CN106073702A - Many time-frequencies yardstick diencephalon myoelectricity coupling analytical method based on small echo transfer entropy - Google Patents

Abstract

The invention discloses a kind of many time-frequencies yardstick diencephalon myoelectricity coupling analytical method based on small echo transfer entropy, described method includes that brain electromyographic signal synchronous acquisition part and signal processing, brain electromyographic signal synchronous acquisition part include eeg signal acquisition and electromyographic signal collection；Signal processing includes that the small echo transfer entropy of Signal Pretreatment and brain myoelectricity analyzes method.The present invention has applicability, admissibility, has important using value in rehabilitation medicine field.

Description

Many time-frequencies yardstick diencephalon myoelectricity coupling analytical method based on small echo-transfer entropy

Technical field

The present invention relates to neural rehabilitation engineering and locomotory mechanism research field, be specifically related to a kind of based on small echo-transfer entropy Many time-frequencies yardstick diencephalon myoelectricity coupling analytical method.

Background technology

Brain electricity (electroencephalogram, EEG) and myoelectricity (electromyographic, EMG) signal wrap respectively The functional response information that brain is controlled to be intended to by information and muscle, the multi-scale coupling between brain electromyographic signal is controlled containing somatic movement Cortex-muscle function coupling (Functional corticomuscular coupling, FCMC) that message reflection is multi-level Link information.At present, brain myoelectricity synchronous characteristic research is based primarily upon coherence analysis, obtains brain sports consciousness and drives and muscle fortune Functional cohesion feature between dynamic response, but traditional coherent analysis can not embody coupling direction character.For being best understood from Function between cerebral cortex and respective muscle is mutual and information transmission characteristic, Granger Causality analysis be applied to brain myoelectricity with Step research, finds to there is two-way (descending EEG → EMG, up EMG → EEG) coupling contact between brain myoelectricity.But due to brain myoelectricity it Between coupling model is unknown and Function Coupling between brain electromyographic signal also exists nonlinear causal relationship, Glan based on set model Outstanding causal analysis method can not effectively describe brain myoelectricity Non-linear coupling feature.Transfer entropy has and does not relies on set model and reality The feature of existing quantitative analysis of nonlinear, it is possible to effectively estimate the Function Coupling intensity between cortex-muscle and information transfer side To.Therefore, transfer entropy model is for estimating the Function Coupling intensity between cortex-muscle and information direction of transfer feature, announcement Motor control and response mechanism between motor process mediopellis and muscle have feasibility.2015, author of the present invention " based on The brain electromyographic signal coupling analysis of multiple dimensioned transfer entropy " in it is proposed that multiple dimensioned transfer entropy method, and based on the method research not With brain electromyographic signal coupling feature in time scale.But along with the increase of coarse yardstick, sequence length reduces, may make Entropy estimate is inaccurate.The proposition of mobile equalization overcomes this drawback so that the length of time series of each yardstick keeps phase With.But more than research still has several drawbacks: when coarse and mobile equalization method are only brain electricity and electromyographic signal to be carried out Between sized, it is impossible to enough depict brain electricity and the time-frequency domain characteristic of myoelectricity and the non-thread of different time-frequency yardstick diencephalon electromyographic signal Property coupling and information transmission.

Summary of the invention

It is an object of the invention to provide one it can be found that non-linear dependencies between cortex muscle, further investigation brain skin Layer and coupling and many time-frequencies yardstick diencephalon myoelectricity coupling analysis based on small echo-transfer entropy of information transfer characteristic between muscle Method.

The step of the method for the invention is as follows:

Step 1, uses 64 to lead Neuroscan equipment synchronous acquisition EEG signals and electromyographic signal；

Step 2, utilizes Neuroscan device data to process the software EEG signals to collecting and electromyographic signal is gone respectively Except baseline drift, spilling, eye move and Hz noise；

Step 3, uses Daubechies class db4 wavelet basis function that brain electricity and electromyographic signal are carried out spectral decomposition, analyzes Synchronizing characteristics between EEG signals time-frequency yardstick different with electromyographic signal, Non-linear coupling and information transmission between quantitative description brain flesh Feature；

Step 4, carries out motor function analysis to Non-linear coupling between brain flesh and information transfer characteristic.

Further, in step 1, electrode for encephalograms uses international 10-20 system standard, using the mastoid process of ears as ginseng Examine, from the EEG signals of 32 top guide skin brain wave acquisition equipment record correspondence motions；Synamp2 equipment is used to gather electromyographic signal, Electrode, along muscle fiber direction, is placed on belly of muscle position.

Further,

The concrete grammar of described step 3 is as follows:

EEG signals x (t) and two groups of time sequences of electromyographic signal y (t) is built based on pretreated measured data in step 2 Row；EEG signals x (t) is carried out wavelet transformation, wavelet transformation is applied among partition of the scale；

First wavelet structure function, formula is as follows:

Ψ_j,k(t)=2^-j/2Ψ(2^-jt-k) (1)

In formula, Ψ (t) is morther wavelet；K is the translational movement in Ψ (t) vertical coordinate direction, and j represents the number of plies of signal, j, k ∈ Z, Z is set of integers；Scale parameter is 2^j, translation parameters 2^jk；T is time index；

Then EEG signals x (t) is carried out 7 layers of spectral decomposition, obtains wavelet conversion coefficient

$C_{j,k}={\∫}_{-\mathrm{\∞}}^{\mathrm{\∞}}x\left(t\right){\mathrm{\Ψ}}_{j,k}^{*}\left(t\right)dt---\left(2\right)$

Wavelet coefficient C_j,kAccording to frequency range order arrangement from high to low, extract the 3rd, 4,5,6,7 layer coefficients reconstruct Gamma (32～64Hz), beta (16～32Hz), alpha (8～16Hz), theta (4～8Hz) and delta (1～4Hz) are frequently The signal of section:

$x_i\left(t\right)=C\underset{-\mathrm{\∞}}{\overset{\mathrm{\∞}}{\Σ}}\underset{-\mathrm{\∞}}{\overset{\mathrm{\∞}}{\Σ}}{C}_{j_i,k}{\mathrm{\Ψ}}_{j,k}\left(t\right),(i=1,2,3,4,5)---\left(3\right)$

In formula,Quality factor for wave filter；

Reconstruction signal x₁(t)、x₂(t)、x₃(t)、x₄(t) and x₅(t) respectively corresponding brain Electricity Functional frequency band delta, theta, The signal of alpha, beta and gamma；

Above-mentioned identical wavelet transform procedure is carried out for electromyographic signal y (t), obtains y_m(t) (m=1,2,3,4,5), point The signal of not corresponding myoelectricity delta, theta, alpha, beta and gamma frequency range；

Based on transfer entropy computational methods, the small echo-transfer entropy WTE of structure x (t) to y (t)_x→y, formula is as follows:

${\mathrm{WTE}}_{x\→y}=\underset{y_{t+u}^m,y_t^m,x_t^i}{\Σ}p(y_{t+u}^m,y_t^m,x_t^i)log\frac{p(y_{t+u}^m,y_t^m,x_t^i)p\left(y_t^m\right)}{p(y_{t+u}^m,y_t^m)p(y_t^m,x_t^i)}---\left(4\right)$

In formula, u is predicted time；P () is the joint probability between variable；Represent brain electricity and myoelectricity respectively The delay vector of delta, theta, alpha, beta, gamma component；ForForecasting sequence；

WTE_x→yThen represent the x of EEG signals EEG_iT () component is to the y of electromyographic signal EMG_mTransfer entropy between (t) component Value；In like manner signal y (t) arrives the small echo-transfer entropy WTE of x (t)_y→xExpression formula be:

${\mathrm{WTE}}_{y\→x}=\underset{x_{t+u}^i,x_t^i,y_t^m}{\Σ}p(x_{t+u}^i,x_t^i,y_t^m)log\frac{p(x_{t+u}^i,x_t^i,y_t^m)p\left(x_t^i\right)}{p(x_{t+u}^i,x_t^i)p(x_t^i,y_t^m)}---\left(5\right)$

In formula,ForForecasting sequence；WTE_y→xRepresent the y of EMG_mT () is to component to the x of EEG_iBiography between (t) component Pass entropy；Transmission entropy is the biggest, illustrates that the coupling of cortex muscle is the strongest between this frequency range；Vice versa.

The present invention compared with prior art has the beneficial effect that: the inventive method utilizes small echo-transfer entropy to analyze brain myoelectricity Signal message transmission characteristic, nonlinear between quantitative description EEG signals with electromyographic signal frequency range couples and information transfer characteristic, Contributing to exploring the functional cohesion between cerebral cortex and muscle, study movement controls feedback mechanism and dyskinesia pathology machine System, sets up rehabilitation state evaluation index based on brain electromyographic signal, builds healing robot kinestate and patient physiological condition Evaluate mechanism, it is possible to obtain considerable Social benefit and economic benefit.

Accompanying drawing explanation

Fig. 1 is the structure diagram of Neuroscan equipment in the present invention.

Fig. 2 is the workflow diagram of the inventive method.

Fig. 3 be experimenter's C4 passage EEG signals wavelet decomposition after time-frequency result figure.

Fig. 4 is time-frequency result figure after the wavelet decomposition of electromyographic signal at experimenter's flexor digitorum superficialis.

Fig. 5 is the brain electromyographic signal small echo-transfer entropy analysis result figure of experimenter.

Drawing reference numeral: 1-electrode for encephalograms, 2-electrode cap, 3-brain myoelectricity Acquisition Instrument, 4-myoelectricity conducting wire, 5-electromyographic electrode.

Detailed description of the invention

EEG signals and electromyographic signal are the faintest, have the features such as non-linear, non-stationary and frequency domain characteristic is prominent.? In motor process, the interactive controlling mechanism between nervous system with muscle can be by the synchronization coupling analysis body of brain electromyographic signal Existing.Wavelet decomposition can extract the specific time-frequency data segment of brain electromyographic signal, and transfer entropy can portray non-thread between signal Property coupling and information transfer characteristic, the present invention by research brain myoelectricity between small echo-transfer entropy analysis, it is thus achieved that different motion shape Information transfering relation between cerebral cortex and muscle under state, and then the physiological mechanism that study movement dysfunction produces.

Embodiment 1:

As in figure 2 it is shown, method step is as follows:

Step 1, uses 64 to lead Neuroscan equipment synchronous acquisition EEG signals and electromyographic signal.

The structure of Neuroscan equipment is as it is shown in figure 1, led by electrode for encephalograms, electrode cap, brain myoelectricity Acquisition Instrument, myoelectricity Line, electromyographic electrode connect composition.

Eeg signal acquisition: electrode for encephalograms adopt international standards 10-20 electrode place standard, realize brain by electrode cap 2 Electricity electrode 1 contacts with scalp.The experiment of brain electromyographic signal synchronous acquisition is carried out under the output motion of hand static state grip.M1, M2 lead After connection is connected respectively to left and right ear, mastoid process overhead hits exactly as reference electrode, ground electrode arrangement, adopts from 32 top guide skin brain electricity Collection equipment selects the EEG signals of C3, C4 and CPZ district record correspondence motion.

Electromyographic signal collection: use Synamp2 equipment to gather flexor digitorum superficialis (flexor digitorum Superficialis, FDS) electromyographic signal at place, first with the skin surface at the tested position of alcohol wipe, remove skin surface Oils and fats and scurf, be then pasted onto belly of muscle position along muscle fiber direction by electromyographic electrode 5, and by suitable for myoelectricity conducting wire 4 The interference that in the fixing motor process of minimizing as far as possible, conducting wire rocks.

Step 2, utilizes Neuroscan device data to process the software EEG signals to collecting and electromyographic signal is gone respectively Except baseline drift, spilling, eye move and Hz noise；

Step 3, uses Daubechies class db4 wavelet basis function that brain electricity and electromyographic signal are carried out spectral decomposition, based on In step 2, pretreated measured data builds EEG signals x (t) and two groups of time serieses of electromyographic signal y (t).With brain telecommunications As a example by number x (t), wavelet transformation is applied among partition of the scale.First wavelet structure function, formula is as follows:

Ψ_j,k(t)=2^-j/2Ψ(2^-jt-k) (1)

In formula, Ψ (t) is morther wavelet；K is the translational movement in Ψ (t) vertical coordinate direction, and j represents the number of plies of signal, j, k ∈ Z, Z is set of integers；Scale parameter is 2^j, translation parameters 2^jk；T is time index.

Then EEG signals x (t) is carried out 7 layers of spectral decomposition, obtains wavelet conversion coefficient

$C_{j,k}={\∫}_{-\mathrm{\∞}}^{\mathrm{\∞}}x\left(t\right){\mathrm{\Ψ}}_{j,k}^{*}\left(t\right)dt---\left(2\right)$

The wavelet coefficient C obtained based on above formula_j,kAccording to frequency range from high to low order arrangement, extract the 3rd, 4,5, 6,7 layer coefficients reconstruct gamma (32～64Hz), beta (16～32Hz), alpha (8～16Hz), theta (4～8Hz) and The signal of delta (1～4Hz) frequency range:

$x_i\left(t\right)=C\underset{-\mathrm{\∞}}{\overset{\mathrm{\∞}}{\Σ}}\underset{-\mathrm{\∞}}{\overset{\mathrm{\∞}}{\Σ}}{C}_{j_i,k}{\mathrm{\Ψ}}_{j,k}\left(t\right),(i=1,2,3,4,5)---\left(3\right)$

In formula,Quality factor for wave filter；

Reconstruction signal x₁(t)、x₂(t)、x₃(t)、x₄(t) and x₅(t) respectively corresponding brain Electricity Functional frequency band delta, theta, The signal of alpha, beta and gamma.

Above-mentioned identical wavelet transform procedure is carried out for electromyographic signal y (t), obtains y_m(t) (m=1,2,3,4,5), point The signal of not corresponding myoelectricity delta, theta, alpha, beta and gamma frequency range.

Based on transfer entropy computational methods, the small echo-transfer entropy WTE of structure x (t) to y (t)_x→y, formula is as follows:

${\mathrm{WTE}}_{x\→y}=\underset{y_{t+u}^m,y_t^m,x_t^i}{\Σ}p(y_{t+u}^m,y_t^m,x_t^i)log\frac{p(y_{t+u}^m,y_t^m,x_t^i)p\left(y_t^m\right)}{p(y_{t+u}^m,y_t^m)p(y_t^m,x_t^i)}---\left(4\right)$

In formula, u is predicted time；P () is the joint probability between variable；Represent brain electricity and myoelectricity respectively The delay vector of delta, theta, alpha, beta, gamma component；ForForecasting sequence；

WTE_x→yThen represent the x of EEG signals EEG_iT () component is to the y of electromyographic signal EMG_mTransfer entropy between (t) component Value.In like manner signal y (t) arrives the small echo-transfer entropy WTE of x (t)_y→xExpression formula be:

${\mathrm{WTE}}_{y\→x}=\underset{x_{t+u}^i,x_t^i,y_t^m}{\Σ}p(x_{t+u}^i,x_t^i,y_t^m)log\frac{p(x_{t+u}^i,x_t^i,y_t^m)p\left(x_t^i\right)}{p(x_{t+u}^i,x_t^i)p(x_t^i,y_t^m)}---\left(5\right)$

In formula,ForForecasting sequence；WTE_y→xRepresent the y of EMG_mT () is to component to the x of EEG_iBiography between (t) component Pass entropy；Transmission entropy is the biggest, illustrates that the coupling of cortex muscle is the strongest between this frequency range；Vice versa.

Based on These parameters, calculate under the output motion of hand static state grip, on different coupling directions, between different time-frequency yardstick WTE value, i.e. can between quantitative description EEG signals EEG and electromyographic signal EMG many time-frequencies yardstick non-linear synchronization coupling spy Levy.

For verifying that brain electromyographic signal small echo of the present invention-transfer entropy analyzes feasibility and the effectiveness of method, raise 8 The experimenter of name health carries out the output experiment of hand static state grip, and experimenter's relevant information is as shown in table 1.According to of the present invention Brain myoelectricity gather with analyze process, synchronous acquisition experimenter's constant force output motion under brain electromyographic signal, be analyzed and grind Study carefully the coupling between subject motion's motor process mediopellis muscle and information transmission mechanism.

This experiment gathers the myoelectricity at left hand flexor digitorum superficialis (flexor digitorum superficialis, FDS) place Signal and offside C4 passage EEG signals, and calculate WTE value.

Fig. 3 and Fig. 4 is respectively experimenter's brain electricity, the electromyographic signal knot of the time-frequency domain on each time-frequency yardstick after wavelet decomposition Really (left side is time-domain diagram, and right side is frequency domain figure), it can be seen that brain electricity and electromyographic signal can obtain after wavelet decomposition Signal to delta, theta, alpha, beta and gamma frequency range.

Fig. 5 is the meansigma methods after the small echo of coupling between 8 experimenter's brain fleshes-transfer entropy analysis.There it can be seen that quiet Between state grip output procedure mediopellis muscle, the stiffness of coupling of beta frequency range is the most notable, and each time-frequency on different coupling direction Stiffness of coupling between yardstick there is also difference, provides theoretical study method for probing into neuromuscular function coupling mechanism.

Table 1 experimenter's relevant information.

Claims (3)

1. many time-frequencies yardstick diencephalon myoelectricity coupling analytical method based on small echo-transfer entropy, it is characterised in that described method Step as follows:

Step 1, uses 64 to lead Neuroscan equipment synchronous acquisition EEG signals and electromyographic signal；

Step 2, utilizes Neuroscan device data to process the software EEG signals to collecting and electromyographic signal removes base respectively Line drift, spilling, eye move and Hz noise；

Step 3, uses Daubechies class db4 wavelet basis function that brain electricity and electromyographic signal are carried out spectral decomposition, analyzes brain electricity Synchronizing characteristics between signal time-frequency yardstick different with electromyographic signal, Non-linear coupling and information transmission spy between quantitative description brain flesh Levy；

Step 4, is analyzed Non-linear coupling and information transfer characteristic between the brain flesh under kinestate.

Many time-frequencies yardstick diencephalon myoelectricity coupling analytical method based on small echo-transfer entropy the most according to claim 1, it is special Levying and be: in step 1, electrode for encephalograms uses international 10-20 system standard, using the mastoid process of ears as reference, from 32 top guide skins The EEG signals of brain wave acquisition equipment record correspondence motion；Using Synamp2 equipment to gather electromyographic signal, electrode is along muscle fiber Direction, is placed on belly of muscle position.

Many time-frequencies yardstick diencephalon myoelectricity coupling analytical method based on small echo-transfer entropy the most according to claim 1, it is special Levying and be, the concrete grammar of described step 3 is as follows:

EEG signals x (t) and two groups of time serieses of electromyographic signal y (t) is built based on pretreated measured data in step 2； EEG signals x (t) is carried out wavelet transformation, wavelet transformation is applied among partition of the scale；

First wavelet structure function, formula is as follows:

Ψ_j,k(t)=2^-j/2Ψ(2^-jt-k) (1)

In formula, Ψ (t) is morther wavelet；K is the translational movement in Ψ (t) vertical coordinate direction, and j represents that the number of plies of signal, j, k ∈ Z, Z are Set of integers；Scale parameter is 2^j, translation parameters 2^jk；T is time index；

Then EEG signals x (t) is carried out 7 layers of spectral decomposition, obtains wavelet conversion coefficient

$C_{j,k}={\∫}_{-\mathrm{\∞}}^{\mathrm{\∞}}x\left(t\right){\mathrm{\Ψ}}_{j,k}^{*}\left(t\right)dt---\left(2\right)$

Wavelet coefficient C_j,kAccording to frequency range order arrangement from high to low, extract the 3rd, 4,5,6,7 layer coefficients reconstruct Gamma (32～64Hz), beta (16～32Hz), alpha (8～16Hz), theta (4～8Hz) and delta (1～4Hz) are frequently The signal of section:

$x_i\left(t\right)=C\underset{-\mathrm{\∞}}{\overset{\mathrm{\∞}}{\Σ}}\underset{-\mathrm{\∞}}{\overset{\mathrm{\∞}}{\Σ}}{C}_{j_i,k}{\mathrm{\Ψ}}_{j,k}\left(t\right),(i=1,2,3,4,5)---\left(3\right)$

In formula,Quality factor for wave filter；

Reconstruction signal x₁(t)、x₂(t)、x₃(t)、x₄(t) and x₅(t) respectively corresponding brain Electricity Functional frequency band delta, theta, The signal of alpha, beta and gamma；

Above-mentioned identical wavelet transform procedure is carried out for electromyographic signal y (t), obtains y_m(t) (m=1,2,3,4,5), the most right Answer the signal of myoelectricity delta, theta, alpha, beta and gamma frequency range；

Based on transfer entropy computational methods, the small echo-transfer entropy WTE of structure x (t) to y (t)_x→y, formula is as follows:

${\mathrm{WTE}}_{x\→y}=\underset{y_{t+u}^m,y_t^m,x_t^i}{\Σ}p(y_{t+u}^m,y_t^m,x_t^i)log\frac{p(y_{t+u}^m,y_t^m,x_t^i)p\left(y_t^m\right)}{p(y_{t+u}^m,y_t^m)p(y_t^m,x_t^i)}---\left(4\right)$

In formula, u is predicted time；P () is the joint probability between variable；Respectively represent brain electricity and myoelectricity delta, The delay vector of theta, alpha, beta, gamma component；ForForecasting sequence；

WTE_x→yThen represent the x of EEG signals EEG_iT () component is to the y of electromyographic signal EMG_mTransmission entropy between (t) component；In like manner Signal y (t) arrives the small echo-transfer entropy WTE of x (t)_y→xExpression formula be:

${\mathrm{WTE}}_{y\→x}=\underset{x_{t+u}^i,x_t^i,y_t^m}{\Σ}p(x_{t+u}^i,x_t^i,y_t^m)log\frac{p(x_{t+u}^i,x_t^i,y_t^m)p\left(x_t^i\right)}{p(x_{t+u}^i,x_t^i)p(x_t^i,y_t^m)}---\left(5\right)$

In formula,ForForecasting sequence；WTE_y→xRepresent the y of EMG_mT () is to component to the x of EEG_iTransfer entropy between (t) component Value；Transmission entropy is the biggest, illustrates that the coupling of cortex muscle is the strongest between this frequency range；Vice versa.