CN102063180B - HHT-based high-frequency combined coding steady state visual evoked potential brain-computer interface method - Google Patents

Abstract

The invention discloses a Hilbert-Huang Transform (HHT)-based high-frequency combined coding steady state visual evoked potential brain-computer interface method, which comprises the following steps of: connecting hardware, expressing different targets by using nn stimulation sequences formed by time sequence permutation and combination through n high-frequency stimulation frequencies, coding the stimulation sequences to form a code base, then executing electroencephalogram data characteristics by adopting a Hilbert-Huang Transform (HHT)-based variable frequency electroencephalogram signal characteristic extraction method, acquiring a Hilbert time-frequency graph of the electroencephalogram data, and comparing the extracted electroencephalogram data characteristics with the code base by using a local frequency spectrum extreme value target identification method to realize quantified target identification accuracy. The method has the advantages of simple operation, little electrode number, more target number, reduction of testee fatigue, and reduction of probability of inducing testee epilepsy.

Description

Based on HHT high frequency assembly coding stable state vision inducting current potential brain-machine interface method

Technical field

(Brain-Computer Interface, BCI) technical field are specifically related to based on HHT high frequency assembly coding stable state vision inducting current potential brain-machine interface method to the present invention relates to brain-computer interface.

Background technology

Brain-computer interface is based on EEG signals and realizes that brain and computing machine or other electronic equipments directly exchange the system of communication and control.Brain-computer interface (BCI) is as man machine interface (Human-computer interface; Be called for short HCI) in a kind of; Owing to do not rely on conventional brain output channel; The brain of behaving has been opened up a brand-new approach that carries out information interchange and control with the external world, and the feasible idea of utilizing people's brain signal directly to control external unit becomes possibility.In recent years, brain-computer interface (BCI) technical development is very fast, embodies significant values in fields such as biomedicine, virtual reality, Entertainment, rehabilitation project and space flight, military affairs.

Stable state vision inducting electricity (Steady State Visually Evoked Potential; SSVEP) be that the brain vision system continues the periodically response of visual stimulus to the outside; It is input signal commonly used in the brain-computer interface BCI system; Compare signals such as P300, incident related synchronization, spontaneous brain electricity; Stable state vision inducting current potential (SSVEP) is the research normal form that has wide application prospect and using value in the BCI system owing to have advantages such as simple to operate, that recording electrode is few, the training time is short, rate of information transmission is high and antijamming capability is strong.

But at present general SSVEP-BCI system convention adopts the low-frequency range of 6～25Hz, basically all adopts the simple normal form of a task with a frequency representation; And because the restriction of frequency resolution and screen refresh rate makes that in limited frequency band range the task object number that can appear is limited; In addition, in practical application, the low frequency flicker of low-frequency range SSVEP stimulates makes the experimenter produce visual fatigue easily, even has the possibility of bringing out experimenter's epileptic attack, the long-time comfortableness of using of the system that therefore can't guarantee.

High frequency SSVEP is near the optical drive threshold frequency of human eye; General frequency range can reach 30～60Hz or higher; Different with low frequency is, and high frequency stimulation can produce flicker fusion (flicker fusion) effect, the subjective imperceptible flicker of user, but in the brain electricity, still can detect the HF-response of SSVEP; So just greatly reduce visual fatigue and the possibility of bringing out epilepsy, realized harmless; Simultaneously, though high frequency SSVEP amplitude is lower, the amplitude of background brain electricity is also very low in this frequency range, so high frequency SSVEP still has certain signal to noise ratio (S/N ratio), has ensured the basic identification accuracy rate of stimuli responsive signal.

(Hilbert-Huang Transform HHT) is a kind of adaptive signal processing method to Hilbert-Huang transform, is fit to very much dealing with nonlinear, non-stationary signal.This method consists two processes: empirical mode decomposition (empirical mode decomposition, EMD) with the Hilbert conversion (Hilbert Transform, HT); Wherein the part of most critical is the EMD method; The EMD method can be decomposed into eigenmodes function (Intrinsic Mode Functions, IMF) sum to sophisticated signal based on the local feature time scale of signal; The frequency content that each IMF comprised is not only relevant with analysis frequency; And the most important thing is to change with signal itself, therefore, the HHT method is fit to handle non-stationary, nonlinear EEG signals very much.

Summary of the invention

In order to overcome the shortcoming of above-mentioned prior art; The invention provides based on HHT high frequency assembly coding stable state vision inducting current potential brain-machine interface method; The normal form based on high frequency sequential assembly coding SSVEP has been proposed on the simple stimulus modality of traditional SSVEP basis; And the temporal aspect proposition that is directed against frequency conversion assembly coding signal has advantage simple to operate, that electrode number is few and number of targets is many based on the HHT feature extracting method.

In order to achieve the above object, the technical scheme taked of the present invention is:

Based on HHT high frequency assembly coding stable state vision inducting current potential brain-machine interface method, may further comprise the steps:

Step 1; The Oz position of sound production potential electrode A in the occipital region of subject's head D; At the one-sided ear-lobe position of sound production reference electrode B of subject's head D, at the frons Fpz of subject's head D position of sound production ground electrode C, the output terminal of potential electrode A inserts the first input end E1 of eeg amplifier E; The output terminal of reference electrode B inserts the second input end E2 of eeg amplifier E; The output terminal of ground electrode C inserts the 3rd input end E3 of eeg amplifier E, and the output terminal of eeg amplifier E links to each other with the input end of computing machine F, and the output terminal of computing machine F is connected with computer screen G;

Step 2, at SSVEP response high band, promptly flicker frequency is higher than 25Hz, selects n frequency of stimulation, and n is the integer greater than 1, with the n of n frequency of stimulation through the formation of sequential permutation and combination ⁿIndividual stimulus sequence is presented in face of the experimenter through computer screen G, and the stimulus sequence coding is formed code database;

Step 3; When the experimenter watches the high frequency sequential assembly coding stimulus sequence of step 2 attentively; The eeg signal acquisition system that connects through step 1 obtains EEG signals; And adopt the feature extraction that realizes EEG signals based on the frequency conversion EEG feature extraction method of Hilbert-Huang transform (HHT), concrete steps are following:

◆ the eeg data to collecting carries out superposed average, improves signal to noise ratio (S/N ratio), obtains pretreated eeg data;

◆ utilize bandpass filtering that pretreated eeg data is carried out bandpass filtering, obtain filtered eeg data;

◆ adopt the EMD method that filtered eeg data is decomposed;

When using the EMD method, to the end points problem, adopt the Boundary Prediction method, the extreme value computing formula that the Boundary Prediction method is confirmed is following:

Extreme value=MAX{| endpoint value |, | the first approximation point |

To the stopping criterion problem, adopt fixing screening number of times, the EMD algorithm is the process of a screening, the mathematical formulae of screening is following:

$x\left(t\right)=\underset{1}{\overset{n}{\mathrm{\Σ}}}{c}_j=x\left(t\right)-\underset{1}{\overset{k}{\mathrm{\Σ}}}{m}_j+(\underset{1}{\overset{k}{\mathrm{\Σ}}}{m}_j-\underset{1}{\overset{p}{\mathrm{\Σ}}}{m}_{2j})+....$

Wherein: x (t) is a signal to be decomposed;

c _jBe the IMF signal after the EMD decomposition;

N is that EMD decomposes back IMF number;

m _j, m _2jThe mean value curve that calculates through last lower envelope for screening process;

K, p is the screening number of times;

For the IMF that guarantees to filter out has enough physical significances on amplitude and frequency, the ratio that promptly guarantees adjacent IMF yardstick is near 2, and to the number of times k of screening process, p must limit, specifically according to the EEG signals characteristics 2 ³-2 ⁴Between optimized choice, promptly select k=p=l, wherein

The ratio of the adjacent IMF yardstick of l=is 2 o'clock a screening number of times;

◆ according to the EMD decomposition result, the IMF reconstruct data of select target frequency place frequency range;

◆ use GZC (Generalized Zero-Crossing) method that the IMF reconstruct data is calculated its instantaneous frequency f _Gzc

Instantaneous frequency f _GzcComputing formula following:

$f_{\mathrm{gzc}}=\frac{4{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="HSA00000328444600012.png,HSA00000328444600031.png"\; state="\{\{state\}\}">f</figure-callout>}}_1+2({\mathrm{<figure-callout\; id="21"\; label="f"\; filenames="HSA00000328444600031.png,HSA00000328444600061.png"\; state="\{\{state\}\}">f</figure-callout>}}_{21}+f_{22})+f_{41}+f_{42}+f_{43}+f_{44}}{12};$

Wherein: ${\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="HSA00000328444600012.png,HSA00000328444600031.png"\; state="\{\{state\}\}">f</figure-callout>}}_1=\frac{1}{{4\mathrm{<figure-callout\; id="1"\; label="T"\; filenames="HSA00000328444600012.png,HSA00000328444600031.png"\; state="\{\{state\}\}">T</figure-callout>}}_1};$

$f_{2i}=\frac{1}{{2T}_{2i}},i=\mathrm{1,2};$

$f_{4i}=\frac{1}{T_{4i}},i=\mathrm{1,2,3,4},$

Wherein T is the time period between zero point and the peak value;

◆ according to instantaneous frequency f _Gzc, and then when obtaining the Hilbert of eeg data-frequency figure;

Step 4, the relevant IMF reconstruct data that step 3 is obtained calculates its FFT frequency spectrum respectively according to 3 sections of total duration average marks; And then calculate the frequency spectrum extreme value; With the frequency spectrum extreme value that obtains and pre-set threshold relatively and coding, with the corresponding encoding ratio in the code database of formation in coding that obtains and the step 2, if in code database; Mark non-zero then; Otherwise be labeled as zero, wherein non-zero number is the correct number of times of target identification, realizes that finally local frequency spectrum extreme value target identification method quantizes discrimination.

Owing to the present invention is directed to the simple shortcomings of bringing with deficiency such as SSVEP low-frequency range such as making experimenter's fatigue easily and bring out that epilepsy, target numbers are lacked of the normal form of existing SSVEP-BCI system; The thought of coding is introduced on system ensemble ground; Utilize the harmless characteristics of high frequency SSVEP and potential advantage mechanism; High frequency sequential assembly coding stable state vision inducting new normal form of current potential brain-computer interface and frequency conversion brain electrical feature method for distilling thereof are proposed; Solve unifrequency and present the limited problem of destination number, after available n basic frequency unit confirmed, make denotable target numbers be increased to n by n ⁿIndividual, so increased the optional number of targets of SSVEP; Utilize the stimulation flicker of high frequency SSVEP to merge effect, improved user's comfort level, reduced the possibility of bringing out tired and epilepsy, so ensured the realization of the efficient lossless function of high frequency SSVEP brain-computer interface system.In view of the above plurality of advantages of high frequency SSVEP, the deficiency of aiming low frequency SSVEP in practical application.The present invention proposes the new normal form of BCI of high frequency sequential assembly coding SSVEP, and through the thinking of a plurality of stable state high frequency sequential permutation and combination realization stimulus codings, n frequency can form n ⁿIndividual target solves unifrequency and presents the limited problem of destination number, compares n frequency and can form n ⁿThe simple normal form of individual target has been expanded number of targets and range of choice greatly; Simultaneously, frequency conversion assembly coding signal has stronger antijamming capability, and its temporal aspect has also increased the echo signal information except that frequency, lays a good foundation with the shortening identification time for further improving the identification accuracy.Obviously, appeared in the number of targets that significantly increases and the identification efficient and the accuracy rate that further promote can improve the communication transfer rate of man-machine interaction greatly; Add the harmless advantage of high frequency SSVEP self; Brain-computer interface technology based on high frequency sequential assembly coding SSVEP can improve the comfortableness and the feasibility of interaction capabilities, reliability and the use of existing BCI technology greatly; Can be applicable to various fields such as aerospace, military affairs, have important academic theoretical research and actual application value.

Description of drawings

Fig. 1 is that hardware of the present invention connects synoptic diagram.

Fig. 2 is a stimulus sequence synoptic diagram of the present invention.

Fig. 3 is the goal stimulus sequential chart and the frequency plot thereof of the embodiment of the invention, and wherein Fig. 3 (a) is the visual stimulus sequence of 27 targets among the embodiment, and Fig. 3 (b) is the frequency of the corresponding stimulus sequence of Fig. 3 (a).

Fig. 4 is time domain waveform and the FFT spectrogram behind 27 target EEG signals superposed averages among the embodiment, and wherein Fig. 4 (a) is the SSVEP response of 27 targets, and Fig. 4 (b) is the FFT frequency spectrum of SSVEP response.

Fig. 5 is based on the process flow diagram of the high frequency sequential assembly coding SSVEP EEG feature extraction method of HHT.

Fig. 6 is a Boundary Prediction method schematic diagram.

Fig. 7 is based on the synoptic diagram that the GZC method is calculated instantaneous frequency.

When Fig. 8 is the Hilbert of 27 goal stimulus sequences and high frequency sequential assembly coding SSVEP response thereof among the embodiment-and frequency figure, wherein: when Fig. 8 (a) is stimulus sequence Hilbert-frequency figure, Fig. 8 (b) is that SSVEP when responding Hilbert-frequency is schemed.

Embodiment

Below in conjunction with accompanying drawing the present invention is done further detailed description.

Based on HHT high frequency assembly coding stable state vision inducting current potential brain-machine interface method, may further comprise the steps:

The first step; With reference to Fig. 1, the Oz position of sound production potential electrode A in the occipital region of subject's head D is at the one-sided ear-lobe position of sound production reference electrode B of subject's head D; At the frons Fpz of subject's head D position of sound production ground electrode C; The output terminal of potential electrode A inserts the first input end E1 of eeg amplifier E, and the output terminal of reference electrode B inserts the second input end E2 of eeg amplifier E, and the output terminal of ground electrode C inserts the 3rd input end E3 of eeg amplifier E; The output terminal of eeg amplifier E links to each other with the input end of computing machine F, and the output terminal of computing machine F is connected with computer screen G;

In second step, with reference to Fig. 2, at SSVEP response high band, promptly flicker frequency is higher than 25Hz, selects n frequency of stimulation, and n is the integer greater than 1, with the n of n frequency of stimulation through the formation of sequential permutation and combination ⁿIndividual stimulus sequence is presented in face of the experimenter through computer screen G, and the stimulus sequence coding is formed code database;

The 3rd step; With reference to Fig. 5; When the experimenter watches the high frequency sequential assembly coding stimulus sequence in second step attentively; The eeg signal acquisition system that connects through the first step obtains EEG signals, and adopts the feature extraction that realizes EEG signals based on the frequency conversion EEG feature extraction method of Hilbert-Huang transform (HHT), and concrete steps are following:

◆ the eeg data to collecting carries out superposed average, improves signal to noise ratio (S/N ratio), obtains pretreated eeg data;

◆ utilize bandpass filtering that pretreated eeg data is carried out bandpass filtering, obtain filtered eeg data;

◆ adopt the EMD method that filtered eeg data is decomposed;

When the EMD method is used,, adopt the Boundary Prediction method to the end points problem. as shown in Figure 6, the extreme value computing formula that the Boundary Prediction method is confirmed is following:

Extreme value=MAX{| endpoint value |, | the first approximation point |;

To the stopping criterion problem, adopt fixing screening number of times, the EMD algorithm is the process of a screening, the mathematical formulae of describing screening process is following:

$x\left(t\right)=\underset{1}{\overset{n}{\mathrm{\&Sigma;}}}{c}_j=x\left(t\right)-\underset{1}{\overset{k}{\mathrm{\&Sigma;}}}{m}_j+(\underset{1}{\overset{k}{\mathrm{\&Sigma;}}}{m}_j-\underset{1}{\overset{p}{\mathrm{\&Sigma;}}}{m}_{2j})+....$

Wherein: x (t) is a signal to be decomposed;

c _jBe the IMF signal after the EMD decomposition;

N is that EMD decomposes back IMF number;

m _j, m _2jThe mean value curve that calculates through last lower envelope for screening process;

K, p is the screening number of times;

For the IMF that guarantees to filter out has enough physical significances on amplitude and frequency, the ratio that promptly guarantees adjacent IMF yardstick is near 2, and to the number of times k of screening process, p must limit, specifically according to the EEG signals characteristics 2 ³-2 ⁴Between optimized choice, promptly select k=p=l, wherein

The ratio of the adjacent IMF yardstick of l=is 2 o'clock a screening number of times;

◆ according to the EMD decomposition result, the IMF reconstruct data of select target frequency place frequency range;

◆ use GZC (Generalized Zero-Crossing) method that the IMF reconstruct data is calculated its instantaneous frequency f _Gzc, as shown in Figure 7;

Instantaneous frequency f _GzcComputing formula following:

$f_{\mathrm{gzc}}=\frac{{4\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="HSA00000328444600012.png,HSA00000328444600031.png"\; state="\{\{state\}\}">f</figure-callout>}}_1+2({\mathrm{<figure-callout\; id="21"\; label="f"\; filenames="HSA00000328444600031.png,HSA00000328444600061.png"\; state="\{\{state\}\}">f</figure-callout>}}_{21}+f_{22})+f_{41}+f_{42}+f_{43}+f_{44}}{12};$

Wherein: ${\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="HSA00000328444600012.png,HSA00000328444600031.png"\; state="\{\{state\}\}">f</figure-callout>}}_1=\frac{1}{{4\mathrm{<figure-callout\; id="1"\; label="T"\; filenames="HSA00000328444600012.png,HSA00000328444600031.png"\; state="\{\{state\}\}">T</figure-callout>}}_1};$

$f_{2i}=\frac{1}{{2T}_{2i}},i=\mathrm{1,2};$

$f_{4i}=\frac{1}{T_{4i}},i=\mathrm{1,2,3,4};$

Wherein T is the time period between zero point and the peak value;

◆ according to instantaneous frequency f _Gzc, and then when obtaining the Hilbert of eeg data-frequency figure, as shown in Figure 8;

In the 4th step, the relevant IMF reconstruct data that the 3rd step was obtained calculates its FFT frequency spectrum respectively according to 3 sections of total duration average marks; And then calculate the frequency spectrum extreme value, with the frequency spectrum extreme value that obtains and pre-set threshold relatively and coding, the corresponding encoding ratio during the coding that obtains gone on foot with second in the code database of formation is; If in code database; Mark non-zero then, otherwise be labeled as zero, wherein non-zero number is the correct number of times of target identification; The local frequency spectrum extreme value target identification method of final realization quantizes discrimination, and identification accuracy computing formula is following:

Below in conjunction with specific embodiment the present invention is done further detailed description.

Based on HHT high frequency assembly coding stable state vision inducting current potential brain-machine interface method, may further comprise the steps:

The first step; The Oz position of sound production potential electrode A in the occipital region of subject's head D; At the one-sided ear-lobe position of sound production reference electrode B of subject's head D, at the frons Fpz of subject's head D position of sound production ground electrode C, the output terminal of potential electrode A inserts the first input end E1 of eeg amplifier E; The output terminal of reference electrode B inserts the second input end E2 of eeg amplifier E; The output terminal of ground electrode C inserts the 3rd input end E3 of eeg amplifier E, and the output terminal of eeg amplifier E links to each other with the input end of computing machine F, and the output terminal of computing machine F is connected with computer screen G;

In second step, with reference to Fig. 2 and Fig. 3, at SSVEP response high band, promptly flicker frequency is higher than 25Hz, and n=3 the frequency of selection 25,33.33 and 40Hz numbers 1,2,3 respectively as basic frequency of stimulation unit, through the sequential permutation and combination, can form n ⁿ=3 ³=27 time serieses of representing different target are respectively

111、112、113、121、122、123131、132、133

211、212、213、221、222、223、231、232、233

311、312、313、321、322、323、331、332、333

The visual stimulus sequence of 27 targets is as shown in Figure 3; For example; 132 represent the stimulus sequence of the time series composition of 25Hz, 40Hz and 33.33Hz, and as shown in Figure 2, each frequency adopts 20 trigger action; Each sequential stimulus sequence time span does not wait at 1.5s and 2.4s, stimulates according to the above representative different target sequential that designs through computer screen G and repeats respectively to appear;

The 3rd step; The experimenter is sitting in from about 70 centimetres position, computing machine CRT screen G dead ahead, watches described different high frequency sequential assembly coding stimulus sequence targets of second step attentively, and the eeg signal acquisition system that connects through the first step obtains the non-linear EEG signals of corresponding typical frequency conversion luffing; Traditional spectral method is difficult to effectively extract its characteristic; Shown in Fig. 4 (b), for this reason, adopt the feature extraction that realizes the frequency conversion EEG signals based on Hilbert-Huang transform (HHT); As shown in Figure 5, specifically may further comprise the steps:

◆ the eeg data to collecting carries out superposed average, improves signal to noise ratio (S/N ratio), obtains pretreated eeg data;

◆ utilize bandpass filtering that pretreated eeg data is carried out bandpass filtering, obtain filtered eeg data;

◆ adopt the EMD method that filtered eeg data is decomposed;

When the EMD method is used,, adopt the Boundary Prediction method to the end points problem. as shown in Figure 6, the extreme value computing formula that the Boundary Prediction method is confirmed is following:

Extreme value=MAX{| endpoint value |, | the first approximation point |;

To the stopping criterion problem, adopt fixing screening number of times, the EMD algorithm is the process of a screening, the mathematical formulae of screening is following:

$x\left(t\right)=\underset{1}{\overset{n}{\mathrm{\&Sigma;}}}{c}_j=x\left(t\right)-\underset{1}{\overset{k}{\mathrm{\&Sigma;}}}{m}_j+(\underset{1}{\overset{k}{\mathrm{\&Sigma;}}}{m}_j-\underset{1}{\overset{p}{\mathrm{\&Sigma;}}}{m}_{2j})+....$

Wherein: x (t) is a signal to be decomposed;

c _jBe the IMF signal after the EMD decomposition;

N is that EMD decomposes back IMF number;

m _j, m _2jThe mean value curve that calculates through last lower envelope for screening process;

K, p is the screening number of times;

For the IMF that guarantees to filter out has enough physical significances on amplitude and frequency, the ratio that promptly guarantees adjacent IMF yardstick is near 2, and to the number of times k of screening process, p must limit, specifically according to the EEG signals characteristics 2 ³-2 ⁴Between optimized choice, promptly select k=p=l, wherein

The ratio of the adjacent IMF yardstick of l=is 2 o'clock a screening number of times;

◆ according to the EMD decomposition result, the IMF reconstruct data of select target frequency place frequency range;

◆ use GZC (Generalized Zero-Crossing) method that the IMF reconstruct data is calculated its instantaneous frequency f _Gzc, as shown in Figure 7;

Instantaneous frequency f _GzcComputing formula following:

$f_{\mathrm{gzc}}=\frac{{4\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="HSA00000328444600012.png,HSA00000328444600031.png"\; state="\{\{state\}\}">f</figure-callout>}}_1+2({\mathrm{<figure-callout\; id="21"\; label="f"\; filenames="HSA00000328444600031.png,HSA00000328444600061.png"\; state="\{\{state\}\}">f</figure-callout>}}_{21}+f_{22})+f_{41}+f_{42}+f_{43}+f_{44}}{12}$

Wherein: ${\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="HSA00000328444600012.png,HSA00000328444600031.png"\; state="\{\{state\}\}">f</figure-callout>}}_1=\frac{1}{{4\mathrm{<figure-callout\; id="1"\; label="T"\; filenames="HSA00000328444600012.png,HSA00000328444600031.png"\; state="\{\{state\}\}">T</figure-callout>}}_1};$

$f_{2i}=\frac{1}{{2T}_{2i}},i=\mathrm{1,2};$

$f_{4i}=\frac{1}{T_{4i}},i=\mathrm{1,2,3,4},$

T is the time period between zero point and the peak value;

◆ according to instantaneous frequency, so when obtaining the Hilbert of eeg data-frequency figure, as shown in Figure 8;

In the 4th step, the relevant IMF reconstruct data that the 3rd step was obtained is asked its FFT frequency spectrum respectively, and is calculated the frequency spectrum extreme value according to 3 sections of total duration average marks, supposes that x is the frequency spectrum extreme value,

The frequency spectrum extreme value that obtains is encoded respectively according to following threshold setting:

A) 20＜x＜=29, x is encoded to 1;

B) 29＜x＜=37, x is encoded to 2;

C) 37＜x＜=37, x is encoded to 3;

Element in the code database of coding that obtains the most at last and formation in second step relatively; If in code database, the mark non-zero, otherwise be labeled as zero; Wherein non-zero number is the correct number of times of target identification, realizes that finally local frequency spectrum extreme value target identification method quantizes discrimination.Identification accuracy computing formula is following:

Last discrimination power is 77.78%.

In the accompanying drawing:

A is a potential electrode; B is a reference electrode; C is a ground electrode; D is a subject's head; E is an eeg amplifier; F is a computing machine; G is a computer screen; E1 is a first input end; E2 is second input end; E3 is the 3rd input end;

■ represents end points;

● represent the first approximation point: near two extreme point straight lines continuation of end points and the intersection point through the end points longitudinal axis;

▲ last the extreme point of representing the border predicted method to confirm;

T ₁；

T _2i，i＝1，2；

T _4i, i=1,2,4,5; More than three T be the time period between zero point and the peak value.

Claims (1)

1. based on HHT high frequency assembly coding stable state vision inducting current potential brain-machine interface method, it is characterized in that: may further comprise the steps:

Step 1; In the occipital region of subject's head (D) (Oz) position of sound production potential electrode (A); At the one-sided ear-lobe position of sound production reference electrode (B) of subject's head (D), at the frons Fpz of subject's head (D) position of sound production ground electrode (C), the output terminal of potential electrode (A) inserts the first input end (E1) of eeg amplifier (E); The output terminal of reference electrode (B) inserts second input end (E2) of eeg amplifier (E); The output terminal of ground electrode (C) inserts the 3rd input end (E3) of eeg amplifier (E), and the output terminal of eeg amplifier (E) links to each other with the input end of computing machine (F), and the output terminal of computing machine (F) is connected with computer screen (G);

Step 2, at stable state vision inducting current potential SSVEP response high band, promptly flicker frequency is higher than 25Hz, selects n frequency of stimulation, and n is the integer greater than 1, with the n of n frequency of stimulation through the formation of sequential permutation and combination ⁿIndividual stimulus sequence is presented in face of the experimenter through computer screen (G), and the stimulus sequence coding is formed code database;

Step 3; When the experimenter watches the high frequency sequential assembly coding stimulus sequence of step 2 attentively; The eeg signal acquisition system that connects through step 1 obtains EEG signals; And adopt the feature extraction that realizes EEG signals based on the frequency conversion EEG feature extraction method of Hilbert-Huang transform HHT, concrete steps are following:

◆ the eeg data to collecting carries out superposed average, improves signal to noise ratio (S/N ratio), obtains pretreated eeg data;

◆ utilize bandpass filtering that pretreated eeg data is carried out bandpass filtering, obtain filtered eeg data;

◆ adopt empirical mode decomposition EMD method that filtered eeg data is decomposed;

When the use experience pattern is decomposed the application of EMD method,, adopt the Boundary Prediction method to the end points problem; The extreme value computing formula that the Boundary Prediction method is confirmed is following:

Extreme value=MAX{| endpoint value |, | the first approximation point |;

To the stopping criterion problem, adopt fixing screening number of times, empirical mode decomposition EMD algorithm is the process of a screening, the mathematical formulae of screening is following:

Wherein: x (t) is a signal to be decomposed;

c _jBe the eigenmodes component IMF signal after the empirical mode decomposition EMD decomposition;

N decomposes back eigenmodes component IMF number for empirical mode decomposition EMD;

m _j, m _2jThe mean value curve that calculates through last lower envelope for screening process;

K, p is the screening number of times;

For the eigenmodes component IMF that guarantees to filter out has enough physical significances on amplitude and frequency; The ratio that promptly guarantees adjacent eigenmodes component IMF yardstick is near 2; To the number of times k of screening process, p must limit, specifically according to EEG signals characteristics optimized choice between 23-24; Promptly select k=p=l, wherein

The ratio of the adjacent eigenmodes component of l=IMF yardstick is 2 o'clock a screening number of times;

◆ rule of thumb pattern is decomposed the EMD decomposition result, the eigenmodes component IMF reconstruct data of select target frequency place frequency range;

◆ use general zero crossing GZC (Generalized Zero-Crossing) method that eigenmodes component IMF reconstruct data is calculated its instantaneous frequency f _Gzc

Instantaneous frequency f _GzcComputing formula following:

Wherein:

Wherein T is the time period between zero point and the peak value;

◆ according to instantaneous frequency f _Gzc, and then when obtaining the Hilbert Hilbert of eeg data-frequency figure;

Step 4, the relevant eigenmodes component IMF reconstruct data that step 3 is obtained calculates its Fast Fourier Transform (FFT) FFT frequency spectrum respectively according to 3 sections of total duration average marks; And then calculate the frequency spectrum extreme value, with the frequency spectrum extreme value that obtains and pre-set threshold relatively and coding, with the corresponding encoding ratio in the code database of formation in coding that obtains and the step 2; If in code database; Mark non-zero then, otherwise be labeled as zero, wherein non-zero number is the correct number of times of target identification; The local frequency spectrum extreme value target identification method of final realization quantizes discrimination, and identification accuracy computing formula is following:

Images