CN102488515B - Conjoint analysis method for electroencephalograph and electromyography signals based on autonomous movement and imagination movement - Google Patents

Abstract

A joint analysis method of EEG signals based on voluntary actions and imaginary actions: system settings, using LabVIEW8.6 to generate square wave pulse signals; EEG signal acquisition and EMG signal acquisition respectively, including: in the voluntary action mode EEG signals and EMG signals, and EEG signals and EMG signals under imaginary action mode; denoising preprocessing is performed on the collected raw data; denoising preprocessed voluntary action mode and imaginary action mode EEG time-domain signal analysis under voluntary action mode and imaginary action mode; EEG time-domain signal analysis of voluntary action and imaginative action mode after denoising preprocessing The time-frequency signal analysis is carried out on the signal, and the time-frequency signal analysis is based on the wavelet transform based on Morlet; the bias-oriented coherent analysis is performed, and the Granger causality is used to conduct the bias-oriented coherent analysis. The invention provides new evaluation parameters for the monitoring of rehabilitation auxiliary equipment and the evaluation of body movement level.

Description

Joint analysis method of brain electromyography signal based on voluntary action and imagined action

technical field

The present invention relates to one. In particular, it involves a method of synchronously collecting EEG signals and EMG signals in related areas in different action modes, and performing data processing and bias-oriented coherent analysis on the EEG signals in the two action modes to obtain EEG signals. The causality of the signal when the action occurs and the flow of information are based on the joint analysis method of brain electromyography signals under voluntary actions and imagined actions.

Background technique

In 1924, Dr. Hans Berger (1873-1941), a professor of psychiatry at Jena University in Germany, first recorded human scalp EEG signals, and named the brain electrical activity electroencephalogram (EEG) for the first time. However, because there were few scholars studying the electrical activity of the central nervous system at that time, Berger's achievements were not recognized by most physiologists and neurologists. Until 1933, the famous British physiologist E.D. Adrain (Nobel Prize winner in 1934) and B. Mathews studied EEG in the Physiology Laboratory of Cambridge University, affirming Berger's research. After that, EEG research developed rapidly and was accepted by the whole world. sEMG (Surface Electromyogram Signal), that is, surface electromyography signal, is a one-dimensional time series signal of neuromuscular activity recorded from the muscle surface through electrodes. It is related to the metabolic state and other factors, and can reflect the muscle activity state and functional state in real time, accurately and in a non-invasive state. Therefore, it can reflect neuromuscular activity to a certain extent, and can be used in the diagnosis of neuromuscular diseases in clinical medicine, the functional analysis of muscle work in the field of ergonomics, the evaluation of muscle function in the field of rehabilitation medicine, and in sports science. Judgment of fatigue, analysis of rationality of sports techniques. Since researchers and scholars began to consider and study EEG and EMG together, there are three commonly used analysis modes:

(1) Automated actions (different tasks) to analyze the correlation of EEG. For example, A.A. Abdul-latif of the University of Melbourne studied the changes in EEG correlations in voluntary movements of the left and right hands in 2004, pointing out the domination and contribution of the ipsilateral motor cortex of the brain when the ipsilateral limbs moved; Li Yan of Jilin University in 2001 in his degree In the paper, the functional MRI study of the brain activation area of normal people in three voluntary finger movement modes showed that the simple movements of the handed hand are mainly dominated by the contralateral brain SM1, while SM1 on both sides is involved in the simple movements of the non-handed hand . Voluntary movements involved more areas than simple movements, and both sides of the SMA participated; imaginary movements were mainly dominated by SMA and PMA.

(2) The characteristics of EEG signals were analyzed by imaginative action mode. Studies have shown that motor imagery therapy can effectively improve the motor function of hemiplegic patients with cerebral infarction. This is mainly based on the plasticity theory of the brain. For incomplete paralysis, in order to produce voluntary movement, there must be a motor idea first, and then muscle contraction and limb movement. One of the functions of rehabilitation is to repeatedly strengthen the normal motor control mode from the brain to the muscle group. Movement ideas based on imaginary actions can effectively promote the formation of this normal motor reflex arc. In the study of Gerardin et al., 8 right-handed normal people were allowed to imagine flexion and extension of the right hand fingers. After MIR, it was found that the bilateral premotor area, parietal lobe, basal ganglia and cerebellum were activated in the same way as the actual exercise. Later, the British Nikhil Sharma confirmed that for stroke patients, "motor imagery" can be used to improve limb movement disorders after stroke. Some scholars in China have also done similar work, which supports the above point of view. These studies suggest that motor imagery can be used to partially activate damaged motor networks in patients with motor disabilities.

(3) Analyze the correlation of EEG under the combination of voluntary action and imaginary action. In 2010, YasunariH explored the EEG-EMG correlation of lower extremity muscles under active action and imaginary action, and obtained the reflection of the overlapping cerebral cortex of motor imagination and actual action; in the same year, Nan Liang of Hiroshima University When studying the imaginary movement of the contralateral limb based on TMS (transcranial magnetic stimulation), it is also pointed out that the excitability of the motor cortex caused by the contralateral movement will be inhibited or disturbed by the movement of the ipsilateral limb, which may be caused by the inhibitory effect of the corpus callosum ; A 2008 study by Andrea Zimmer of the University of Zurich showed that imagery therapy for stroke patients can have a greater effect than other physical therapies.

Compared with independent signal analysis, causal analysis of EEG signals can more directly and accurately reflect the relationship between signals, which makes research based on EEG correlation analysis have a wide range of application values: it can gain insight into certain movement disorders The pathological mechanism of diseases (such as Parkinson's disease), providing effective basis and new ways for functional restoration and replacement after diseases; rehabilitation methods that can effectively improve the recovery period of sports injuries; evaluation and research methods that can improve human exercise levels and balance ability .

Contents of the invention

The technical problem to be solved by the present invention is to provide a causal analysis of EEG and EMG signals in different modalities by using biased coherence. By synchronously stimulating EEG and EMG data in different modalities, the overall It analyzes the full-band information of EEG signals and electromechanical signals, and gives the causality and information flow between the two in the frequency domain, so as to provide an evaluation parameter and a motor mechanism evaluation parameter based on autonomous actions for rehabilitation. , Brain electromyography signal joint analysis method under imaginative action.

The technical scheme adopted in the present invention is: a method for joint analysis of brain electromyography signals based on voluntary actions and imagined actions, including the following steps:

1) Perform system settings, that is, use LabVIEW8.6 to generate a square wave pulse signal with a period of 10s and a duty cycle of 0.2;

2) Collect EEG signals and EMG signals respectively, including: EEG signals and EMG signals under voluntary action mode, EEG signals and EMG signals under imaginative action mode;

3) Perform denoising preprocessing on the collected raw data;

4) Analyze the EEG time-domain signals under the voluntary action mode and the imagined action mode on the EEG signal time-domain images under the voluntary action mode and the imaginary action mode after the denoising preprocessing;

5) Carrying out time-frequency signal analysis of the brain electromyography signal under the voluntary action after the denoising pretreatment and the imaginary action mode, and the described time-frequency signal analysis adopts wavelet transform based on Morlet;

6) Perform bias-oriented coherence analysis, specifically using Granger causality to perform bias-oriented coherence analysis.

The use of LabVIEW8.6 described in step 1 to generate a square wave pulse signal with a period of 10s and a duty cycle of 0.2 includes the following process:

(1) Set the sampling rate and sampling mode;

(2) Generate a single-cycle analog waveform;

(3) start output;

(4) Judgment: Whether the remainder of the acquisition time/10 is greater than 2, if yes, continue to judge after turning off the light, otherwise continue to judge after turning on the light.

The voluntary action mode described in step 2 is specifically: the subject rests for 10 seconds; starts LabVIEW, generates a square wave pulse with a period of 10s and a duty cycle of 0.2, so that the indicator light is lit when the high level is triggered, And last for 2s, and then the indicator light is turned off for 8s. When the indicator light is on, the subject quickly flexes the right middle finger; at the same time, the EEG signals at C3 and C4 in the EEG lead diagram and the EMG signals at the finger superficial flexors are recorded , Square wave pulse signal for 60 seconds.

The imaginary action mode described in step 2 is specifically: the subject rests for 10 seconds; starts LabVIEW, generates a square wave pulse with a period of 10s and a duty cycle of 0.2, so that the indicator light is lit when the high level is triggered, And last for 2s, and then the indicator light is off for 8s. When the indicator light was on, the subject quickly flexed the right middle finger; at the same time, the EEG signals at C3 and C4, the EMG signals at the flexor finger superficialis (FDS), and the synchronous pulse signal were recorded for 60 seconds.

The denoising preprocessing of the collected raw data described in step 3 is to use the Butterworth third-order bandpass filter (including 50Hz power frequency notch) to filter the EEG signal and the EMG signal respectively, according to the signal The effective frequency band characteristics of the selected EEG signal cut-off frequency: 0.5Hz and 40Hz; EMG signal cut-off frequency: 0.5Hz and 200Hz. Afterwards, due to the high sampling rate of the original data, the EEG signal and EMG signal were down-sampled to 512Hz.

The method for joint analysis of EEG signals based on voluntary actions and imaginative actions of the present invention provides a method for analyzing the deviation and coherence of EEG signals and EMG signals in different action modes, aiming at EEG and EMG signals The relationship between EEG, EEG and EMG signals has been judged and analyzed for coherence, and obvious results have been obtained, thus providing new evaluation parameters for rehabilitation auxiliary equipment monitoring and body movement level assessment. There are practical application prospects in the field of motion mechanism research.

Description of drawings

Figure 1 is a block diagram of joint analysis of brain electromyography signals based on voluntary actions and imagined actions;

Figure 2 is a block diagram of the Labview 8.6 synchronous pulse flow;

Figure 3 is a schematic diagram of EEG leads;

Figure 4(a) is a time-domain diagram of the EEG signal of the subject in the voluntary action mode;

Figure 4(b) is the time-domain diagram of the brain electromyography signal of the subject in the imaginary action mode;

Fig. 5 (a) Spectrogram of the EEG signal of the subject in the voluntary action mode;

Figure 5(b) The spectrogram of the EEG signal of the subject in the imaginary action mode;

Figure 6(a) EEG bias-oriented coherence analysis of subjects in voluntary action mode

Figure 6(b) EEG bias-oriented coherence analysis of subjects in the imaginative and active mode

In the picture:

1: Cerebral cortex 2: Upper limb muscles

3: Surface electrode 4: Surface electrode

5: voluntary action 6: stimulating action

7: EEG amplifier 8: EMG amplifier

9: Digital bioelectricity acquisition 10: Data processing

Detailed ways

The method for joint analysis of brain electromyography signals based on voluntary actions and imaginary actions of the present invention will be described in detail below in conjunction with the embodiments and drawings.

The joint analysis method of brain electromyography based on voluntary action and imaginary action of the present invention is a joint analysis method of brain electromyography that maps time domain information to frequency domain based on Granger causality. This design belongs to the technical field of rehabilitation of the disabled. Its technical process is: by synchronously collecting EEG signals and EMG signals in related areas under different action modes, performing data processing and bias-oriented coherent analysis on the EEG signals in the two action modes to obtain EEG signals. The causality of the signal and the flow of information when the action occurs.

The method for joint analysis of brain electromyography signals based on voluntary actions and imagined actions of the present invention includes the following steps:

1) Perform system settings, that is, use LabVIEW8.6 to generate a square wave pulse signal with a period of 10s and a duty cycle of 0.2. LabVIEW (Laboratory Virtual instrument Engineering Workbench) is a graphical programming language development environment, which is widely accepted by industry, academia and research laboratories as a standard data acquisition and instrument control software. LabVIEW integrates all functions of communication with hardware and data acquisition cards meeting GPIB, VXI, RS-232 and RS-485 protocols. It also has built-in library functions for easy application of software standards such as TCP/IP and ActiveX. This is a powerful and flexible software. It is easy to use it to build your own virtual instrument, and its graphical interface makes the process of programming and use lively and interesting.

The use of LabVIEW8.6 to generate a square wave pulse signal with a period of 10s and a duty cycle of 0.2 includes the following process:

(1) Set the sampling rate and sampling mode;

(2) Generate a single-cycle analog waveform;

(3) start output;

(4) Judgment: Whether the remainder of the acquisition time/10 is greater than 2, if yes, continue to judge after turning off the light, otherwise continue to judge after turning on the light.

2) Collect EEG signals and EMG signals respectively, including: EEG signals and EMG signals under voluntary action mode, EEG signals and EMG signals under imaginative action mode;

The collection of EEG signals adopts the international standard 10-20 electrode placement standard, and the electrodes are connected to the scalp through the electrode cap. Because the areas where the brain controls the movement of the human body are relatively obvious in C3 and C4, the EEG signals are collected at C3 and C4. Using the single-level lead method, the A1 and A2 leads are respectively connected to the left and right earlobes as irrelevant electrodes, as shown in Figure 3.

In this experiment, under different action modes, the middle finger of the subjects is required to move actively or passively, and the target muscle that needs to be mobilized and involved is the flexor digitorum superficialis (FDS).

The voluntary action mode is specifically: the subject rests for 10 seconds; starts LabVIEW, generates a square wave pulse with a period of 10s and a duty cycle of 0.2, so that the indicator light is lit when the high level is triggered, and lasts 2s, and then the indicator light was turned off for 8s. When the indicator light was on, the subject quickly flexed the right middle finger; at the same time, the EEG signals at C3 and C4 in the EEG lead diagram, the EMG signals at the superficial flexor muscles, and square Wave pulse signal for 60 seconds.

The specific imaginary action mode is: the subject rests for 10 seconds; starts LabVIEW, generates a square wave pulse with a period of 10s and a duty cycle of 0.2, so that the indicator light is lit when the high level is triggered, and lasts 2s, and then the indicator light turns off for 8s. When the indicator light was on, the subject quickly flexed the right middle finger; at the same time, the EEG signals at C3 and C4, the EMG signals at the flexor finger superficialis (FDS), and the synchronous pulse signal were recorded for 60 seconds.

3) Perform denoising preprocessing on the collected raw data;

Because the collected raw data is mixed with a large amount of background noise, before the data analysis, the raw data needs to be preprocessed, and the described raw data collected is denoised and preprocessed using a Butterworth third-order bandpass filter ( Including 50Hz power frequency notch) to filter the EEG signal and EMG signal respectively, according to the effective frequency band characteristics of the signal, select the EEG signal cutoff frequency: 0.5Hz and 40Hz; EMG signal cutoff frequency: 0.5Hz and 200Hz. Afterwards, due to the high sampling rate of the original data, the EEG signal and EMG signal were down-sampled to 512Hz.

4) Analyze the EEG time-domain signals under the voluntary action mode and the imagined action mode on the EEG signal time-domain images under the voluntary action mode and the imaginary action mode after the denoising preprocessing;

Figure 4 shows the EEG time-domain signals of the subject in the voluntary action mode and the imaginative action mode. There is a significant change in the value, indicating that when the muscle contracts or relaxes, the electrical activity of the muscle is more active than when it is resting. In Fig. 4(b), it shows the time-domain diagram of the EEG signal of the subject in the imaginary action mode. Since the EMG signal under imaginary action is almost stationary, no clear change in amplitude can be seen in the time-domain diagram of the EMG signal.

5) Carrying out time-frequency signal analysis of the brain electromyography signal under the voluntary action after the denoising pretreatment and the imaginary action mode, and the described time-frequency signal analysis adopts wavelet transform based on Morlet;

Morlet wavelet transform is a kind of continuous wavelet transform. The basic idea is: to convolve the continuous time signal s(t) with Morlet wavelet w(t, f), so as to obtain the time-frequency energy distribution changing with time, that is,

TF(t, f) = |w(t, f) ^* s(t)| ²

Morlet wavelet w(t, f) is a Gaussian function that reproduces modulation, and has a Gaussian distribution in the time domain (standard deviation σ _t ) and frequency domain (standard deviation σ _f ), for a certain frequency f, its expression is :

$\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; Aexp</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="t"\; filenames="HDA0000117967410000011.png,HDA0000117967410000031.png"\; state="\{\{state\}\}">t</figure-callout></span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; i\&pi;ft</span>}<span\; class="notranslate">\; )</span>$

where, σ _t = 1/2πσ _f , $A={\left({\mathrm{\&sigma;}}_t{\mathrm{\&pi;}}^{\frac{1}{2}}\right)}^{-\frac{1}{2}}$

A is a normalization factor, and its purpose is to ensure that the energy of the wavelet basis itself is 1.

The Morlet wavelet family has a constant ratio f/ _σf (in practical applications, the value is generally greater than -5), so the _σf and _σt corresponding to different frequencies f are different, that is, they have a variable frequency in the entire time-frequency plane. Variable time-frequency resolution: It can provide high time resolution in the high frequency area and high frequency resolution in the low frequency area.

As shown in Figure 5(a) and Figure 5(b), it can be clearly seen that in the voluntary action mode, when the action occurs, the low-frequency component of the EEG signal, that is, the α frequency band (8-13Hz) produces Event-related desynchronization, that is, the ERD phenomenon; and then, about a second or so, the contralateral EEG produced an event-related desynchronization phenomenon in the β frequency band (14-30), that is, the ERS phenomenon. When the subjects start to imagine, they will find that the energy at C3 of the EEG signal begins to decrease (Figure 5(a)), which is reflected in the α-band and β-band, which is also a common phenomenon that brain waves will appear when imagining actions. The phenomenon, which we call event-related desynchronization or ERD phenomenon, appears at the C3 EEG lead, which also explains the contralateral dominance principle of the brain to limb movements.

6) Perform bias-oriented coherence analysis, specifically using Granger causality to perform bias-oriented coherence analysis.

In neuroscience, studying the cognitive functions contained in different cortical regions and evaluating the connections between regions are two core issues. In order to study the connection between various brain regions, scholars often use some mathematical methods to evaluate the interaction between various neural clusters. For example: coherence, correlation and phase synchronization, etc. However, these methods cannot distinguish the direction of information flow or causality between brain regions. Therefore, Granger causality formulated the concept of causality in 1969 and applied it to linear time series models. Nowadays, Granger causality is widely used in economics, physiology, computer neuroscience and many other fields to study the internal relationship between various variables.

a) Granger causality

The basic idea of Granger causality is: if the current value of the first time series is estimated by the past value of the first time series and the past value of the second time series, and compared with only the first time series If the forecast error variance decreases when the past values are estimated, then the second time series is said to be the cause of the first time series, otherwise it is not. Time is a very important element in Granger causality, what happens in the front is the cause, and what happens in the back is the effect. For two-dimensional time series data, Granger causality can be directly used to analyze the internal relationship between the data. For multidimensional time series data, due to the interaction between variables, Granger causality cannot be directly used to analyze the direct interaction of variables, and the partial correlation causality introduced earlier can be used to study the internal direct relationship between these variables .

The steps of Granger causality test are as follows:

(1) Regress the current y on all the lagging items y and other variables (if any), that is, y on the lagging items of y y _t-1 , y _t-2 , ..., y _tq and others Variables, but in this regression does not include the lag term x, which is a constrained regression. The constrained residual sum of squares RSSR is then derived from this regression.

(2) Do a regression with a lagging term x, that is, add a lagging term x to the previous regression formula, which is an unconstrained regression, from which the unconstrained residual sum of squares RSSUR is obtained.

(3) The null hypothesis is H0: α1=α2=...=αq=0, that is, the lag item x does not belong to this regression.

(4) In order to test this hypothesis, use the F test, namely

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; RSS</span>}}_{\mathrm{<span\; class="notranslate">\; R</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; RSS</span>}}_{\mathrm{<span\; class="notranslate">\; UR</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; q</span>}}{{\mathrm{<span\; class="notranslate">\; RSS</span>}}_{\mathrm{<span\; class="notranslate">\; UR</span>}}<span\; class="notranslate">\; /</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>}$

It follows an F-distribution with degrees of freedom q and (n-k).

(5) The null hypothesis is rejected if the calculated F value at the chosen significance level α is a critical Fα value such that the lagged x term belongs to this regression, showing that x is the cause of y.

(6) Similarly, in order to test whether y is the cause of x, the variables y and x can be replaced with each other, and steps (1) to (5) can be repeated.

As we all know, in many applications of signal processing, it is often necessary to perform frequency domain analysis, especially similar to signals with inconspicuous time domain characteristics, such as EEG signals, or the acquisition of target speed information in radar systems, sonar The analysis of noise signals in the system, and the spectral analysis in the speech processing system, etc., all involve frequency domain analysis, and frequency domain analysis is one of its most important parameter research methods. Therefore, the Granger causality is extended to the frequency domain space, so as to further study the causal relationship between variables in the frequency domain, and the results obtained are more practical.

The definition of Granger causality originally came from economics. It is a very important tool to describe the directional dynamic relationship between elements in the process of multivariate processing, so it has been applied to the research of neuroscience in recent years [18]. The earliest causal relationship is the influence in the time domain. Based on this idea, the time structure of the signal is developed, and the causal relationship of determinability is defined. In the linear structure, Granger causality is closely related to the VAR model. The details are as follows: Let, x=(x(t))t∈Z, where: x(t)=(x ₁ (t), Λ, x _n (t))′ is a stable n-dimensional sequentially. Then a short p-order VAR model VAR[p] is:

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; p</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

where a(r) is the n×n dimensional coefficient matrix of the model, and ε(t) is the multivariate Gaussian white noise. The covariance matrix of the noise process is denoted by Σ. In order to ensure the stability of the model assumptions:

det(Ia(1)z-...-a(p)z ^p )≠0 (2)

where |z| ≤ 1 for all z ∈ C.

In this model, the coefficient a _ij (r) describes the linear dependence of x _i on each element of x _j in the past. If in the autoregressive expression of (1), for all r=1, A, p, a _ij (r) are 0, then it is said that there is no Granger causality between x _j and x _i in the whole process. In other words, if applying the variables of x _j does not improve the prediction of x _i (t+1) when using past values to predict the value of x _i (t+1), then there is no Granger causality between x _j and x _i . Note that the VAR model method can only describe the linear relationship between variables, so strictly speaking, it involves linear Granger causality.

b) Bias-oriented coherence analysis

In order to obtain a frequency-domain description method for describing Granger causality, Baccala and Sameshima proposed the concept of biased coherence in 2001, making:

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&omega;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; p</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; i\&omega;r</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

It represents the difference between the n-dimensional identity matrix I and the Fourier transform of the coefficient sequence. Then the bias-oriented coherence |π _i←j (ω)| for an autoregressive process of order p is defined as follows:

$<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; \&pi;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; \&LeftArrow;</span>\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&omega;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&omega;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}{\sqrt{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; kj</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&omega;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

Condition (2) guarantees that the denominator is always positive, so this equation can well define biased coherence. It can be seen from the definition that |π _i←j (ω)| does not exist for all frequencies ω if and only if all coefficients a _ij (r) are zero, so x _j pairs do not exist Granger causality. This shows that the bias-oriented coherence |π _i←j (ω)| provides a way to measure the linear effect of orientation on pairs in the frequency domain ω. Also, since equation (4) is in normalized form, the value of the bias-oriented coherence is on [0, 1]. It compares the influence of x _j on x _i in the past with the influence of x _j on other variables in the past. Therefore, for a given signal source, bias-oriented coherence analysis can rank variables according to the strength of influence.

As shown in Figure 6(a), the three graphs on the diagonal (1, 5, 9) represent the power spectrum of the signal, the black curves in other graphs indicate that there is obvious bias towards coherence, and the dotted line represents the power spectrum of the p=0.05 Confidence level. The abscissa indicates the source, and the ordinate indicates the target. In the figure above, it can be seen that the EMG signal at the C3 lead and the C4 lead EEG signal are coherent, and the EEG signal at the C3 lead has an effect on the muscle. The coherence of the electrical signal is slightly higher than that at the C4 lead. And there is obvious coherence between EEG leads.

As shown in Figure 6(b), in the imagined action mode, the three plots on the diagonal (1, 5, 9) represent the power spectrum of the signal, and the black curves in the other plots indicate that there is a clear bias toward coherence , dashed line indicates confidence level of p = 0.05. The abscissa represents the source, and the ordinate represents the target, although this can also be shown, and it can be clearly seen that the influence of the C3 lead on the C4 lead is greater than the influence of the C4 lead on the C3 lead. It shows that in the imaginative action mode, the EEG signal is still very active.

The joint analysis method of brain electromyographic signals based on voluntary action and imaginary action of the present invention preliminarily verified the practical prospect of biased coherence in the joint analysis of brain electromyographic signals, and analyzed the electroencephalographic signals and electromyographic signals as well as the The correlation of EEG signals has achieved obvious test results, which provides a good scientific basis and application basis for further application in practical fields, such as the monitoring of sports assistance facilities and the evaluation of human exercise levels.

Claims (2)

1. one kind based on autonomous action, imagination action hypencephalon electromyographic signal conjoint analysis method, it is characterized in that, includes following steps:

1) carry out system's setting, that is, using the LabVIEW8.6 cycle that produces is 0.2 square-wave pulse signal as 10s, dutycycle;

2) carry out eeg signal acquisition and electromyographic signal collection respectively, comprising: independently move EEG signals and electromyographic signal under the mode, EEG signals and electromyographic signal under the imagination action mode;

Described autonomous action mode is specifically: the experimenter had a rest for 10 seconds; Open LabVIEW, the generation cycle is that 10s, dutycycle are 0.2 square-wave pulse, make display lamp trigger and lighted at high level, and lasting 2s, display lamp is closed 8s then, and when display lamp was bright, the experimenter did the action of right hand middle finger flexing fast; Write down 60 seconds of C3 among the brain electric conductance connection figure, C4 place EEG signals, flexor digitorum superficialis place electromyographic signal, square-wave pulse signal simultaneously;

The described imagination is moved mode specifically: the experimenter had a rest for 10 seconds; Open LabVIEW, the generation cycle is that 10s, dutycycle are 0.2 square-wave pulse, make display lamp trigger and lighted at high level, and lasting 2s, display lamp is closed 8s then, and when display lamp was bright, the experimenter did the action of right hand middle finger flexing fast; Write down C3, C4 place EEG signals, flexor digitorum superficialis (FDS) simultaneously and locate 60 seconds of electromyographic signal, synchronization pulse;

3) initial data of gathering is carried out the denoising pretreatment;

Described initial data to collection carries out the denoising pretreatment, be to use the Butterworth three rank band filters that contain 50Hz power frequency trap respectively EEG signals and electromyographic signal to be carried out Filtering Processing, according to effective frequency range feature of signal, choose EEG signals cut-off frequency: 0.5Hz and 40Hz; Electromyographic signal cut-off frequency: 0.5Hz and 200Hz afterwards, because the sample rate of initial data is very high, carry out down-sampled the processing to 512Hz to EEG signals and electromyographic signal respectively;

4) under the pretreated autonomous action mode of denoising and the brain electromyographic signal time-domain diagram under the imagination action mode independently moves under the mode and brain myoelectricity time-domain signal analysis under the imagination action mode;

5) to the brain electromyographic signal under the pretreated autonomous action of denoising, the imagination action mode carry out the time frequency signal analysis, described time frequency signal analysis is the wavelet transformation that adopts based on Morlet;

6) carrying out inclined to one side oriented phase dry analysis, specifically is to adopt the Granger causality to carry out inclined to one side oriented phase dry analysis.

2. according to claim 1 based on autonomous action, imagination action hypencephalon electromyographic signal conjoint analysis method, it is characterized in that the described LabVIEW8.6 of use of the step 1) cycle that produces is that 0.2 square-wave pulse signal comprises following process as 10s, dutycycle:

(1) sample rate and sampling configuration are set;

(2) generate the monocycle analog waveform;

(3) begin output;

(4) judge: whether acquisition time greater than 2, is to continue after the light-off to judge again, otherwise continue behind the bright lamp to judge again divided by 10 remainder.

Images