阅读说明：本技术 基于非对称多重分形去趋势相关分析脑肌电信号监测方法 (Electroencephalogram and electromyogram signal monitoring method based on asymmetric multi-fractal detrending correlation analysis ) 是由 徐光华 张凯 陈晓璧 于 2019-10-21 设计创作，主要内容包括：基于非对称多重分形去趋势相关分析脑肌电信号监测方法,先通过奇异谱分析、带通滤波和正则化,实现对脑肌电信号的预处理；然后再通过非对称多重分形去趋势相关分析,完成对脑肌电信号多重分形特征的提取和脑肌电耦合关系的解释,同时,根据多重分形特征表达运动过程中,不同阶段的脑肌电信号特征,并对比正常人与患者的特征区别,完成诊断；本发明实现了对脑肌电耦合关系的有效评估。(The method for monitoring the electroencephalogram and electromyogram signals based on the asymmetric multi-fractal detrended correlation analysis comprises the steps of firstly, preprocessing the electroencephalogram and electromyogram signals through singular spectrum analysis, band-pass filtering and regularization; then, by means of asymmetric multi-fractal detrending correlation analysis, extraction of multi-fractal features of the brain-muscle electrical signal and interpretation of the brain-muscle electrical coupling relation are completed, meanwhile, the characteristics of the brain-muscle electrical signal at different stages in the exercise process are expressed according to the multi-fractal features, and the features of a normal person and a patient are compared for distinguishing, so that diagnosis is completed; the invention realizes effective evaluation of the brain-muscle electrical coupling relation.)

1. The electroencephalogram and electromyogram signal monitoring method based on the asymmetric multi-fractal detrending correlation analysis is characterized by comprising the following steps of: preprocessing the brain and muscle electrical signals by singular spectrum analysis, band-pass filtering and regularization; and then, extracting the multi-fractal characteristics of the brain-muscle electrical signal and explaining the brain-muscle electrical coupling relation through asymmetric multi-fractal detrending correlation analysis, expressing the characteristics of the brain-muscle electrical signal at different stages in the movement process according to the multi-fractal characteristics, and comparing the characteristics of normal people and patients to finish diagnosis.

2. The method for monitoring the electroencephalogram and electromyogram signal based on the asymmetric multi-fractal detrapping correlation analysis according to claim 1, characterized by comprising the following steps:

1) designing a motion strategy, constructing a brain-muscle electrical signal monitoring system, acquiring brain-muscle electrical signals while executing the motion strategy through a subject, and transmitting the brain-muscle electrical signals to a computer for analysis in real time;

2) designing a brain electromyographic signal preprocessing method, and removing artifacts and noises in the brain electrical signals and the electromyographic signals by a singular spectrum analysis method; filtering the denoised signal to 8-45Hz by band-pass filtering; then, regularization processing is carried out on the electroencephalogram signals and the electromyogram signals, and signal data of a standard measurement interval are calculated;

3) adopting asymmetric multi-fractal detrending correlation analysis to process normalized signals, and the specific operation steps are as follows:

3.1) calculating the accumulated deviation of two time sequences of the electroencephalogram signal and the electromyogram signal, and constructing a trend outline of the signals;

in the formula, x and y respectively represent original one-dimensional time sequence signals of an electroencephalogram signal and an electromyogram signal, and mean (x (t)) and mean (y (t)) respectively represent mean values of the electroencephalogram signal and the electromyogram signal; calculating the difference value between the original signal and the signal mean value at different moments to obtain the profile of the variation trend of the electroencephalogram signal and the electromyogram signal, wherein the profile is p (x)/p (y);

3.2) dividing the trend contour signals px (j) and py (j) into k small boxes with equal length, wherein each section of signals is not overlapped,

dividing the trend outlines generated by accumulation at different moments into s small boxes with equal length, wherein,

3.3) fitting fpS by applying a polynomial of order r in each small box s of equal length_xAnd fpS_yThe fitted equation is called the box trend;

3.4) calculating the covariance of the remaining part of each box, determining the fluctuation function of each segment:

wherein, F_r(n) represents the fluctuation function of the different stages;

3.5) if the two signals analyzed are asymmetric signals, calculating a q-order fluctuation function:

wherein the content of the first and second substances,represents the slope, F_r(n) represents the fluctuation function of the different stages;

then, positive and negative fluctuation trends are calculated respectively

M⁺And M^-The number of subsequences representing positive and negative trends respectively;

3.6) to measure the degree of asymmetry of the signal cross-correlation, the following formula is used for calculation:

wherein the content of the first and second substances,

Technical Field

The invention belongs to the technical field of bioelectricity signal movement function detection, and particularly relates to a electroencephalogram and electromyographic signal monitoring method based on asymmetric multi-fractal detrending correlation analysis.

Background

In recent years, bioelectric signals have been widely used in neurological diseases and motor dysfunction-related diseases as a diagnostic means and interactive form. Brain-computer interface (BCI) is essentially a set of human-computer interface systems that enable communication between a person and external devices, and it does not rely on normal peripheral neuromuscular channels, but interacts with the outside world with electroencephalogram signals as a carrier of brain intent. The electromyographic signals are one of the most common human body bioelectricity signals, can fully reflect the motion state of muscles, are easy to acquire the interrelation and the cooperation mechanism of the electroencephalogram signals and the surface electromyographic signals, and have important research values in the research fields of motor nerve disease diagnosis and man-machine integrated intelligent rehabilitation robots.

However, electroencephalogram signals and electromyogram signals have weak signals, serious noise interference, strong non-stationarity and non-linearity, so that the traditional signal processing method is difficult to effectively extract the brain activity intention, and particularly in the limb movement behavior process, the interference of various bioelectricity signals and the complex coupling of the bioelectricity signals cause the submergence of useful signal characteristics, so that the movement intention cannot be accurately captured.

The forms and degrees of the injuries of the human body movement functions caused by brain diseases and accidental injuries are different, the clinical manifestations have obvious personalized characteristics, the limbs are used as human movement organs, the area of the cortex layer of the brain movement center occupies a large proportion, the nerve function composition is complex, and the symptomatic treatment of the personalized injuries of the limb functions is urgently needed. Therefore, according to the basic structure of the motor central nervous circuit, aiming at the functional expression of the brain motor central, peripheral nerves and limbs, the evaluation method of the brain nervous central, peripheral nerve conduction and action function of the limb function is specifically researched, and the damage analysis and the comprehensive evaluation of the different link functions of the nervous circuit are a topic which needs to be researched urgently.

Disclosure of Invention

In order to overcome the defects of the prior art, the invention aims to provide a method for monitoring the electroencephalogram and electromyogram signal based on asymmetric multi-fractal detrending correlation analysis, so that the electroencephalogram and electromyogram coupling relation can be effectively evaluated.

In order to achieve the purpose, the invention adopts the technical scheme that:

the method for monitoring the electroencephalogram and electromyogram signals based on the asymmetric multi-fractal detrended correlation analysis comprises the steps of firstly, preprocessing the electroencephalogram and electromyogram signals through singular spectrum analysis, band-pass filtering and regularization; and then, extracting the multi-fractal characteristics of the brain-muscle electrical signal and explaining the brain-muscle electrical coupling relation through asymmetric multi-fractal detrending correlation analysis, expressing the characteristics of the brain-muscle electrical signal at different stages in the movement process according to the multi-fractal characteristics, and comparing the characteristics of normal people and patients to finish diagnosis.

The method for monitoring the electroencephalogram and electromyogram signal based on the asymmetric multi-fractal detrending correlation analysis comprises the following steps:

1) designing a motion strategy, constructing a brain-muscle electrical signal monitoring system, acquiring brain-muscle electrical signals while executing the motion strategy through a subject, and transmitting the brain-muscle electrical signals to a computer for analysis in real time;

2) designing a brain electromyographic signal preprocessing method, and removing artifacts and noises in the brain electrical signals and the electromyographic signals by a singular spectrum analysis method; filtering the denoised signal to 8-45Hz by band-pass filtering; then, regularization processing is carried out on the electroencephalogram signals and the electromyogram signals, and signal data of a standard measurement interval are calculated;

3) adopting asymmetric multi-fractal detrending correlation analysis to process normalized signals, and the specific operation steps are as follows:

3.1) calculating the accumulated deviation of two time sequences of the electroencephalogram signal and the electromyogram signal, and constructing a trend outline of the signals;

in the formula, x and y respectively represent original one-dimensional time sequence signals of an electroencephalogram signal and an electromyogram signal, and mean (x (t)) and mean (y (t)) respectively represent mean values of the electroencephalogram signal and the electromyogram signal; calculating the difference value between the original signal and the signal mean value at different moments to obtain the profile of the variation trend of the electroencephalogram signal and the electromyogram signal, wherein the profile is p (x)/p (y);

3.2) dividing the trend contour signals px (j) and py (j) into k small boxes with equal length, wherein each section of signals is not overlapped,

dividing the trend outlines generated by accumulation at different moments into s small boxes with equal length, wherein,and

represents a collection of these boxes;

3.3) fitting fpS by applying a polynomial of order r in each small box s of equal length_xAnd fpS_yThe fitted equation is called the box trend;

and

respectively representing the fractal trend of the kth box of the electroencephalogram signal and the electromyogram signal at a certain moment, a and b respectively representing the parameters of a polynomial, and r being the order of the polynomial;

3.4) calculating the covariance of the remaining part of each box, determining the fluctuation function of each segment:

wherein, F_r(n) represents the fluctuation function of the different stages;

3.5) if the two signals analyzed are asymmetric signals, calculating a q-order fluctuation function:

wherein the content of the first and second substances,represents the slope, F_r(n) represents the fluctuation function of the different stages;

then, positive and negative fluctuation trends are calculated respectively

M⁺And M^-The number of subsequences representing positive and negative trends respectively;

3.6) to measure the degree of asymmetry of the signal cross-correlation, the following formula is used for calculation:

wherein the content of the first and second substances,

and

representing the scale index of the rise and fall respectively.

The invention has the beneficial effects that:

the invention aims at the characteristics of the brain electromyographic signals, applies an asymmetric multi-fractal detrending related analysis method, and effectively extracts effective brain electromyographic signal characteristics from muscle activities of specific tasks, thereby realizing effective evaluation on the brain electromyographic coupling relation, revealing the rule of the dynamics change of the brain electromyography in partial muscle movement behaviors, and providing a simple diagnosis scheme for the problem of dyskinesia caused by brain injury.

Drawings

Fig. 1 is a flow chart of a motion strategy according to an embodiment of the present invention.

Fig. 2 is a schematic diagram of a process flow of a brain-muscle electrical signal according to an embodiment of the present invention.

FIG. 3 is a diagram illustrating the effect of the present invention.

Detailed Description

The invention is described in detail below with reference to the figures and examples.

The method for monitoring the electroencephalogram and electromyogram signal based on the asymmetric multi-fractal detrending correlation analysis comprises the following steps:

1) designing a motion strategy, constructing a brain-muscle electrical signal monitoring system, acquiring brain-muscle electrical signals while executing the motion strategy through a subject, and transmitting the brain-muscle electrical signals to a computer for analysis in real time;

a subject wears an electroencephalogram cap and a myoelectricity acquisition device, sits on a chair with the forearm bent and laid on a table top, a patch of the myoelectricity acquisition device is distributed at a wrist flexor, and a pressure sensor is placed on the table top; referring to fig. 1, the exercise strategy of the present embodiment has a one-time experiment duration of 4 seconds, and the subject keeps in a motional state for 0-2 seconds, that is, the subject keeps in a relaxed state for the first 2 seconds; exerting upper arm force within 2-3 s, namely pressing the sensor by a subject with fingers for 2-3 seconds, and continuously increasing to 10N; the force is maintained within 3-4 s, namely, the force applying state is kept for 3-4 s; the method comprises the following steps that when a subject executes a task, original signals are collected through electroencephalogram collection equipment and myoelectricity collection equipment;

before the experiment, the test subject needs to be pre-trained, so as to ensure the standardization of force application and maintenance;

2) designing a brain electromyographic signal preprocessing method, and removing artifacts and noises in brain electrical signals and electromyographic signals by a singular spectrum analysis method, wherein the artifacts and the noises comprise power frequency interference, crosstalk and other irrelevant biological signals; filtering the denoised signal to 8-45Hz by band-pass filtering; then, regularization processing is carried out on the electroencephalogram and myoelectric signals, and signal data of a standard measurement interval are calculated;

referring to fig. 2, the present embodiment performs signal processing on the acquired signal according to the following steps: firstly, removing signals which have large errors and are irrelevant to a test task from acquired signals, and then carrying out denoising processing on the signals by using singular spectrum analysis, wherein the specific implementation steps of the singular spectrum are divided into the following parts: firstly, selecting a window length of 0.5s, and converting observed one-dimensional time sequence data into a multi-dimensional time sequence to obtain a track matrix; secondly, performing singular value decomposition on the converted signal to obtain a singular value of a track matrix; thirdly, on the premise that the contribution rate of the singular value meets the requirement of a threshold value, the singular value is sorted according to the contribution rate, then signal recombination is carried out aiming at the sorting result, singular value decomposition and reconstruction are realized, and through singular spectrum analysis, the effective elimination of irrelevant noise signals and interference signals can be realized; then, taking into account the characteristic distribution rule of the brain and muscle electrical signals, extracting 8-45Hz components in the denoised signals by using band-pass filtering, and in order to ensure that the efficiency of signal calculation is improved and the error is reduced, regularizing the signals and calculating the signal data of a standard measurement interval;

3) adopting asymmetric multi-fractal detrending correlation analysis to process normalized signals, and the specific operation steps are as follows:

3.1) respectively carrying out mean value calculation on the electroencephalogram signals and the electromyogram signals in the tasks, then respectively calculating the accumulated deviation of two time signals by subtracting the mean value from the original signals, and constructing the trend profile of the signals:

in the formula, x and y respectively represent original one-dimensional time sequence signals of an electroencephalogram signal and an electromyogram signal, and mean (x (t)) and mean (y (t)) respectively represent mean values of the electroencephalogram signal and the electromyogram signal; calculating the difference value between the original signal and the signal mean value at different moments to obtain the profile of the variation trend of the electroencephalogram signal and the electromyogram signal, wherein the profile is p (x)/p (y);

3.2) dividing the trend contour signals px (j) and py (j) into k small boxes with equal length on average, wherein each signal section has no overlap,

dividing the trend outlines generated by accumulation at different moments into s small boxes with equal length, wherein,

and

represents a collection of these boxes;

3.3) fitting each small box s by using a least square method and a polynomial with the motion order r, wherein the fitted equation is called a box trend;

wherein，

And

respectively representing the fractal trend of the kth box of the electroencephalogram signal and the electromyogram signal at a certain moment, a and b respectively representing the parameters of a polynomial, and r being the order of the polynomial;

is pS^rThe linear trend of (a) of (b),

3.4) calculating the covariance of the remaining part of each box by

The fractal trend can be judged, and the fluctuation function of each segment is determined:

wherein, F_r(n) represents the fluctuation function of the different stages;

3.5) if the two signals analyzed are asymmetric signals, calculating a q-order fluctuation function:

wherein the content of the first and second substances,

represents the slope, F_r(n) represents the fluctuation function of the different stages;

then, positive and negative fluctuation trends are calculated respectively

Wherein M is⁺And M^-The number of subsequences representing positive and negative trends respectively;

andby mixingAnd

is subjected to logarithmic calculation to obtain_xy(q) the positive and negative trends respectively represent the symmetry of the two signals; if q is greater than 0, the process is repeated,

and

the up or down scaling behavior of the sequence at large fluctuations can be described separately; if q is less than 0, the process is repeated,

and

the zooming behavior of the sequence under small fluctuation can be respectively described; if the two parameters are greatly influenced by q, the two analyzed signals have a multi-fractal characteristic; on the contrary, the fractal feature is not available;

3.6) if two signals have an asymmetric relationship, in order to measure the degree of asymmetry of the cross-correlation of the signals, the degree of asymmetry can be quantified by the following equation:

referring to fig. 3, which is a graph showing analysis results of a healthy subject after an experiment is performed in a standard experimental environment, it can be seen from fig. 3 that two analyzed signals have a multi-fractal characteristic, and a brain-muscle electrical signal has asymmetry in a muscle resting-exertion-maintaining task. From FIG. 3, it can be analyzed that: the electroencephalogram signal and the electromyogram signal have almost no correlation in a motionless state, and have no obvious zooming action under the fluctuation action, and the two signals are almost in a symmetrical relation;

in the stage of continuous exertion of muscle, the brain myoelectric signals have strong correlation, and the relation of dynamic coupling is concentrated on the low-frequency band of 8-26Hz, the dynamic characteristics show that the scaling behavior is obvious under large fluctuation, and the two signals have high asymmetry and multi-fractal phenomenon;

in the muscle retention maintenance stage, the brain myoelectric signals have obvious correlation in a high-frequency band of 27-45Hz, the dynamic characteristic shows that the brain myoelectric signals have obvious scaling behavior under small fluctuation, and the two signals also have high asymmetry and multi-fractal phenomenon under the task;

for patients with impaired motor function, the task signals of the muscle continuous force-applying stage and the force maintaining stage can be analyzed through the task design and signal monitoring strategy.

For a patient with normal brain function but blocked neural pathway, the electroencephalogram signals can show the dynamic change of the zooming behavior of muscle in the force stage under large fluctuation and the zooming behavior of muscle in the force maintenance stage under small fluctuation, but the electromyogram signals are still consistent with the muscle static electromyogram signals;

for patients with impaired brain function, the brain myoelectric signals of the patients are irrelevant in the whole task state, and the dynamic characteristics are consistent with the resting state.

For the patient in the convalescent period, the characteristics of the brain electromyographic signals in the force application stage are not greatly different from those of the normal person, but the characteristics of the brain electromyographic signals in the force maintenance stage may be obviously different from those of the normal person, and the rehabilitation degree of the patient can be quantified by comparing the difference.