阅读说明：本技术 长时程脑-肌电耦合的上肢运动功能训练与评测方法 (Long-time-course brain-myoelectric coupled upper limb movement function training and evaluating method ) 是由 王仲朋 明东 陈龙 刘爽 许敏鹏 何峰 于 2020-12-23 设计创作，主要内容包括：本发明涉及脑-肌电耦合技术、医疗器械领域,为提出长时程的运动神经反馈训练系统,构建长时程脑-肌电特征的上肢运动功能训练装置和评测方法,训练提升上肢运动神经功能,建立运动神经反馈训练效果的客观评测标准,进一步研究可以得到完善脑-机交互式运动神经反馈训练系统,本发明采取的技术方案是,长时程脑-肌电耦合的上肢运动功能训练与评测方法,设计用户长时程运动神经反馈训练系统范式；搭建脑-肌电同步采集装置,以同步采集脑-肌电数据；实时计算匹配神经反馈的用户脑电特征参数,进行正向激励反馈训练；评测运动神经反馈训练前后的脑-肌电耦合关系和效果差异。本发明主要应用于脑-肌电耦合场合。(The invention relates to the field of brain-myoelectricity coupling technology and medical instruments, in order to provide a long-time-range motor nerve feedback training system, construct a long-time-range brain-myoelectricity characteristic upper limb motor function training device and an evaluation method, train and promote the upper limb motor nerve function, establish an objective evaluation standard of the motor nerve feedback training effect, and further research can obtain a perfect brain-machine interactive motor nerve feedback training system; a brain-myoelectricity synchronous acquisition device is built to synchronously acquire brain-myoelectricity data; calculating user electroencephalogram characteristic parameters matched with neural feedback in real time, and performing forward excitation feedback training; evaluating the brain-myoelectricity coupling relation and effect difference before and after the motor nerve feedback training. The invention is mainly applied to the occasion of brain-myoelectricity coupling.)

1. A long-time brain-myoelectric coupled upper limb motor function training and evaluating method is characterized in that a user long-time motor nerve feedback training system paradigm is designed; a brain-myoelectricity synchronous acquisition device is built to synchronously acquire brain-myoelectricity data; calculating user electroencephalogram characteristic parameters matched with neural feedback in real time, and performing forward excitation feedback training; evaluating the brain-myoelectricity coupling relation and effect difference before and after the motor nerve feedback training.

2. The long-term brain-myoelectric coupled upper limb motor function training and evaluating method of claim 1, wherein the specific content of the paradigm for designing the long-term motor nerve feedback training system of the user comprises 5 task stages in total, and the result is used as the evaluation standard of the initial motor imagery level in the standard evaluation stage before training; then a motor nerve feedback training stage; then, performing a standard evaluation stage in training; a motor nerve feedback training stage again; and finally, in the post-training standard task stage, analyzing the electroencephalogram and electromyogram data of the user, which are acquired in the standard task stage, and taking the analyzed data as a judgment standard of the exercise function training effect of the user.

3. The long-term brain-myoelectric coupled upper limb motor function training and evaluating method of claim 1, wherein the standard evaluation task paradigm design comprises 4 partial sessions, firstly, the user performs 30 trial tries each for 1 session left-hand LHME and right-hand RHME motor execution tasks, the order is random, the tasks are used for familiarizing the system paradigm and the motor feeling, and in addition, EMG signals are synchronously collected for the user; then calibrating each 30 dials by 1 session based on a system of left-hand LHMI and right-hand RHMI motor imagery tasks, wherein the sequence is random and is used for constructing an electroencephalogram recognition model of a subsequent online motor imagery task, a green cross appears on a display of each dial, the display is used as a Cue task Cue preparation, a red arrow is shown after 1s, the hand motor imagery task in the direction of the corresponding arrow needs to be carried out by a user, the user can go to rest after 4s, and the user can relax to wait for the start of the next dial; then 2 sessions are conducted on an online left-hand and right-hand motor imagery brain-computer interface MI-BCI, each session comprises 30 dials of the left hand and the right hand, the sequence is random and used for judging the motor imagery feature level and the classification recognition performance of the user, a green cross appears on a display of each dial, the display is prepared for a prompt task, a red arrow is prompted to appear after 1s to indicate that the user needs to conduct a hand motor imagery task in the corresponding arrow direction, the length of the arrow changes along with the electroencephalogram feature of the user motor imagery, namely the corresponding ERD energy value, so that visual feedback information is provided, the recognition decision result of the dial is fed back in a sound mode after 4s, the user can enter a rest mode, and the user can relax and wait for the start of the next dial.

4. The long-term brain-myoelectric coupled upper limb motor function training and evaluating method as claimed in claim 1, wherein the brain-myoelectric data synchronous acquisition content comprises:

the user sits on back seat, apart from 75 ~ 90cm of computer display, and both hands keep comfortable gesture, can suitably have a rest for each session interval of every stage of two weeks of period, and the head is worn EEG and is gathered the headgear and train the task, and specific signal acquisition sensor includes: EEG acquisition electrodes are 15 electrode lead positions according to the international standard 10-20 system: f3, FZ, F4, FC3, FC4, C5, C3, CZ, C4, C6, CP3, CP4, P3, PZ and P4, a scalp electroencephalogram collecting electrode made of standard Ag/AgCl materials is adopted, a special electroencephalogram medium is adopted between the scalp and the electrode to ensure good conduction characteristics, the impedance is controlled to be below 10k omega in the collecting process, the nose tip is used as a reference in the collecting process, the forehead, namely the middle of the FPZ and the FZ, is used as grounding, the EMG collecting electrode is a standard electromyogram signal collecting electrode configured by special equipment, the collecting electrode is attached to the middle of the outer side of the forearm, namely the extensor muscle position of the ulnar, the distance between the two electrodes is 2cm, and the grounding electrode is attached to the far end of;

the EEG acquisition part applies a 64-lead EEG acquisition system and acquisition software thereof, data acquisition parameters are set to be a sampling rate of 1000Hz, a hardware band-pass filtering of 0.5-100 Hz and a 50Hz power frequency notch, the EMG acquisition part applies a multi-physiological signal acquisition system and a special EMG acquisition module thereof, the data acquisition parameters are set to be a sampling rate of 5000Hz, a hardware band-pass filtering of 0.5-1000 Hz and a 50Hz power frequency notch, the EEG and the EMG equipment are connected with stimulation computer hardware, and a precise time mark triggering mode is supported in the acquisition process to ensure data synchronization; all system paradigm interfaces are written and realized by adopting an MATLAB special tool box Psychtoolbox, and synchronous event codes are sent to an electroencephalogram amplifier and EMG acquisition equipment while visual prompts are presented so as to ensure data synchronism.

5. The long-term brain-myoelectric coupled upper limb motor function training and evaluating method as claimed in claim 1, wherein the real-time brain electrical data feature calculation step is as follows:

the method is characterized in that real-time electroencephalogram features are used for controlling neural feedback instruction triggering, a user can continuously train and adjust the characteristic level of self motor imagery according to real-time feedback contents to form a motor imagery mode taking somatosensory dynamics as a main factor, for Event-related desynchronized ERD/synchronized ERS (Event-related) electroencephalogram feature signals induced by the motor imagery, short-time Fourier analysis is used for current common time-frequency analysis, specifically, an observation window W (t) with limited width is used for observing a signal x (t), and then Fourier transformation is carried out on the windowed signal to obtain the method:

where W is the angular frequency, W^*(tau-t) is a complex conjugate function of W (tau-t), when an observation window with limited value length is translated along a time axis, information that the frequency spectrum distribution of the signal changes along with time can be obtained on a two-dimensional time-frequency plane, a two-dimensional time-frequency map of the electroencephalogram signal is obtained, and further the real-time change condition of the ERD energy value is obtained;

the method for calculating the characteristic energy of the left-hand and right-hand motor imagery is a relative ERD energy difference value, and the calculation formula is as follows:

wherein the content of the first and second substances,P_nfor instantaneous energy spectrum, P_relaxIs a ground state average energy spectrum, P_taskAnd if the task state average energy spectrum is obtained, correspondingly calculating real-time left and right hand motor imagery electroencephalogram nerve feedback training parameters as follows:

left hand MI: NF_eeg|left(t)＝[RP_left(relax)-RP_left(t)]/RP_left(relax) (3)

Right-hand MI: NF_eeg|right(t)＝[RP_right(relax)-RP_right(t)]/RP_right(relax) (4)

In the formula, when the parameters of the two parameters exceed the EEG characteristic 0 threshold value of the corresponding side limb relaxation state, a system is triggered to positively stimulate feedback instructions, including neural feedback modes such as visual, auditory and somatosensory stimulation.

6. The long-term brain-myoelectric coupled upper limb motor function training and evaluating method as claimed in claim 1, wherein the long-term brain-myoelectric motor function evaluating step is as follows:

assume that the time domain EEG signal sequence is represented as: x ═ X₁,x₂,x₃…,x_NThe EMG signal sequence is expressed as: y ═ Y₁,y₂,y₃…,y_NN is the length of the time sequence, and the time sequences corresponding to different characteristic frequency bands are characterized as { X }^kAnd { Y }^k1,2,3, …, m, where m is the number of EEG and EMG characteristic frequency bands of interest, and on the basis of the two sets of time series signals, the respective entropy values are first calculated, and the entropy rate of the sequence X is calculated according to the following formula:

in the case where the two time series signals are independent, another set of entropy rates is defined as follows:

in the formula, the step length of the time prediction window is set as t, the width of the discrete time window is set as n, then x_n ^kAnd y_n ^kA signal window, which is the k-th component of the two time series signals, respectively, p () is the joint probability between the two. Further calculating from Y^kTo X^kThe entropy of transfer is:

in the same way, can obtain X^kTo Y^kTransfer entropy, the magnitude of which characterizes a particular transferThe cortical muscle coupling of the specific frequency band signal component in the delivery direction is strong or weak.

7. The method for training and evaluating the motor function of the long-term brain-electromyogram coupled upper limbs according to claim 1, wherein the classical multivariate Empirical Mode decomposition (memd) method is applied to decompose the components of the EEG and EMG signals into several eigenmode functions IMF (intrinsic Mode function), each IMF component contains the local feature signals of the original signal at different scales, and then the signal components at specific frequency bands are obtained for calculating the transmission entropy, and in addition, for comparing the cortical muscle coupling relationship of different limb movements at different transmission directions, the percentage of the transmission entropy of each component to the total sum of all components is further calculated, and expressed as:

the coupling relation between the sensory motor cortex signal and the muscle movement information is calculated by the formula, the larger the value is, the stronger the coupling relation representing the information flow direction of the sensory motor cortex signal and the muscle movement information is, and further, a basic method basis for evaluating the long-term motor nerve feedback training effect is provided.

Technical Field

The invention relates to the field of brain-myoelectricity coupling technology and medical instruments, in particular to a long-time brain-myoelectricity coupling upper limb movement function training device and an evaluation method.

Background

The Biofeedback Training technology (Biofeedback Training) achieves the goal of self-regulation (1) by presenting physiological information which can be perceived by human bodies and taking the physiological information as a signal identifier to continuously train the human bodies to learn, regulate and control the inherent or specific functions of the human bodies. In recent years, biofeedback has been used for neural function training of specific groups or tasks, and improving or improving certain specific control abilities has become a hot approach for active neural training. Particularly, Neurofeedback training (NFT) is used as an active intervention means for directly training brain functions, neurophysiological signals are acquired in real time through a sensing device, rapid decoding of brain information is completed by means of advanced information processing technology, and closed-loop training of on-line feedback of brain working states and operation performance can be achieved by matching with various stimulation modes such as vision, touch and the like (a Neurofeedback training system model is shown in fig. 1).

The nerve feedback has the characteristics of simplicity, safety and convenience, and has the advantages of no wound, no stimulation, no side effect and the like, and is worthy of application and popularization in the field of artificial intelligence. It has been found that the neural feedback technique is used to regulate brain activity and simultaneously to change the individual's cognitive and behavioral functions [2], so the technique is increasingly applied to the enhancement of brain neural functions in the fields of military, aerospace, etc. In the aspect of Motor nerve training, research shows that through a Motor imagery task, a user can learn to enhance control over a body Motor cortex through real-time nerve feedback training, and induce activation of an auxiliary Motor Area (SMA), and the degree can reach the activation level of the body Motor cortex caused by real motion, so that a possibility is provided for Motor function training [3 ].

However, how to decode complex brain information and establish a complete and effective feedback loop by using a neurofeedback technology is an important challenge in intelligent control, so that a corresponding training method is formulated according to different performance indexes (including resting state activities, anatomical brain structures and personality traits) obtained by effectively extracting relevant brain features and comprehensively evaluating neural function performance and based on neurofeedback training of different populations, and further intensive research is urgently needed [4] [5 ]. In addition, in the research of the brain-computer interface, the Motor Imagery is taken as an active human-computer interaction paradigm, which is more consistent with the thinking and activity mode of the normal human brain, and after a certain degree of training, a user can carry out interaction control of an on-line Motor Imagery brain-computer interface (MI-BCI) system. In addition, the system control signal source generally uses the energy characteristics of the scalp electroencephalogram in the motor imagery process of the user, accumulation or optimization can be carried out for a long period of time to control instruction output, and the operability is high. Based on the advantages, the brain-computer interactive motor nerve feedback system method is deeply researched and developed for motor imagery, so that the human brain is more clearly understood, real human-computer interaction is realized, and the method has strong theoretical and application values [6] to [8 ].

Reference documents:

[1]Oujamaa L,Relave I,Froger J,et al.Rehabilitation of arm function after stroke.Literature review[J].Annals of physical and rehabilitation medicine,2009,52(3):269-293.

[2]Sitaram R,Ros T,Stoeckel L,et al.Closed-loop brain training:the science of neurofeedback[J].Nature Reviews Neuroscience,2017,18(2):86.

[3]Zich C,Debener S,Kranczioch C,et al.Real-time EEG feedback during simultaneous EEG–fMRI identifies the cortical signature of motor imagery[J].Neuroimage,2015,114:438-447.

[4]Young B M,Nigogosyan Z,Walton L M,et al.Changes in functional brain organization and behavioral correlations after rehabilitative therapy using a brain-computer interface[J].Frontiers in Neuroengineering,2014,7:26.

[5]Reynolds C,Osuagwu B A,Vuckovic A.Influence of motor imagination on cortical activation during functional electrical stimulation[J].Clinical Neurophysiology,2015,126(7):1360-1369.

[6]Kaiser V,Bauernfeind G,Kreilinger A,et al.Cortical effects of user training in a motor imagery based brain-computer interface measured by fNIRS and EEG.[J].Neuroimage,2014,85(1):432-444.

[7]Kaiser V,Daly I,Pichiorri F,et al.Relationship between electrical brain responses to motor imagery and motor impairment in stroke[J].Stroke,2012,43(10):2735-2740.

[8]Várkuti B,Guan C,Pan Y,et al.Resting state changes in functional connectivity correlate with movement recovery for BCI and robot-assisted upper-extremity training after stroke[J].Neurorehabilitation and Neural Repair,2013,27(1):53-62。

disclosure of Invention

In order to overcome the defects of the prior art, the invention aims to provide a long-time-course motor nerve feedback training system, and construct a long-time-course brain-myoelectric characteristic upper limb motor function training device and an evaluation method. The MI-BCI system instruction output and idea control process is visualized and sensible, the motor function of the upper limbs is improved by training, the objective evaluation standard of the motor feedback training effect is established, the key technical guarantee is hopefully provided for the novel motor function training, the brain-machine interactive motor feedback training system can be perfected by further research, and considerable social benefit and economic benefit are obtained. Therefore, the technical scheme adopted by the invention is that a long-term brain-myoelectricity coupled upper limb movement function training and evaluating method designs a user long-term movement nerve feedback training system paradigm; a brain-myoelectricity synchronous acquisition device is built to synchronously acquire brain-myoelectricity data; calculating user electroencephalogram characteristic parameters matched with neural feedback in real time, and performing forward excitation feedback training; evaluating the brain-myoelectricity coupling relation and effect difference before and after the motor nerve feedback training.

Designing specific content of a user long-time-range motor nerve feedback training system paradigm, wherein the specific content comprises 5 task stages in total, and a standard evaluation stage before training is adopted, and the result is used as an evaluation standard of an initial motor imagery level; then a motor nerve feedback training stage; then, performing a standard evaluation stage in training; a motor nerve feedback training stage again; and finally, in the post-training standard task stage, analyzing the electroencephalogram and electromyogram data of the user, which are acquired in the standard task stage, and taking the analyzed data as a judgment standard of the exercise function training effect of the user.

The standard evaluation task paradigm design comprises 4 partial sessions, and a user performs 30 trial tries for 1 session left-hand LHME and right-hand RHME motion execution tasks respectively, wherein the sequence is random, the tasks are used for familiarizing the system paradigm and motion feeling, and EMG signals are synchronously acquired for the user; then calibrating each 30 dials by 1 session based on a system of left-hand LHMI and right-hand RHMI motor imagery tasks, wherein the sequence is random and is used for constructing an electroencephalogram recognition model of a subsequent online motor imagery task, a green cross appears on a display of each dial, the display is used as a Cue task Cue preparation, a red arrow is shown after 1s, the hand motor imagery task in the direction of the corresponding arrow needs to be carried out by a user, the user can go to rest after 4s, and the user can relax to wait for the start of the next dial; then 2 sessions are conducted on an online left-hand and right-hand motor imagery brain-computer interface MI-BCI, each session comprises 30 dials of the left hand and the right hand, the sequence is random and used for judging the motor imagery feature level and the classification recognition performance of the user, a green cross appears on a display of each dial, the display is prepared for a prompt task, a red arrow is prompted to appear after 1s to indicate that the user needs to conduct a hand motor imagery task in the corresponding arrow direction, the length of the arrow changes along with the electroencephalogram feature of the user motor imagery, namely the corresponding ERD energy value, so that visual feedback information is provided, the recognition decision result of the dial is fed back in a sound mode after 4s, the user can enter a rest mode, and the user can relax and wait for the start of the next dial.

The synchronous acquisition content of the brain-myoelectricity data comprises the following steps:

the user sits on back seat, apart from 75 ~ 90cm of computer display, and both hands keep comfortable gesture, can suitably have a rest for each session interval of every stage of two weeks of period, and the head is worn EEG and is gathered the headgear and train the task, and specific signal acquisition sensor includes: EEG acquisition electrodes are 15 electrode lead positions according to the international standard 10-20 system: f3, FZ, F4, FC3, FC4, C5, C3, CZ, C4, C6, CP3, CP4, P3, PZ and P4, a scalp electroencephalogram collecting electrode made of standard Ag/AgCl materials is adopted, a special electroencephalogram medium is adopted between the scalp and the electrode to ensure good conduction characteristics, the impedance is controlled to be below 10k omega in the collecting process, the nose tip is used as a reference in the collecting process, the forehead, namely the middle of the FPZ and the FZ, is used as grounding, the EMG collecting electrode is a standard electromyogram signal collecting electrode configured by special equipment, the collecting electrode is attached to the middle of the outer side of the forearm, namely the extensor muscle position of the ulnar, the distance between the two electrodes is 2cm, and the grounding electrode is attached to the far end of;

the EEG acquisition part applies a 64-lead EEG acquisition system and acquisition software thereof, data acquisition parameters are set to be a sampling rate of 1000Hz, a hardware band-pass filtering of 0.5-100 Hz and a 50Hz power frequency notch, the EMG acquisition part applies a multi-physiological signal acquisition system and a special EMG acquisition module thereof, the data acquisition parameters are set to be a sampling rate of 5000Hz, a hardware band-pass filtering of 0.5-1000 Hz and a 50Hz power frequency notch, the EEG and the EMG equipment are connected with stimulation computer hardware, and a precise time mark triggering mode is supported in the acquisition process to ensure data synchronization; all system paradigm interfaces are written and realized by adopting an MATLAB special tool box Psychtoolbox, and synchronous event codes are sent to an electroencephalogram amplifier and EMG acquisition equipment while visual prompts are presented so as to ensure data synchronism.

Real-time electroencephalogram data feature calculation step

The real-time electroencephalogram characteristics are used for controlling the neural feedback instruction to trigger, a user can continuously train and adjust the characteristic level of self motor imagery according to real-time feedback contents to form a motor imagery mode taking somatosensory dynamics as a main part, short-time Fourier analysis is used for the Event-related desynchronized ERD/synchronized ERS (Event-related) electroencephalogram characteristic signals induced by the motor imagery, the current common time-frequency analysis is specifically obtained by firstly using an observation window W (t) with limited width to observe signals x (t) and then carrying out Fourier transformation on the windowed signals,

where W is the angular frequency, W^*(tau-t) is a complex conjugate function of W (tau-t), when an observation window with limited value length is translated along a time axis, information that the frequency spectrum distribution of the signal changes along with time can be obtained on a two-dimensional time-frequency plane, a two-dimensional time-frequency map of the electroencephalogram signal is obtained, and further the real-time change condition of the ERD energy value is obtained;

the method for calculating the characteristic energy of the left-hand and right-hand motor imagery is a relative ERD energy difference value, and the calculation formula is as follows:

wherein the content of the first and second substances,P_nfor instantaneous energy spectrum, P_relaxIs a ground state average energy spectrum, P_taskAnd if the task state average energy spectrum is obtained, correspondingly calculating real-time left and right hand motor imagery electroencephalogram nerve feedback training parameters as follows:

left hand MI: NF_eegleft(t)＝[RP_left(relax)-RP_left(t)]/RP_left(relax) (3)

Right-hand MI: NF_eegright(t)＝[RP_right(relax)-RP_right(t)]/RP_right(relax) (4)

In the formula, when the parameters of the two parameters exceed the EEG characteristic 0 threshold value of the corresponding side limb relaxation state, a system is triggered to positively stimulate feedback instructions, including neural feedback modes such as visual, auditory and somatosensory stimulation.

Long-time-course brain-myoelectric movement function evaluation method

Assume that the time domain EEG signal sequence is represented as: x ═ X₁,x₂,x₃…,x_NThe EMG signal sequence is expressed as: y ═ Y₁,y₂,y₃…,y_NN is the length of the time sequence, and the time sequences corresponding to different characteristic frequency bands are characterized as { X }^kAnd { Y }^k1,2,3, …, m, where m is the number of EEG and EMG characteristic frequency bands of interest, given the two sets of time series signals,first, respective entropy values are calculated, and the entropy rate calculation formula of the sequence X is as follows:

in the case where the two time series signals are independent, another set of entropy rates is defined as follows:

in the formula, the step length of the time prediction window is set as t, the width of the discrete time window is set as n, then x_n ^kAnd y_n ^kA signal window, which is the k-th component of the two time series signals, respectively, p () is the joint probability between the two. Further calculating from Y^kTo X^kThe entropy of transfer is:

in the same way, can obtain X^kTo Y^kAnd the magnitude of the transfer entropy value represents the cortical muscle coupling strength of the specific frequency band signal component in the specific transfer direction.

The invention applies a classical Multivariate Empirical Mode Decomposition (MEMD) method to carry out component decomposition on EEG and EMG signals, and decomposes the signals into a plurality of Intrinsic Mode Functions (IMFs), wherein each IMF component comprises local characteristic signals of the original signals in different scales. And further calculating signal components under a specific frequency band for calculating a transmission entropy value, and further calculating the percentage of each component transmission entropy value in the sum of all components for comparing the cortical muscle coupling relationship of different limb movements in different transmission directions, wherein the percentage is expressed as:

the coupling relation between the sensory motor cortex signal and the muscle movement information is calculated by the formula, the larger the value is, the stronger the coupling relation representing the information flow direction of the sensory motor cortex signal and the muscle movement information is, and further, a basic method basis for evaluating the long-term motor nerve feedback training effect is provided.

The invention has the characteristics and beneficial effects that:

the invention designs a long-term brain-myoelectricity coupled upper limb movement function training device and an evaluation method, provides a nerve feedback training mode with multi-sensory channel fusion, can dynamically improve the motor imagery correlation capability of a user in real time, overcomes the one-way control defect of the traditional MI-BCI system, optimizes the training process of the guided movement function, evaluates the movement function training effect according to the brain-myoelectricity associated coupling effect, is closer to practical application, is expected to provide a key technology for the design of a novel nerve feedback training system, and lays a foundation for the brain-machine interface associated artificial intelligence system to rapidly step into a large-scale application stage.

Description of the drawings:

FIG. 1 is a long-term motor feedback training system framework.

Fig. 2 is a long-term motor feedback training system paradigm design. In the figure:

(a) designing an overall training time course flow; (b) and designing a standard evaluation task flow.

Fig. 3 training scenario and signal acquisition setup. In the figure:

(a) schematic diagram of EEG electrode coverage area and EMG electrode position; (b) application scenarios and acquisition devices.

Detailed Description

The invention belongs to the field of biomedical engineering, and designs a long-term brain-myoelectric coupled upper limb movement function training device and an evaluation method. According to the basic principle of brain neural plasticity, the motor nerves of the human body can be functionally changed through effective training. In recent years, with the progress of technologies in the related fields of artificial intelligence and neural engineering, specific neural functions can be effectively replaced, repaired, enhanced, improved and supplemented through the combination of a specific task paradigm design and a novel BCI technology. Therefore, the invention aims to construct a long-time-course upper limb movement function evaluating device and method by applying electroencephalogram and myoelectricity coupling analysis technology.

The technical process comprises the following steps: designing a user long-time-course motor nerve feedback training system paradigm, building a brain-myoelectricity synchronous acquisition device, calculating user electroencephalogram characteristic parameters matched with nerve feedback in real time, and evaluating brain-myoelectricity coupling relation and effect difference before and after motor nerve feedback training.

The technical scheme of the invention is detailed as follows:

first, upper limb movement function training model and system device design

The overall system design of the present invention is shown in fig. 1, and the system architecture and technical process thereof include: designing a paradigm presentation of a long-term motor nerve feedback training system of a user, building an Electroencephalogram (EEG) and Electromyogram (EMG) signal data synchronous acquisition device, calculating user Electroencephalogram characteristic parameters matched with nerve feedback in real time, and evaluating the brain-Electromyogram coupling relation and effect difference of long-term motor nerve feedback training. Each system module is detailed as follows:

model presentation module of neural feedback training system

The task operation flow design of the present invention is shown in fig. 2(a), and is described by taking a training time interval as an example of two weeks (14 days in total, recorded as day 1-day 14), which includes 5 task phases. The main tasks of user system training include: participating in a pre-training standard evaluation task (day1 is carried out) as an evaluation standard of the initial motor imagery level; then participating in a motor nerve feedback training task (day 2-6, training 3 times every other day, each time for 30 min); then, performing a standard evaluation task in training (day7 is performed); participating in a motor nerve feedback training task again (day 8-13, training 3 times every other day, each time for 30 min); and finally, performing a standard task (day 14) after training, and analyzing the electroencephalogram and electromyogram data of the user, which are acquired in the process of the standard evaluation task, as an evaluation standard of the exercise function training effect of the user.

The canonical design of the standard evaluation task is shown in fig. 2(b), and comprises 4 parts (sessions) in total. Firstly, a user carries out 30 trial times (dials) of 1 session left-hand and right-hand motion execution (marked as LHME and RHME) tasks respectively, the sequence is random, the tasks are used for familiarizing a system paradigm and motion feeling, and EMG signals are synchronously acquired for the user; then 1 session is based on system calibration (30 each) of left-hand and right-hand motor imagery (LHMI and RHMI) tasks, the sequence is random, an electroencephalogram recognition model of a subsequent online motor imagery task is constructed, each trial starts to display a green cross to be prepared as a prompt task (Cue), a red arrow is prompted to appear after 1s to indicate that a user needs to perform a hand motor imagery task in the direction of the corresponding arrow, the user enters rest after 4s, and the user can relax to wait for the start of the next trial; then 2 sessions are conducted on-line left-hand and right-hand motor imagery brain-computer interfaces (MI-BCI), each session comprises 30 dials of the left hand and the right hand, the sequence is random and used for judging the motor imagery feature level and the classification recognition performance of the user, a green cross appears on a display of each dial, the display is prepared for a prompt task, a red arrow is prompted to appear after 1s, the hand motor imagery task in the corresponding arrow direction needs to be conducted by the user, the length of the arrow changes along with the motor imagery electroencephalogram features (corresponding to ERD energy values) of the user to provide visual feedback information, the recognition decision result of the dial is fed back in 4s in a voice mode, the user can enter a rest mode, and the user can relax to wait for the start of the next dial.

Brain-myoelectricity data synchronous acquisition module

A user sits on a backrest seat, the distance between the user and a computer display is 75-90 cm, the two hands keep comfortable postures, and the system application scene is shown in fig. 3. The session intervals may be appropriately rested for each phase of two weeks. The head wears an EEG acquisition headgear for a training task, with a specific signal acquisition sensor configuration as shown in fig. 3 (a). The EEG collecting electrode is characterized by comprising 15 electrode lead positions (F3, FZ, F4, FC3, FC4, C5, C3, CZ, C4, C6, CP3, CP4, P3, PZ and P4) according to an international standard 10-20 system, a scalp electroencephalogram collecting electrode made of standard Ag/AgCl materials is adopted, a special electroencephalogram medium is adopted between the scalp and the electrode to ensure good conduction characteristics, the impedance is controlled to be below 10k omega in the collecting process, the nose tip is used as a reference in the collecting process, and the forehead (between the FPZ and the FZ) is used as the ground. The EMG collecting electrode is a standard electromyographic signal collecting electrode configured by special equipment, the collecting electrode is attached to the middle of the outer side of the forearm (extensor carpi ulnaris), the distance between the two electrodes is 2cm, and the grounding electrode is attached to the far end of the ulna.

The equipment required by the system mainly relates to electroencephalogram acquisition equipment and myoelectricity acquisition equipment. The EEG acquisition part applies a 64-lead EEG acquisition system and acquisition software thereof, and data acquisition parameters are set to be 1000Hz sampling rate, 0.5-100 Hz hardware band-pass filtering and 50Hz power frequency notch. The EMG acquisition part adopts a multi-physiological signal acquisition system and a special electromyography acquisition module thereof, and data acquisition parameters are set to be 5000Hz sampling rate, 0.5-1000 Hz hardware band-pass filtering and 50Hz power frequency notch. EEG and EMG equipment are connected with stimulation computer hardware (serial port/parallel port communication mode), and accurate time mark triggering mode is supported in the collection process to guarantee data synchronization. In addition, all system paradigm interfaces are written and realized by adopting an MATLAB special tool box (Psychtoolbox), and synchronous event codes are sent to an electroencephalogram amplifier and EMG acquisition equipment while visual prompts are presented so as to ensure data synchronism.

Calculating real-time electroencephalogram data characteristic module

The real-time electroencephalogram characteristics are used for controlling the neural feedback instruction to trigger, and a user can continuously train and adjust the self motor imagery characteristic level according to the real-time feedback content to form a motor imagery mode (left and right wrist stretching actions) taking somatosensory dynamics as the leading factor. For Event-related desynchronization/synchronization (ERD/ERS) electroencephalogram characteristic signals induced by motor imagery, power spectrum time-frequency analysis is generally used, short-time fourier analysis is one of the currently common time-frequency analysis methods, and it is assumed that electroencephalogram signals have a certain degree of short-time stationarity, i.e., the frequency spectrum distribution of the signals is not changed in a limited time window. The short-time Fourier transform is obtained by first observing the signal x (t) using an observation window W (t) of finite width, and then Fourier transforming the windowed signal,

where W is the angular frequency, W^*(τ -t) is the complex conjugate of W (τ -t)A function. When the observation window with limited value length is translated along the time axis, the information of the time-varying frequency spectrum distribution of the signal can be obtained on a two-dimensional time-frequency plane, so that a two-dimensional time-frequency map of the electroencephalogram signal can be obtained, and the real-time variation condition of the ERD energy value can be further obtained.

The method for calculating the characteristic energy of the motor imagery of the left hand and the right hand by the neural feedback training system is a relative ERD energy difference value, and the calculation formula is as follows:

wherein the content of the first and second substances,P_nfor instantaneous energy spectrum, P_relaxIs the mean energy spectrum of the ground state (relaxed state of the limb), P_taskThe task state averaged energy spectrum. Correspondingly calculating real-time left and right hand motor imagery electroencephalogram nerve feedback training parameters as follows:

left hand MI: NF_eegleft(t)＝[RP_left(relax)-RP_left(t)]/RP_left(relax) (3)

Right-hand MI: NF_eegright(t)＝[RP_right(relax)-RP_right(t)]/RP_right(relax) (4)

In the formula, when the parameters of the two parameters exceed the EEG characteristic 0 threshold value of the corresponding side limb relaxation state, a system is triggered to positively stimulate a feedback instruction, and the feedback instruction can be a neural feedback mode such as visual, auditory and somatosensory stimulation.

Method for evaluating brain-myoelectric movement function of second and long term

Transfer Entropy (TE) is proposed on the basis of information entropy and can be used to characterize directional information Transfer between two time series signals, and the Transfer entropy from sequence I to sequence II represents the change of uncertainty of information I to information II, i.e. the size of information I transferred to II.

Assume that the time domain EEG signal sequence is represented as: x ═ X₁,X₂,X₃…,X_NThe EMG signal sequence is expressed as: y ═ Y₁,y₂,y₃…,y_NN is the length of time series, and the time series corresponding to different characteristic frequency bands can be further characterized as { X }^kAnd { Y }^kAnd (k ═ 1,2,3, …, m), where m is the number of EEG and EMG feature bins of interest. On the basis of the two groups of time series signals, respective entropy values (rates) are firstly calculated, and the entropy rate calculation formula of the sequence X is as follows:

in the case where the two time series signals are independent, another set of entropy rates is defined as follows:

in the formula, the step length of the time prediction window is set as t, the width of the discrete time window is set as n, then x_n ^kAnd y_n ^kA signal window, which is the k-th component of the two time series signals, respectively, p () is the joint probability between the two. Further calculating from Y^kTo X^kThe entropy of transfer is:

in the same way, can obtain X^kTo Y^kThe transfer entropy, the magnitude of which can characterize the cortical muscle coupling strength of the specific frequency band signal component in the specific transfer direction.

The invention applies a classical Multivariate Empirical Mode Decomposition (MEMD) method to carry out component Decomposition on EEG and EMG signals, and decomposes the signals into a plurality of Intrinsic Mode Functions (IMFs), wherein each IMF component contains local characteristic signals of the original signals in different scales. Further, the signal components in a specific frequency band are obtained to be used for calculating the transmission entropy, and in addition, for comparing the cortical muscle coupling relationship of different limb movements in different transmission directions, the percentage of the transmission entropy of each component in the sum of all the components is further calculated, which can be expressed as:

according to the method, the coupling relation between the sensory motor cortex signal and the muscle movement information can be calculated, the larger the value is, the stronger the coupling relation representing the information flow direction of the sensory motor cortex signal and the muscle movement information is, and further, a basic method basis for evaluating the long-term motor nerve feedback training effect is provided.

The invention designs a long-term brain-myoelectricity coupled upper limb movement function training device and an evaluation method. The invention can be used in the fields of disabled person rehabilitation, electronic entertainment, industrial control, aerospace engineering and the like, can obtain a complete neural feedback intelligent training system through further research, and is expected to obtain considerable social and economic benefits.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.