CN103735262A - Dual-tree complex wavelet and common spatial pattern combined electroencephalogram characteristic extraction method - Google Patents

Abstract

The present invention relates to a kind of dual-tree complex wavelet feature extracting methods that spatial model combines together. The present invention chooses the EEG signals in appropriate channel first, then according to dual-tree complex wavelet frequency segmentation the characteristics of, to original frequency carry out up-sampling or down-sampling, then utilize dual-tree complex wavelet multi-resolution decomposition, to obtain

,

,

With

The corresponding frequency range of frequency range of four species rhythm waves, and signal reconstruction is carried out under the scale, obtain multiple reconstruction signals under corresponding band, then identical decomposition and reconstruction is carried out to each appropriate channel, again the reconstruction signal of each frequency range in each channel is joined together to be input in spatial filter, the feature vector of 6 dimensions is obtained, Mental imagery classification of task is finally completed using support vector machines. Method proposed by the present invention not only carries out frequency information analysis to Mental imagery EEG signals, moreover it is possible to it is insufficient effectively to overcome the problems, such as that electrode is chosen.

Description

A kind of dual-tree complex wavelet is the brain electrical feature extracting method of spatial model combination together

Technical field

The invention belongs to EEG Processing field, relate to a kind of dual-tree complex wavelet brain electrical feature extracting method of spatial model combination together.

Background technology

For the patient of those neuromuscular system function major injuries, be badly in need of a kind of new method exchanging with the external world.And brain-computer interface a kind of such method just, it does not rely on brain outside nervous system and muscular tissue, is a kind of a kind of direct communication and control passage of setting up between human brain and computer or other electronic equipments.At present, the EEG signals type that is applicable to brain-computer interface has a variety of, and wherein motion imagination EEG signals is one of maximum type of application, because motion imagination EEG signals has without environmental stimuli, can realize the advantages such as asynchronous communication, meet the demand for development of BCI technology.

Up to the present, people still know little about it to the forming process of human thinking, and the various thinking activitiess that read people by brain electricity are also unrealistic.But EEG signals is used for moving, imagination identification aspect, has obtained certain progress.More typical application example is that the Niels Birbaurmer of Tübingen, Germany university utilizes paralytic's EEG signals successfully to control computer cursor in nineteen ninety to move.The researchs such as Pfurtscheller of Austria Graz University of Science and Technology show, when human body is made monolateral limb motion or imagery motion, the relevant current potential (ERD) that desynchronizes of brain offside generation event, and brain homonymy produces event related synchronization current potential (ERS).The experiment of Pfurtscheller can dope left hand or the motion of the right hand simultaneously, and classification accuracy rate reaches 85%.2000, Graz research group carried out the research of two kinds of motion imaginations of right-hand man classification.At present, the research group that professor Pfurtscheller leads has successfully been set up GrazI and GrazII system, first, experimenter controls system by imagining the EEG signals that right-hand man or forefinger, right crus of diaphragm motion produce, part experimenter is by after training, the motion imagination discrimination of two classes obtained up to 85% at linear resolution, yet the discrimination of three classes only has 77%.

The research that U.S. Wadsworth carries out at center is by recording the EEG signals of brain motorsensory cortex, to come one dimension or the two dimensional motion of steering needle.In its experiment, first to train subjects to learn freely to control the amplitude of μ rhythm and beta response, thereby realize the functions such as cursor movement, letter spelling and artificial limb control by the variation of these rhythm and pace of moving things.The Germany Berlin Muller of polytechnical university etc. has researched and developed Hex-O-Spell typewriter, the spelling that experimenter typewrites by the motion of imagination trick.Japanese scholars Yamawaki sets up time-frequency model and carries out sort research for motion imagination EEG signals, has also obtained good classifying quality.Subsequently, the feature extracting method based on public space pattern (CSP), has improved the signal to noise ratio of EEG signals greatly.The people such as the Kai Keng Ang of Singapore improve CSP method, the methods such as spatial filter group (FBCSP), the space filtering based on robustness bank of filters (RFBCSP), Time-Frequency Domain Filtering network (OSSFN) have been proposed respectively, progressively improved the degree of accuracy of pattern classification, and obtained excellent achievement in the BCI contest of 2008.Recently, the helicopter that the He of U.S. Minnesota university etc. utilizes multitask imagery motion EEG signals to control in virtual environment completes the motions such as forward/backward, rise/fall, left/right rotation, and three experimenters have obtained about 85% accuracy generally.

Domestic, the EEG signals research group of Tsing-Hua University carried out the systematic studys such as BCI cursor movement based on the motion imagination, rehabilitation supplemental training, obtained 79.48% online average classification accuracy rate and 85% the average classification accuracy rate of off-line in the Study of recognition of three type games imagination tasks.The Zheng Chong of Xi'an Communications University merit etc. utilizes complexity algorithm to extract EEG signals in addition, by the analysis of Mahalanobis distance discrimination formula, realizes two class mental tasks classifications, has obtained good effect.The Zhang Liqing of Shanghai Communications University etc. has studied the characteristic vector that common tensor Discrimination Analysis Algorithm is extracted the EEG sample of single, and recycling support vector machine is classified, and result shows that this algorithm has improved classification accuracy.The patriotic philosophy of Song of Southeast China University adopts wavelet transform and AR model to carry out feature extraction to the EEG signals of the motion imagination, two generic task motion imagination discrimination average out to 89.5%.

It is generally acknowledged, BCI systems attempt adopts the passage of few way, and the benefit of doing is like this apparent, and required electrode is few, not only shortens time, and low volume data needs little information processing cost.Corresponding with it, the scholars such as the German Berlin Blankertz of polytechnical university point out, adopt a small amount of passage of nervous physiology priori selection might not produce the result better than full tunnel collection, and electrode is chosen deficiency also can reduce classification accuracy rate.Subsequently, Sannelli, Schroder, Barachant etc. imagine that to several experimenters the EEG data of the motions such as right-hand man, foot study, when especially optionally selection portion subchannel is studied, the nicety of grading that different combination of channels obtain differs greatly.The minimum eeg data of take obtains best classifying quality as target, the scholar such as Arvaneh, Gao further studies CHANNEL OPTIMIZATION and selects problem, result shows to adopt the methods such as CSP and expansion algorithm, SVM recurrence passage exclusive method can both under certain selects criterion, find the channel position of applicable particular subject, not only reduce number of electrodes, and raising classification performance, also point out that a large amount of passage capable of beings provide the more information of horn of plenty simultaneously.

For CSP algorithm, need a large amount of electrodes, lack the shortcoming that frequency information is analyzed, the present invention reduces electrodeplate, reduces information processing cost, meanwhile increases frequency information analysis, makes up the deficiency of CSP algorithm in EEG Processing.

Summary of the invention

Object of the present invention is widely used aspect EEG feature extraction for CSP algorithm exactly, but the method needs a large amount of electrodes, lacks frequency information analysis, and a kind of dual-tree complex wavelet feature extracting method that spatial model combines together proposing.In order to have made up the defect of CSP space filtering shortage frequency information, improved the discrimination of the corresponding eigenvalue of different motion imagination task, and then improved the recognition accuracy of grader.

In order to realize above object, the inventive method mainly comprises the following steps:

Step (1) is obtained the motion imagination EEG signals of suitable passage.Specifically, carry out different motion imagination tasks can activate the zones of different on brain motor cortex according to brain, select corresponding electrode, thereby obtain the motion imagination EEG signals that this electrode gathers, its sample frequency is

Step (2) is asked for suitable double sampling frequency f _s, guarantee that the frequency range of the EEG signals that step (3) reconstruct obtains is 0～30Hz.Specifically, the frequency range of the l layer wavelet coefficient that multiple dimensioned decomposition obtains according to dual-tree complex wavelet, solves suitable f _sand l,

f _s/2 ^l≈30Hz (1)

Wherein, 30Hz is the maximum of EEG signals beta response ripple.Again according to f _sto original frequency carry out corresponding up-sampling or down-sampling, the sample frequency finally obtaining is f _s.If concrete sampling selection principle is

time, carry out up-sampling, otherwise carry out down-sampling.

Step (3) is carried out dual-tree complex wavelet decomposition and reconstruction to EEG signals.Specifically, first the EEG signals after double sampling is carried out to dual-tree complex wavelet decomposition, then choose δ (0.5～4Hz), θ (4～8Hz), α (8～14Hz) and frequency range corresponding to β (14～30Hz) 4 species rhythm wave frequency scope, and by its reconstruct.

Step (4) combines the reconstruction signal of each passage and carries out CSP space filtering.Specifically, the reconstruction signal of each suitable passage is combined, be then input in CSP spatial filter simultaneously and carry out filtering, filtered signal obtains the required characteristic vector f of grader through formula (2) _p.

$f_p=\lg (\mathrm{var}\left(Z_p\right)/\underset{i=1}{\overset{2m}{\mathrm{\Σ}}}\mathrm{var}\left(Z_i\right)),p=1,...,2m---\left(2\right)$

Wherein var represents the variance of asking vectorial, and m represents the logarithm of CSP wave filter, and log represents natural logrithm computing, Z _p(p=1,2 ..., 2m) represent the filtered characteristic vector of CSP.

Compare with existing motion imagination brain electrical feature extracting method, the method that the present invention proposes not only can be removed and the irrelevant frequency band of the imagination that moves, make up the defect that CSP space filtering lacks frequency information, thereby improve the discrimination of the corresponding eigenvalue of different motion imagination task; The requirement of CSP to port number be can also reduce, thereby the redundancy that some are uncorrelated or sound pollution passage provides and noise reduced to a certain extent.

The inventive method can meet the requirements for extracting features in pattern recognition task preferably, at brain-computer interface, disease of brain diagnostic field, has broad application prospects.

Accompanying drawing explanation

Fig. 1 is implementing procedure figure of the present invention, i.e. EEG feature extraction handling process.

The specific embodiment

Below in conjunction with accompanying drawing, describe in detail and the present invention is based on the dual-tree complex wavelet brain electrical feature extracting method of spatial model together, Fig. 1 is implementing procedure figure.

As Fig. 1, the enforcement of the inventive method mainly comprises following step:

(1) obtain multichannel motion imagination EEG signals eeg data;

(2) select corresponding electrode, obtain the EEG signals data of respective channel;

(3) calculate the sample frequency relevant to the multiple dimensioned frequency division of dual-tree complex wavelet, and original sample frequency is carried out to up-sampling or down-sampling;

(4) each passage EEG signals is carried out dual-tree complex wavelet decomposition;

(5) choose relevant frequency range, and by its reconstruct;

(6) combine the reconstruction signal of each passage, be input in CSP wave filter;

(7) filtered signal is taken the logarithm variation, is input to support vector machine classifier and carries out training and testing.

One by one each step is elaborated below.

Step 1: obtain multichannel motion imagination EEG signals sample data

Adopt 40 conduction polar caps in the U.S. Neuro Scan Scan4.3 of the company collecting device imagination process eeg signal acquisition that moves.Experimenter has worn brain electricity cap recoil on request on wheelchair, keeps quite, nature, watches the sight prompting of setting in experimental situation attentively.Adopt following two kinds of motion thought experiment normal forms: the right hand control wheelchair control bar forward, left hand controls the wheelchair control bar controlled motion form that corresponding wheelchair advances, brakes respectively backward, in implementation process, also can experimental concrete condition do suitable correction to the design of experiment model.

Step 2: select corresponding electrode, obtain the EEG signals data of respective channel

According to brain, carry out different motion imagination tasks and can activate the zones of different on brain motor cortex, then according to experimental paradigm, select corresponding electrode, obtain the EEG signals data of this passage, sample frequency is

Step 3: the selection of double sampling frequency

In order to guarantee that the frequency range of the EEG signals that step 4 reconstruct obtains is 0～30Hz, the frequency range of the l layer wavelet coefficient that multiple dimensioned decomposition obtains according to dual-tree complex wavelet, asks suitable f according to formula (1) _sand l.Again according to f _sto original frequency

carry out corresponding up-sampling or down-sampling, the sample frequency finally obtaining is f _s.If concrete sampling selection principle is

time, carry out up-sampling, otherwise carry out down-sampling.

Step 4: dual-tree complex wavelet decomposes

Each passage EEG signals data of selecting are carried out to L layer dual-tree complex wavelet and decompose, obtain the signal frequency range of L+1 reconstruct.If the sample frequency of EEG signals is f _s, carry out the decomposition of L layer, obtain reconstruction signal, wherein the frequency range of l layer wavelet coefficient and scale coefficient is as follows:

f(d _l)∈[f _s/2 ^l+1,f _s/2 ^l]Hz，l∈1,2,...,L

f(c _L)∈[0,f _s/2 ^L+1]

Step 5: choose relevant frequency range, and by its reconstruct

First the EEG signals after sampling is carried out to dual-tree complex wavelet decomposition, then choose δ, θ, α and frequency range corresponding to β tetra-species rhythm wave frequency scope, and by its reconstruct.

Step 6: CSP filtering

The reconstruction signal of each frequency range of each passage is combined, be then input in CSP spatial filter simultaneously and carry out filtering, its median filter logarithm is m.

Step 7: the motion imagination classification of task based on support vector machine

To through equation (2), obtain the required characteristic vector of grader through the filtered signal of step 6, the input using it as support vector machine classifier, carries out training and testing, completes the classification of multi-motion imagination task.

Claims (1)

1. a dual-tree complex wavelet brain electrical feature extracting method for spatial model combination together, is characterized in that the method comprises the steps:

Step (1) is obtained the motion imagination EEG signals of suitable passage; Specifically, according to brain, carry out the zones of different on different motion imagination mission-enabling brain motor cortexs, select corresponding electrode, thereby obtain the motion imagination EEG signals that this electrode gathers, its sample frequency is

Step (2) is asked for suitable double sampling frequency f _s, guarantee that the frequency range of the EEG signals that step (3) reconstruct obtains is 0～30Hz; Specifically, the frequency range of the l layer wavelet coefficient that multiple dimensioned decomposition obtains according to dual-tree complex wavelet, solves suitable f _sand l,

f _s/2 ^l≈30Hz (1)

Wherein, 30Hz is the maximum of EEG signals beta response ripple; Again according to f _sto original frequency

carry out corresponding up-sampling or down-sampling, the sample frequency finally obtaining is f _s; If concrete sampling selection principle is time, carry out up-sampling, otherwise carry out down-sampling;

Step (3) is carried out dual-tree complex wavelet decomposition and reconstruction to EEG signals; Specifically, first the EEG signals after double sampling is carried out to dual-tree complex wavelet decomposition, then choose δ, θ, α and frequency range corresponding to β tetra-species rhythm wave frequency scope, and by its reconstruct, wherein δ corresponding to 0.5Hz～4Hz, θ corresponding to 4Hz～8Hz, α corresponding to 8Hz～14Hz, β corresponding to 14Hz～30Hz;

Step (4) combines the reconstruction signal of each passage and carries out CSP space filtering; Specifically, the reconstruction signal of each suitable passage is combined, be then input in CSP spatial filter simultaneously and carry out filtering, filtered signal obtains the required characteristic vector f of grader through formula (2) _p;

$f_p=\lg (\mathrm{var}\left(Z_p\right)/\underset{i=1}{\overset{2m}{\mathrm{\Σ}}}\mathrm{var}\left(Z_i\right)),p=1,...,2m---\left(2\right)$

Wherein, var represents the variance of asking vectorial, and m represents the logarithm of CSP wave filter, and lg represents natural logrithm computing, Z _prepresent the filtered characteristic vector of CSP.

Images