CN102755162A - Audio-visual cognitive event-related electroencephalogram-based identification method - Google Patents

Abstract

The invention discloses an audio-visual cognitive event-related electroencephalogram-based identification method, which comprises the following steps of: performing wavelet packet decomposition and reconstruction on a series of filtered independent components to acquire electroencephalogram signals within electroencephalogram activity source rhythm ranges, and performing empirical mode decomposition on electroencephalogram signals of each rhythm to obtain corresponding feature mode components which reflect feature rhythm vibration modes; extracting inter-cognitive event phase locking strength features, inter-event synchronous features and driving and response coupling features from the feature mode components by adopting an inter-trial coherence index, a phase intrinsic coherence index, a phase shift intrinsic coherence index and a partial direct coherence index; extracting the power spectrum features of an audio-visual cognitive event-related electroencephalogram by adopting event-related spectral perturbation; and classifying the inter-cognitive event phase locking strength features, the inter-event synchronous features, the driving and response coupling features and the power spectrum features of cognitive events of a plurality of subjects by adopting a support vector machine multi-classifier to obtain identification results. According to the audio-visual cognitive event-related electroencephalogram-based identification method increases the identification rate.

Description

A kind of personal identification method based on the relevant brain electricity of the cognitive incident of audiovisual

Technical field

The invention belongs to the identification field, particularly a kind of personal identification method based on the relevant brain electricity of the cognitive incident of audiovisual.

Background technology

Along with the progress of society and the raising of quality of life, the protection of information security will concern the interests of individual and even entire society.Now, the identification and the authentication question of identity have all been related in a lot of fields.For example: aspect information security, be applied to the safe transfer of information Code And Decode and resource.The mode of tradition " recognizing people with thing " can't satisfy the demand in the present age because reliability is low, lose etc. drawback easily, the substitute is the biological characteristic that has by people self, and for example: fingerprint and iris wait and carry out identification.Yet each technology all has limitation, and researcheres are made great efforts for exploring new living things feature recognition means untiringly.Research shows, though for same outside stimulus or main body in the same incident of thinking, the EEG signals that brain produced of different people also are different, promptly the brain electricity has significant individual difference.Based on this point, consider simultaneously that the brain electricity has that universality is strong, relatively stable, high specificity, be difficult to duplicate and advantage such as forgery, the someone proposes the brain electricity is applied in the middle of the identification as a kind of novel biological characteristic.Particularly complicated task, need that the experimenter highly participates in, with the closely-related task of thinking activities, the brain that is inspired electricity individual difference is also obvious more, thereby is more suitable for the normal form of bringing out as identification.

For cognitive EEG signals, to investigate its corresponding time and spatial energy variation and phase change in each subtask incident usually.The traditional analysis method of cognitive brain electricity can be carried out along time domain and frequency domain both direction respectively.On the time domain, carry out superposed average in cognitive incident in the time of will observing brain electric array lock, obtain ERP (event related potential) signal.On the frequency domain, change by its mean power of frequency computation part in same incident when then the power spectrum of whole EEG (brain) data being locked, obtain two dimension (time-frequency) image that is referred to as relevant perturbation of spectra of incident or disturbance (ERSP).

But no matter be the neither characteristic that can comprehensively provide the relevant cognitive brain electricity of generation incident of the average gained ERSP of time domain average gained ERP or frequency domain.Because ERP through the superposed average filtering can not be in the observation brain electricity but still induce relevant a large amount of useful informations with cognitive incident with stimulating periods of events accurately to repeat, when only staying minority and bringing out relevant lock and lock the phase partial information with cognitive incident.And among the ERSP by cognitive incident induce the characteristic wave bands that changes of the power that produces or energy oscillation almost be difficult in to bring out and find corresponding characteristic component among the ERP that is produced by same incident.Vice versa.Therefore, even ERP and ERSP are combined, also can't intactly extract the whole characteristics in the cognitive brain electricity comprehensively.

The inventor finds to exist at least in the prior art following shortcoming and defect in realizing process of the present invention:

But remain in some shortcomings at present, thereby reduced the separability of the brain electrical feature that causes owing to individual difference, can better not carry out identification on the signal decomposition of cognitive EEG signals and the feature extracting method.

Summary of the invention

The invention provides a kind of personal identification method based on the relevant brain electricity of the cognitive incident of audiovisual, this method has strengthened the separability of the brain electrical feature that is caused by individual difference, makes to carry out identification better, sees hereinafter for details and describes:

A kind of personal identification method based on the relevant brain electricity of the cognitive incident of audiovisual said method comprising the steps of:

(1) adopt expansion maximum informational entropy independent component analysis method to carry out the filtering of evoked brain potential time frequency space; With the expansion maximum informational entropy is object function; Go back to primary time domain distribution of electrodes space through the evoked brain potential signal being separated mixed matrix back projection, obtain a series of independent element components; Employing is carried out space filtering based on wavelet packet zone independent component analysis method to said a series of independent element components;

(2) adopt the WAVELET PACKET DECOMPOSITION reconstructing method to obtain the EEG signals that δ, θ, α, β and γ wave band are the cortex brain electrical acti source rhythm and pace of moving things scope of representative to filtered a series of independent element components; Each rhythm and pace of moving things EEG signals is carried out empirical modal decompose, obtain the characteristic modal components of each self-corresponding reflection characteristic rhythm and pace of moving things vibration mode;

(3) the inherent index that is concerned with of employing incident, the inherent relevant index of phase place, the partly directly relevant index of the relevant exponential sum in phase shift inherence extract respectively and lock synchronous characteristic between phase strength characteristic, incident between the cognitive incident in the said characteristic modal components, drive and respond coupling feature; The relevant perturbation of spectrum of employing incident extracts the power spectrum characteristic of the relevant brain electricity of cognitive incident;

(4) adopt the SVMs multi-categorizer that the power spectrum characteristic of locking synchronous characteristic between phase strength characteristic, said incident, said driving and response coupling feature and said cognitive incident between a plurality of experimenters' said cognitive incident is classified, obtain the identification result.

The beneficial effect of technical scheme provided by the invention is: adopting independent component analysis is that a series of locus are clear and definite with empirical mode decomposition method with spontaneous EEG or the relevant EEG signal decomposition of incident, has the characteristic component sum that function connects meaning; In the time of can realizing the dynamic monitoring signal lock through the inherent relevant index of relevant perturbation of spectrum of incident and incident, lock phase character; Use simultaneously that phase place is inherent relevant, the inherent relevant and directly relevant index of part of phase shift can more deeply detect information change and the transformational relation of cortex brain electrical acti source before and after cognitive incident, from multi-angular analysis and extract cognitive brain electrical feature; To the signal detection and the processing of event related potential, carry out above-mentioned characteristic parameter extraction, obtained very good effect; Reduced the influence of interfering signal to EEG signals; EEG signals are used for identification, and then have improved the identification rate, satisfied the needs in the practical application.

Description of drawings

Fig. 1 is provided by the invention based on the sketch map of the relevant brain electrical feature extraction of the cognitive incident of audiovisual with pattern recognition analysis overall plan;

Fig. 2 a is the sketch map of time-frequency characteristics before the ICA filtering provided by the invention;

Fig. 2 b is the sketch map of time-frequency characteristics after the ICA filtering provided by the invention;

Fig. 2 c is the sketch map of signal power spectrum density time domain variation characteristic provided by the invention;

Fig. 2 d is ERSP value provided by the invention and the ERP signal that contains P300 sketch map relatively;

Fig. 2 e is ITC value provided by the invention and the ERP signal that contains P300 sketch map relatively;

Fig. 2 f is ERSP value provided by the invention and the ERP signal that contains P300 sketch map relatively;

Fig. 2 g is ITC value provided by the invention and the ERP signal that contains P300 sketch map relatively;

Fig. 3 is a kind of flow chart based on the electric personal identification method of the relevant brain of the cognitive incident of audiovisual provided by the invention.

The specific embodiment

For making the object of the invention, technical scheme and advantage clearer, embodiment of the present invention is done to describe in detail further below in conjunction with accompanying drawing.

The relevant brain electricity of cognitive incident and sensation, cognitive activities are closely related, are the combined reactions of stimulus to the sense organ and inherent cognitive thinking to external world of each functional areas of perception nervous pathway and cerebral cortex.The cognitive incident EEG signals that audio-visual combined stimulation obtains with respect to unipath vision or auditory stimulus, wave-shape amplitude is higher, incubation period is more stable and space-time characteristic is obvious.

Owing to bring out the mixed signal of the multiple signals that the cognitive EEG signals that obtain produce by homology not often; Therefore and EEG signals generally have non-stationary and time variation, more are suitable for to adopt independent component analysis (ICA) and empirical modal to decompose (EMD) method scalp brain electricity (spontaneous and cognitive EEG) to be decomposed into a series of locus are definite, relatively independent component sum of time.

The introducing incident is correlated with that inherent relevant (ITC) index of perturbation of spectrum (ERSP), cognitive incident, phase place inherent relevant (PIC), inherent relevant (PsIC) parameter of phase shift and directly relevant (PDC) various features parameter index of part detect when locking between the important frequencies component in cortex brain electrical acti source, lock phase, phase place replacement and synchronized oscillation characteristic; Thereby strengthen the separability of brain electricity biological characteristic, carry out identification better.

In order to strengthen the separability of the brain electrical feature that causes by individual difference, make and can carry out identification better that referring to Fig. 1 and Fig. 3, the embodiment of the invention provides a kind of personal identification method based on the relevant brain electricity of the cognitive incident of audiovisual, sees hereinafter for details and describes:

Independent component analysis (ICA) is meant the technology of from the linear hybrid signal of multiple source signals, isolating source signal.Except known source signal is statistics independence, there are not other prioris, ICA provides a kind of effective blind source to separate (BSS) method.ICA the details of other signal is not almost destroyed, and denoising performance is also often well a lot of than traditional filtering method when eliminating noise.Empirical modal decomposes (EMD) method and is suitable for analyzing non-linear and the non-stationary signal sequence, has very high signal to noise ratio.The key of this method is an empirical mode decomposition, and it can make sophisticated signal be decomposed into limited intrinsic mode functions (IMF), and each the IMF component that decomposes has out comprised the local feature signal of the different time yardstick of original signal.Empirical mode decomposition can make the non-stationary data carry out tranquilization and handle, and carries out Hilbert transform then and obtains time-frequency spectrum, obtains the frequency of physical significance.Compare with methods such as wavelet decomposition with short time discrete Fourier transform, this method be intuitively, direct, posterior and adaptive because basic function is to be decomposed by data itself to obtain.

Owing to bring out the mixed signal of the multiple signals that the cognitive EEG signals that obtain produce by homology not often; Therefore and EEG signals generally have non-stationary and time variation, adopt independent component analysis (ICA) and empirical modal to decompose (EMD) method and scalp brain electricity (spontaneous and cognitive EEG) are decomposed into a series of locus are definite, relatively independent component sum of time.

101: adopt expansion maximum informational entropy independent component analysis method (ICA) to carry out the filtering of evoked brain potential time frequency space; With the expansion maximum informational entropy is object function; Go back to primary time domain distribution of electrodes space through the evoked brain potential signal being separated mixed matrix back projection, obtain a series of independent element components; Employing is carried out space filtering based on wavelet packet zone independent component analysis (WPICA) method to a series of independent element components;

The wavelet packet character subspace is chosen in feature band distribution according to EEG signals, respectively a series of independent element components is launched isolated components and decomposes, and is decomposed into a series of height frequency various signals; Through in the isolated component catabolic process, embedding wavelet packet multiresolution analysis method a series of independent element components are carried out the Wavelet Packet Domain uncoiling; Again through the wavelet packet inverse transformation, domain space when original brain electricity is returned in the brain power supply signal projection of feature band with the interfering signal filtering, obtains more purified a series of independent element component from the angle of frequency.

Wherein, it is conventionally known to one of skill in the art adopting expansion maximum informational entropy independent component analysis method (ICA) to carry out the concrete steps of evoked brain potential space filtering and separating mixed matrix, and the embodiment of the invention does not limit at this.

102: adopt the WAVELET PACKET DECOMPOSITION reconstructing method to obtain the EEG signals that δ, θ, α, β and γ wave band are the cortex brain electrical acti source rhythm and pace of moving things scope of representative to filtered a series of independent element components; Each rhythm and pace of moving things EEG signals is carried out EMD decompose, obtain the characteristic modal components of each self-corresponding reflection characteristic rhythm and pace of moving things vibration mode;

Relevant perturbation of spectrum (ERSP) characteristic of incident with the energy envelope CURVE STUDY EEG signals of the characteristic modal components of match in the algorithmic procedure.

Wherein, directly carry out EMD for fear of original eeg data and decompose the easy false mode function problem that produces, adopt the WAVELET PACKET DECOMPOSITION reconstructing method to obtain the EEG signals that δ, θ, α, β and γ wave band are the cortex brain electrical acti source rhythm and pace of moving things scope of representative.Relevant perturbation of spectrum (ERSP) characteristic of incident through energy envelope CURVE STUDY EEG signals.

103: inherent relevant (ITC) index of employing incident, inherent relevant (PIC) index of phase place, directly relevant (PDC) index of inherent relevant (PsIC) exponential sum part of phase shift extract synchronous characteristic between lock phase strength characteristic in the characteristic modal components, incident respectively, drive and respond coupling feature; The relevant perturbation of spectrum ERSP of employing incident extracts the power spectrum characteristic of the relevant brain electricity of cognitive incident;

Wherein, for n cognitive incident, ERSP is defined as

$\mathrm{ERSP}(f,t)=\frac{1}{n}\mathrm{\Σ}\left|F_k(f,t)\right|---\left(1\right)$

F in the formula _k(f, (t)-frequently (f) distributes when being the cognitive brain electricity of the k time incident t).ERSP has reflected the influence of cognitive process to each frequency band power spectrum.

Wherein, ITC then is used for detecting the degree that lock brings out mutually takes place in this cognitive process, is defined as

$\mathrm{ITC}(f,t)=\frac{1}{n}\left|\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\frac{F_k(f,t)}{|F_k(f,t)}\right|---\left(2\right)$

The ITC value is between 0 and 1, and ITC=1 representes that the strictness of cognitive brain electricity is locked in the stimulation initial state of cognitive incident.

Change though ERSP can reflect the power spectral density in the cognitive process, ITC can detect lock phase degree, does not provide synchronized oscillation information yet.For estimating the synchronized oscillation characteristic between the cortex brain electrical acti source, introduce phase place inherent relevant (PIC) and measure, be defined as

$C_{\mathrm{PIC}}(f,t)=\left|\underset{k}{\mathrm{\Σ}}{F}_k(f,t)\right|/\underset{k}{\mathrm{\Σ}}\left|F_k(f,t)\right|\≤1---\left(3\right)$

C _PIC(f, t) value is also between 0 and 1.C _PIC(f, t)=1 the cognitive brain electricity of expression is locked phase fully.And for the non-lock phase behavior of signal, also can introduce inherent relevant (PsIC) parameter of phase shift to measure, be defined as

$C_{\mathrm{PsIC}}(f,t)=\underset{k}{\mathrm{\Σ}}{\left|F_k(f,t)\right|}^2/\underset{f,t}{\mathrm{Max}}\underset{k}{\mathrm{\Σ}}|F_k(f,t){|}^2\≤1---\left(4\right)$

C _PsIC(f, t) value is still between 0 and 1.C _PsIC(f, t)=the 1 complete non-lock phase of the cognitive brain electricity of expression.

For further detecting the kinetics coupled relation between cognitive each frequency range component of brain electricity, introduce directly relevant (PDC) index of part, be defined as

In the formula A _{I, j}(f t) is i, j frequency range cross-spectral density coefficient.D ^{I ← j} _PDCExpression j component is to the directly relevant effect of i component.D _PDCValue between 0 and 1, D _PDC=1 expression maximum coherence.D _PDCDescribe the linear correlation that two time series signals are asked from frequency domain, connect degree to estimate its function, drive and the response coupled relation.

104: adopt the SVMs multi-categorizer that the power spectrum characteristic of locking synchronous characteristic between phase strength characteristic, incident between a plurality of experimenters' cognitive incident, drive and respond coupling feature and cognitive incident is classified, obtain the identification result.

Verify a kind of personal identification method that the embodiment of the invention provides with a concrete experiment below, see hereinafter for details and describe based on the relevant brain electricity of the cognitive incident of audiovisual:

Referring to Fig. 2 a, Fig. 2 b, Fig. 2 c, Fig. 2 d, Fig. 2 e, Fig. 2 f and Fig. 2 g, time-frequency filtering: (a) time-frequency characteristics is not obvious before the ICA filtering; (b) after the filtering, the energy rising phenomenon of corresponding band is obvious, appears time-frequency characteristics suddenly.Space filtering: A ERP time-frequency characteristics behind wavelet packet independent component analysis (WPICA) space filtering is outstanding, and spatial distribution differences is big; The corresponding signal power spectrum density of B time domain variation characteristic is obvious.In the time of can be through inherent relevant (ITC) index dynamic monitoring signal lock of relevant perturbation of spectrum (ERSP) of incident and incident, lock phase character.P300 is the typical endogenous component of ERP signal; Only extract the ERP signal by superposed average; The lock phase character detect the time-the frequency analysis example; ERSP (dB), ITC value and the ERP signal that contains P300 be relatively: the P300 signal goes out the corresponding ERSP of having and the peak value of ITC now, when the indication induced response is accompanied by lock, lock the phase effect; After the pretreatment of ICA space filtering, this effect is more outstanding.

Make that through above-mentioned processing the characteristic of extracting is better, reduced the influence of interfering signal, EEG signals are used for identification, and then have improved the identification rate, satisfied the needs in the practical application EEG signals.

In sum; The embodiment of the invention provides a kind of personal identification method based on the relevant brain electricity of the cognitive incident of audiovisual; It is that a series of locus are clear and definite with empirical mode decomposition method with spontaneous EEG or the relevant EEG signal decomposition of incident that this method adopts independent component analysis, has the characteristic component sum that function connects meaning; In the time of can realizing the dynamic monitoring signal lock through the inherent relevant index of relevant perturbation of spectrum of incident and incident, lock phase character; Use simultaneously that phase place is inherent relevant, the inherent relevant and directly relevant index of part of phase shift can more deeply detect information change and the transformational relation of cortex brain electrical acti source before and after cognitive incident, from multi-angular analysis and extract cognitive brain electrical feature; Signal detection and processing to event related potential; Reduced the influence of interfering signal, carried out above-mentioned characteristic parameter extraction, obtained very good effect EEG signals; And then improved the recognition correct rate that EEG signals is used for identification, satisfied the needs in the practical application.

It will be appreciated by those skilled in the art that accompanying drawing is the sketch map of a preferred embodiment, the invention described above embodiment sequence number is not represented the quality of embodiment just to description.

The above is merely preferred embodiment of the present invention, and is in order to restriction the present invention, not all within spirit of the present invention and principle, any modification of being done, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (1)

1. the personal identification method based on the relevant brain electricity of the cognitive incident of audiovisual is characterized in that, said method comprising the steps of:

(1) adopt expansion maximum informational entropy independent component analysis method to carry out the filtering of evoked brain potential time frequency space; With the expansion maximum informational entropy is object function; Go back to primary time domain distribution of electrodes space through the evoked brain potential signal being separated mixed matrix back projection, obtain a series of independent element components; Employing is carried out space filtering based on wavelet packet zone independent component analysis method to said a series of independent element components;

(2) adopt the WAVELET PACKET DECOMPOSITION reconstructing method to obtain the EEG signals that δ, θ, α, β and γ wave band are the cortex brain electrical acti source rhythm and pace of moving things scope of representative to filtered a series of independent element components; Each rhythm and pace of moving things EEG signals is carried out empirical modal decompose, obtain the characteristic modal components of each self-corresponding reflection characteristic rhythm and pace of moving things vibration mode;

(3) the inherent index that is concerned with of employing incident, the inherent relevant index of phase place, the partly directly relevant index of the relevant exponential sum in phase shift inherence extract respectively and lock synchronous characteristic between phase strength characteristic, incident between the cognitive incident in the said characteristic modal components, drive and respond coupling feature; The relevant perturbation of spectrum of employing incident extracts the power spectrum characteristic of the relevant brain electricity of cognitive incident;

(4) adopt the SVMs multi-categorizer that the power spectrum characteristic of locking synchronous characteristic between phase strength characteristic, said incident, said driving and response coupling feature and said cognitive incident between a plurality of experimenters' said cognitive incident is classified, obtain the identification result.

Images