, and then this sample can not authenticate or be identified as any known sample, needs authentication again, declares will descend to rate; And threshold value is located at 5%, and this sample may just be authenticated to be a plurality of known sample, and False Rate will improve.So threshold value can be used for regulating level of security, different threshold values are to the influence of classifying quality, as Fig. 6.

5) the eeg data input neural network of the unknown is discerned and authenticated.The experimenter has determined BP neural network structure and suitable he (she's) motion imagination type by behind the above-mentioned steps 1-4, and just can discern and authenticate this moment.

Remarks: identification refers to and select this section EEG signals from several experimenter is for which experimenter, multiselect one; Then for determining that whether this section EEG signals is certain experimenter, "Yes" is " not being " still, alternative for verification process.

The present invention only need add a cover eeg signal acquisition device (for example Neuroscan eeg recording system) and can implement.And be integrated in algorithm in the embedded chip or the computer software system in.

The specific embodiment is pressed accompanying drawing 1,2,3.Can realize through the following steps:

1) experimenter is trained,, gather experimenter's EEG signals feature of different motion imagination process, gather the signal that meets 6 electrodes of 10/20 international standard by the stimulus modelity training; Concrete electrode position is (number of electrode and position can change, and change the back the possibility of result and have difference) as shown in Figure 1.

2) EEG signals that collects is carried out pretreatment (specific algorithm as previously shown) back input feature vector extraction system.

3) referring to Fig. 2, the EEG feature extraction system extracts each experimenter's EEG signals feature, by BP neural network learning and identification, produces personalized neural network classifier, and training process finishes.

4) referring to Fig. 3, according to each experimenter's who extracts EEG signals feature, definite personalized neural network classifier is discerned or is authenticated each experimenter's identity.

As identification, on method, single signal processing method of no use is analyzed, and adopts the multiple signal processing method of current extraction EEG signals feature with motion imagination EEG signals in the present invention.Experimental result shows, selects appropriate threshold, and discrimination can surpass 90%.The biological identity identifying technology of brain stricture of vagina among the present invention can change biological characteristic (by imagination different motion), this and other biological identification technology is different, the experimenter only need imagine certain motion, it can be stored as password, do not need key, password and other instrument during application, only need think about it their password, just can complete operation.From now on, even can or recall with dream, custom as password.The present invention is intended to by many Feature Fusion, the advantage of different characteristic is concentrated in together, and characteristics algorithm to be not limited only to herein this several, may increase and decrease.Preliminary experiment is the result also prove, the present invention has solved the not high shortcoming of Brainprint identity Verification System stability in the past substantially.Certainly, feature neither be The more the better, and number of features is many more, calculates complicatedly more, and the time is long more; Feature is few more, and the effect of fusion is just poor more.As long as select suitable feature, just may obtain very satisfied effect.

The Brainprint identity Verification System can be applied to for identification and the higher place of authentication requesting, such as national security, police and judicial, finance etc. Since EEG signals possess be difficult to copy, can not force, live body exists, the specific characteristics such as repeat, in case after the electric sample devices miniaturization of brain, simplification and the portability, application prospect of the present invention will be very bright.

Claims (4)

1. the Brainprint identity identification authentication method based on multi-characteristics algorithm adopts the input signal of EEG signals as identification, it is characterized in that, may further comprise the steps:

1) set stimulation programs in computer, the experimenter makes the different motion imaginations according to the requirement of experiment in the stimulation programs;

2) original EEG signals is gathered by the EEG amplifier;

3) EEG signals of gathering is amplified with A/D change;

4) with the EEG signals of gathering according to be integrated in the embedded chip or the computer software system in algorithm extract characteristic quantity, EEG signals is handled, comprise public average reference, filtering, extraction characteristic quantity, produce personalized grader, set up many tagsorts device, eeg data input neural network with the unknown, determine BP neural network structure and suitable experimenter's motion imagination type, discern or authenticate;

5) the threshold value variable is set, many tagsorts device output is judged, be higher than the authentication success that is considered as of threshold value.

2. the Brainprint identity identification authentication method based on multi-characteristics algorithm according to claim 1, it is characterized in that: described personalized grader is that known eeg data input neural network is learnt, determine the BP neural network structure, every kind of characteristic quantity is applied to train the BP neutral net separately, and designs separately at every experimenter, each feature and every kind of motion imagination type.

3. the Brainprint identity identification authentication method based on multi-characteristics algorithm according to claim 1, it is characterized in that: described many tagsorts device is to be gathered by the personalized grader of difference to form, can arbitrarily change different characteristic quantity inputs, the number of characteristic quantity can be regulated, and adjusts characteristic quantity automatically according to classifying quality.

4. the Brainprint identity identification authentication method based on multi-characteristics algorithm according to claim 1 is characterized in that: described identification or authentication are with the form of probability distribution the every class result who occurs to be selected.

Images