CN113807402A - System for inhibiting MIs-triggering of MI-BCI system and training and testing method thereof - Google Patents
System for inhibiting MIs-triggering of MI-BCI system and training and testing method thereof Download PDFInfo
- Publication number
- CN113807402A CN113807402A CN202110966440.2A CN202110966440A CN113807402A CN 113807402 A CN113807402 A CN 113807402A CN 202110966440 A CN202110966440 A CN 202110966440A CN 113807402 A CN113807402 A CN 113807402A
- Authority
- CN
- China
- Prior art keywords
- electroencephalogram
- unit
- bci
- test
- signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Abstract
The invention discloses a system for inhibiting false triggering of an MI-BCI system, which comprises the MI-BCI system, an electric stimulation unit, a signal acquisition unit, a feature extraction unit and a feature classification unit, wherein the electric stimulation unit is used for electrically stimulating limbs to generate somatosensory evoked potentials; the feature extraction unit inputs the electroencephalogram signal and extracts the electroencephalogram signal features by using a common space mode algorithm; the characteristic classification unit is used for inputting electroencephalogram signal characteristics, performing mode recognition classification on the extracted electroencephalogram signal characteristics by using a support vector machine, classifying the electroencephalogram signal characteristics into a target/non-target task, and feeding back a classification result to the MI-BCI system; when the EEG signals are classified into target tasks, the MI-BCI system drives the actuator to drive the affected limb to do corresponding actions, otherwise, the affected limb does not act. The invention also discloses a training and testing method of the system. The invention effectively inhibits the problem of false triggering in MI-BCI.
Description
Technical Field
The invention relates to a brain-computer interface system based on motion imagery, in particular to a system for inhibiting MIs-triggering of an MI-BCI system and a training and testing method thereof.
Background
The brain-computer interface (MI-BCI) system based on motion imagery can decode subjective motor intention of human brain through scalp brain electrical signals. In clinical application, MI-BCI technology can provide patients with intact brain but severely impaired limb motor function, such as sci (spinal cord in therapy), als (amyotrophic vascular surgery), etc., with an information pathway for directly using the brain to control mechanical peripherals, which is very helpful for improving the quality of life of these patients. Research in recent years shows that MI-BCI can also be used for rehabilitation treatment of brain injury caused by diseases such as stroke, and the novel treatment method has good rehabilitation effect in some reports, so that the potential of MI-BCI in the field of stroke rehabilitation is fully shown.
Different limb movements correspond to activation of different parts of the sensory motor cortex of the brain. The motor imagery and the motor execution have similar brain neuron activities, so that the electroencephalogram signals generated by the motor imagery can be decoded to deduce what actions the subject wants to execute, and accordingly the external device can be controlled. It was found that similar to motion performance, motion imagery can induce event-related desynchronization (ERD), i.e., energy drop in the alpha (8-13Hz) and beta (13-30Hz) frequency bands. Different motor imagination tasks are performed, and the induced brain nerve activity has certain difference, namely the cortical position where ERD occurs is different, and a specific imagination action mode can be identified by detecting the characteristic difference of the ERD.
Based on the above, the MI-BCI mainly has the following two characteristics that firstly, in the whole operation process, the MI-BCI system assumes that the brain only executes a plurality of preset action tasks; next, the system determines whether a certain motion task has occurred by determining whether ERD has occurred in the corresponding brain region of the motion. Thus, ideally, MI-BCI for stroke rehabilitation only presets the brain to perform both the imagination of limb movement and limb immobilization. When the motor intention of the apoplexy patient on the affected limb is decoded, the system drives the peripheral equipment to pull the affected limb to perform passive motion, and a closed loop of the motor intention and the motion of the affected limb is formed, so that a better nerve induction effect is formed, and a better rehabilitation effect is generated.
However, the MI-BCI system does not directly recognize motion awareness about the target limb, and it essentially does so by distinguishing the target motion from other mental tasks. The decoding of motion imagery relies primarily on ERD features of the corresponding brain regions, so in typical left and right hand recognition, if ERD features are detected on the right sensory motion region, the BCI system decides that the left hand imagery action is being attempted, and vice versa. If the right hand of a stroke patient is the affected limb, MI-BCI determines whether the patient has motor imagery on the affected limb by detecting whether an ERD of sufficient intensity has occurred in the sensory-motor cortex of the contralateral side. Unfortunately, ERD is a common phenomenon in human brain function, and not only can ERD be induced in the contralateral limb MI, but also in many other mental activities such as thinking, memory, and even actual or imaginary movements of the ipsilateral limb. Although the ERD intensity and the position of the spatial dominance distribution induced by the tasks are different, the characteristics of dispersion and low signal-to-noise ratio of the electroencephalogram characteristic spatial distribution lead a discriminator in an MI-BCI system to be difficult to judge whether the ERD of the brain area on the affected side is generated by the motor imagery of the affected limb. This creates a MIs-triggering effect of MI-BCI, i.e., some unrelated mental activities of the user's brain may cause MI-BCI to misunderstand that the user is performing a targeted task (affected limb MI) mental activity, and further cause MI-BCI to drive peripheral traction to move when the patient is not performing the motor imagery of the affected limb. Such passive movement of the affected limb, which is erroneously driven by unrelated mental activities, hardly has a positive effect on the recovery of the damaged brain area, and may even have a wrong nerve-inducing effect. Clearly, efforts should be made to avoid such false triggering in the ideal stroke rehabilitation MI-BCI.
The false triggering problem stems primarily from the low task specificity of the ERD feature, i.e., many unrelated tasks can also elicit ERD of the targeted brain region. Therefore, the introduction of electroencephalogram features with stronger task specificity is a simple idea for improving the problem of false triggering. In recent years, a new MI modulation SSSEP paradigm has received more and more attention to MI decoding problems. Research shows that the SSSEP characteristic caused by applying somatosensory stimulation with fixed frequency to a target limb part is only influenced by mental awareness of brain on motor imagery of the limb part, so that a new electroencephalogram characteristic related to motor awareness is formed, and the characteristic has better task specificity compared with an ERD characteristic. Based on the method, SSSEP characteristics of MI modulation are induced by an MI-SSSEP paradigm so as to effectively inhibit false triggering problems in MI-BCI.
Disclosure of Invention
The invention provides a system for inhibiting MIs-triggering of an MI-BCI system and a training and testing method thereof, aiming at solving the technical problems in the prior art.
The technical scheme adopted by the invention for solving the technical problems in the prior art is as follows: a system for suppressing MIs-triggering of an MI-BCI system, comprising the MI-BCI system, and further comprising: the device comprises an electrical stimulation unit, a signal acquisition unit, a feature extraction unit and a feature classification unit, wherein the electrical stimulation unit is used for electrically stimulating limbs to generate somatosensory induction potential; the characteristic extraction unit is used for inputting the electroencephalogram signals collected by the signal collection unit and extracting the characteristics of the electroencephalogram signals by adopting a common space mode algorithm; the characteristic classification unit is used for inputting the electroencephalogram signal characteristics extracted by the characteristic extraction unit, performing mode recognition classification on the extracted electroencephalogram signal characteristics by adopting a support vector machine, classifying the electroencephalogram signal characteristics into a target task and a non-target task, and feeding back a classification result to an MI-BCI system; the MI-BCI system receives the electroencephalogram signals collected by the signal collecting unit, and drives the actuator to act according to the feedback signals of the feature classification unit in the following mode: when the electroencephalogram signals are classified into target tasks, the MI-BCI system sends out signals corresponding to the electroencephalogram signals, and drives the actuator to drive the affected limb to do corresponding actions; when the electroencephalogram signals are identified as non-target tasks, the MI-BCI system does not drive the actuator to act.
Furthermore, the signal acquisition unit adopts a 64-lead electroencephalogram acquisition system produced by Neuroscan company to acquire 60-lead 0.5-100Hz electroencephalogram signals through a silver or silver chloride alloy electrode cap.
Further, the lead distribution of the electrode cap is according to the international standard 10 or 20 electrode system; wherein the reference electrode is attached to the tip of the nose and the ground electrode is attached to the forehead.
Furthermore, the sampling frequency of the signal acquisition unit is 1000Hz, and 50Hz power frequency interference is filtered.
Furthermore, the electrical stimulation unit applies stimulation to the limbs through the self-adhesive electrocardio electrode by adopting a two-phase pulse current with the pulse width of 100-200 mu s, and the stimulation frequency is 30-32 Hz.
The electroencephalogram signal pre-processing unit is used for outputting the electroencephalogram signals from the signal acquisition unit to the feature extraction unit after filtering pre-processing, spatially filtering the acquired data by adopting a common average reference method, and down-sampling the signals to 200 Hz.
Further, the feature extraction unit comprises an 8-13Hz band-pass filter, a 13-30Hz band-pass filter and a 30-32Hz band-pass filter, the feature extraction unit obtains electroencephalogram data corresponding to three frequency bands after the acquired electroencephalogram signals are subjected to band-pass filtering of the three frequency bands respectively, and feature extraction is performed on the electroencephalogram data of each frequency band respectively.
Further, the feature extraction unit calculates a CSP projection matrix for the EEG component of each frequency band, and further extracts the spatial feature of each EEG component, respectively.
Further, training and testing samples of the feature extraction unit and the feature classification unit are obtained by performing experiments on healthy subjects; during the experiment, an electrical stimulation signal is applied to the limb on one side to be tested; the electrical stimulation adopts a two-phase pulse current with the pulse width of 200 mus, and the stimulation is applied through two self-adhesive electrocardio electrodes, the stimulation frequency is 31Hz, and the stimulation intensity is adjusted until the fingers of a user slightly vibrate to generate stable and clearly visible SSSEP; setting three interference tasks as test tasks, namely limb movement imagination, limb movement execution and mental calculation tasks; collecting 4s electroencephalogram data in a task execution period, and sequentially performing 8-13Hz, 13-30Hz and 30-32Hz band-pass filtering processing to be used as a training or testing sample.
The invention also provides a training and testing method of the system for inhibiting the MIs-triggering of the MI-BCI system, which comprises the following steps:
m with proper size is selected to obtain a characteristic vector f under a certain frequency bandtrain_iThen training feature ftrain＝ [ftrain_1，ftrain_2，ftrain_3]A combination of three frequency band features;
The invention has the advantages and positive effects that: the false triggering problem stems primarily from the low task specificity of the ERD feature, i.e., many unrelated tasks can also elicit ERD of the targeted brain region. Therefore, the introduction of electroencephalogram features with stronger task specificity is a simple idea for improving the problem of false triggering. In recent years, a new MI modulation SSSEP paradigm has received more and more attention to MI decoding problems. Research shows that the SSSEP characteristic caused by applying somatosensory stimulation with fixed frequency to a target limb part is only influenced by mental consciousness of the brain on motor imagery of the limb part, so that a new electroencephalogram characteristic related to the motor consciousness is formed, and the characteristic has better task specificity compared with an ERD characteristic. Based on the method, SSSEP characteristics of MI modulation are induced by an MI-SSSEP paradigm so as to effectively inhibit false triggering problems in MI-BCI.
Drawings
FIG. 1 is a schematic diagram of an embodiment of the present invention.
Fig. 2 is an algorithmic flow chart of the present invention.
Figure 3 is a schematic of a single trial run of the present invention.
Fig. 4 is a schematic diagram of the operation of an electrical stimulation of the present invention.
Detailed Description
For further understanding of the contents, features and effects of the present invention, the following embodiments are enumerated in conjunction with the accompanying drawings, and the following detailed description is given:
the Chinese-English words and English abbreviations have the following meanings:
MI-BCI: brain-computer interface based on motor imagery.
MI: and (4) motor imagery.
BCI: a brain-computer interface.
ERD: event correlation desynchronization.
EEG: and (4) electroencephalogram.
CSP: and sharing the spatial mode.
ERSP: event-related spectral perturbations.
SSSEP: evoked potentials are felt in a steady state.
MI-SSSEP: motor imagery (brain-computer interface) based on steady state somatosensory evoked potentials.
A hang: a hanning window.
SVM is a support vector machine.
Referring to fig. 1 to 4, a system for suppressing MIs-triggering of an MI-BCI system includes the MI-BCI system, and further includes: the system comprises an electrical stimulation unit, a signal acquisition unit, a feature extraction unit and a feature classification unit, wherein the electrical stimulation unit is used for electrically stimulating limbs to generate somatosensory evoked potentials; the characteristic extraction unit is used for inputting the electroencephalogram signals collected by the signal collection unit and extracting the characteristics of the electroencephalogram signals by adopting a common space mode algorithm; the characteristic classification unit is used for inputting the electroencephalogram signal characteristics extracted by the characteristic extraction unit, performing mode recognition classification on the extracted electroencephalogram signal characteristics by adopting a support vector machine, classifying the electroencephalogram signal characteristics into a target task and a non-target task, and feeding back a classification result to an MI-BCI system; the MI-BCI system receives the electroencephalogram signals collected by the signal collecting unit, and drives the actuator to act according to the real-time classification feedback signals of the electroencephalogram signals by the characteristic classification unit in the following mode: when the electroencephalogram signals are classified into target tasks, the MI-BCI system sends out signals corresponding to the electroencephalogram signals, and drives the actuator to drive the affected limb to do corresponding actions; when the electroencephalogram signals are identified as non-target tasks, the MI-BCI system does not drive the actuator to act.
Furthermore, the signal acquisition unit can adopt a 64-lead electroencephalogram acquisition system produced by Neuroscan company, and 60-lead 0.5-100Hz electroencephalogram signals can be acquired through a silver or silver chloride alloy electrode cap.
Further, the lead distribution of the electrode cap may be in accordance with the international standard 10 or 20 electrode system; wherein the reference electrode is attached to the tip of the nose and the ground electrode is attached to the forehead.
Furthermore, the sampling frequency of the signal acquisition unit can be 1000Hz, and 50Hz power frequency interference is filtered.
Furthermore, the electrical stimulation unit adopts a two-phase pulse current with the pulse width of 100-200 mus, and can apply stimulation to limbs through the self-adhesive electrocardio electrode, and the stimulation frequency can be 30-32 Hz. The stimulation frequency may be selected to be other narrow band signals greater than 30Hz to ensure discrimination from the frequency range in which the ERD signature is located.
The electroencephalogram signal pre-processing unit can perform filtering pre-processing on the electroencephalogram signals from the signal acquisition unit and then output the electroencephalogram signals to the feature extraction unit, the common average reference method can be adopted to perform spatial filtering on acquired data, and the signals can be down-sampled to 200 Hz.
Further, the feature extraction unit may include an 8-13Hz band-pass filter, a 13-30Hz band-pass filter, and a 30-32Hz band-pass filter, and the feature extraction unit may obtain electroencephalogram data corresponding to the three frequency bands by respectively passing the acquired electroencephalogram signals through the band-pass filters of the three frequency bands, and may perform feature extraction on the electroencephalogram data of each frequency band.
Further, the feature extraction unit may calculate a CSP projection matrix for the EEG components of each frequency band, and may further extract the spatial features of each EEG component separately.
Further, training and testing samples of the feature extraction unit and the feature classification unit can be obtained through experiments on healthy subjects; during the experiment, an electrical stimulation signal can be applied to the limb on one side to be tested; the electrical stimulation can adopt a two-phase pulse current with the pulse width of 200 mus, stimulation can be applied through two self-adhesive electrocardio electrodes, the stimulation frequency can be 31Hz, and the stimulation intensity can be adjusted to slight tremor of fingers of a user until stable and clearly visible SSSEP is generated; three interference tasks can be set as test tasks, namely limb movement imagination, limb movement execution and mental calculation tasks; 4s of electroencephalogram data in a task execution period can be collected, and can be used as a training or testing sample after 8-13Hz, 13-30Hz and 30-32Hz band-pass filtering treatment in sequence.
During the experiment, the experimental process of a single round can comprise three stages, and the duration time can be 10 seconds; the first stage can be set as a preparation period, a white circle appears in the center of the screen, and the first stage lasts for 2s to remind the tested person to start the experiment and enter an experiment state; then the red circle is lightened to remind the user that the task is about to start, and the duration lasts for 2 seconds; the second stage can be set as a imagination period lasting 4 seconds, and the subject is tried to execute corresponding imagination actions according to the prompt; the third stage can be set as rest period which lasts for 2 seconds and is adjusted to prepare the next experiment; in each round, the red circle is on and electrical stimulation is applied to the subject, which reaches a maximum value over 0.5 seconds and ends at second 8; the whole experiment is completed in a quiet and non-interfering environment.
The invention also provides an embodiment of a training and testing method of the system for inhibiting the MIs-triggering of the MI-BCI system, which comprises the following steps:
m with proper size is selected to obtain a characteristic vector f under a certain frequency bandtrain_iThen training feature ftrain＝ [ftrain_1，ftrain_2，ftrain_3]A combination of three frequency band features;
The working process and working principle of the present invention are further explained by a preferred embodiment of the present invention as follows:
the present invention contemplates control experiments with both the traditional MI paradigm and the mixed MI-SSSEP paradigm for the MI-BCI system of stroke rehabilitation therapy. In MI-BCI for stroke rehabilitation, the target task is generally motor imagery of the affected limb and the non-target task is generally at rest. The system collects electroencephalogram signals, after feature extraction and classification identification, if the target task of a user is predicted to be carried out, the system drives the mechanical peripheral to pull the affected limb to do passive motion, and the process is considered to be finished with one-time triggering. The right hand is used for simulating the affected limb, and under the MI-SSSEP mixed paradigm, the affected limb is tested to perform corresponding tasks and receive electric stimulation with certain frequency on the right median nerve at the same time so as to induce SSSEP. The schematic structure is shown in fig. 1, and the electrical stimulation is only needed in a mixed paradigm.
Before the experiment, the tested subject carried out the adjustment of the electrical stimulation intensity and the electrode paste position. The electrical stimulation adopts a two-phase pulse current with the pulse width of 200 mus, and the stimulation is applied through two self-adhesive electrocardio electrodes. The stimulation frequency was 31Hz and the stimulation intensity was adjusted until the user's fingers were slightly tremored to produce a steady and clearly visible SSSEP. The stimulation intensity of each test subject was determined according to the condition of each test subject.
In order to study the MIs-triggering situation of MI-BCI, the study introduced three interference tasks as testing tasks, respectively left hand motor imagery, left hand motor performance and mental arithmetic tasks. The interference task data is not included in the modeling process and is only used for inputting the test output result in the discriminant model. An ideal MI-BCI should be triggered only by the targeting task, and any mental activities that are not targeting tasks should not trigger MI-BCI. Therefore, if an interfering task triggers MI-BCI, it should be considered a false trigger.
During the experiment, the person is tried to sit on a seat which is about 1m away from the screen, the comfortable state is kept, and the body is prevented from greatly moving as far as possible. The experimental protocol for a single run contained a total of 4 stages with a duration of 10 s. The first stage is a preparation period, a white circle appears in the center of a screen, and the beginning of the experiment of the test round is reminded for 2s continuously, and the self state needs to be adjusted; then the red circle is lightened to remind the user that the task is about to start, and the duration lasts for 2 seconds; then, a imagination period lasts for 4s, the subject is tested to execute corresponding imagination actions according to the prompt, if the prompt is 'right hand imagination', the subject is tested to carry out imagination actions of right hand fist, and if the prompt is 'rest', the subject is kept still; if the prompt is 'left hand imagination', the left hand fist making imagination action is tried to be carried out; if the prompt is 'left-hand execution', the actual motion of making a fist with the left hand is tried to be carried out; if the prompt is a mental arithmetic task, the tested brain adds two random three-digit numbers presented on the screen; finally, the rest period lasts for 2s, and the test can be slightly adjusted to prepare the next experiment. In each round of the mixed paradigm, the subject was given an electrical stimulus when the red circle was bright, which stimulus reached a maximum over 0.5s and ended at 8s, and the entire experiment was completed in a quiet, non-interfering environment.
The whole experiment is divided into 8 groups, 4 groups are traditional paradigm, 4 groups are mixed paradigm, and the random sequence is carried out. Each set contained 20 modeled samples (10 for each of the target and non-target tasks) and 15 test samples (5 for each of the interference tasks). In each paradigm, 40 samples are respectively used for the target task and the non-target task, and 20 samples are respectively used for the left-hand motor imagery, the left-hand motor execution and the mental calculation task, so that the total number of the samples is 140.
A64-lead electroencephalogram acquisition system developed by Neuroscan company is adopted to acquire 60-lead 0.5-100Hz electroencephalogram signals (CB 1, CB2, HEO and VEO leads are removed) through a silver/silver chloride (Ag/AgCl) alloy electrode cap. The sampling frequency is 1000Hz, and 50Hz power frequency interference is filtered. The lead distribution of the electrode cap is according to the international standard 10/20 electrode system. Wherein the reference electrode is attached to the tip of the nose and the ground electrode is attached to the forehead. In the pre-processing, the raw data is spatially filtered using a Common Average Reference (CAR) and the signal is down-sampled to 200 Hz.
1. Time-frequency analysis
An event-related spectral perturbation (ERSP) method is used to analyze the time-frequency domain characteristics of eeg (electroencephalography) signals, and to analyze ERD/SSSEP patterns under different imaginative tasks. The definition formula of ERSP is as follows:
wherein n represents the number of experimental runs, Fk(f, t) refers to the spectral estimation at frequency f at time t of the k-th experiment. And (3) adopting short-time Fourier transform when ERSP is calculated, setting the tuning window width to be 256 sampling points, and subtracting the frequency spectrum average value in 2s before a task from the original data so as to remove the baseline.
The method adopts Common Spatial Pattern (CSP) algorithm to extract effective electroencephalogram signal characteristics and Support Vector Machine (SVM) to carry out pattern recognition. And intercepting 4s data in the task execution period, and performing band-pass filtering on the data at 8-13Hz, 13-30Hz and 30-32Hz to effectively extract ERD characteristics and SSSEP characteristics. And then, calculating a CSP projection matrix for the EEG component of each frequency band, and further extracting the spatial feature of each EEG component. And constructing a target/non-target classifier by using a target task and a non-target task as training sets, then respectively inputting the three test tasks into the classifier, and recording the number of samples identified as the target task, wherein the proportion of the number of the samples in the target task to the total number of the samples in the class task is called the false triggering rate of the interference task.
The specific algorithm flow is shown in fig. 3. Firstly, preprocessing an original signal to obtain X, selecting task period data of each sample, and performing band-pass filtering on corresponding characteristic frequency bands to obtain XiWherein, i is 1,2 and 3 respectively corresponding to alpha, beta and SSSEP frequency bands. For each band, the training set X is dividedtrain_iAnd test set Xtest_iConstructing a CSP filter based on the training set sample to obtain a projection matrix Wi. From WiObtaining Z after spatial filteringi＝Wi TXtrain_iWherein Z isip(p ═ 1,2, …,2m) represents the filtered signal ZiM lines before (corresponding to the largest m eigenvalues) and m lines after (corresponding to the smallest m eigenvalues), the characteristic calculation formula of a single trial is
Taking m as 2, obtaining a feature vector f with the dimension of 1 x 4train_iThen f istrain＝[ftrain_1,ftrain_2,ftrain_3]The combination of the three frequency band features is a feature vector of 1 × 12. Test set characteristics ftestAnd the extraction process is similar, and the prediction result is obtained by sending the prediction result into a classifier constructed by the characteristics of the training set.
The above-mentioned embodiments are only for illustrating the technical ideas and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the contents of the present invention and to carry out the same, and the present invention should not be limited to the embodiments, i.e. equivalent variations or modifications made within the spirit of the present invention are still within the scope of the present invention.
Claims (10)
1. A system for suppressing false triggering of an MI-BCI system, comprising the MI-BCI system, characterized by further comprising: the system comprises an electrical stimulation unit, a signal acquisition unit, a feature extraction unit and a feature classification unit, wherein the electrical stimulation unit is used for electrically stimulating limbs to generate somatosensory evoked potentials; the characteristic extraction unit is used for inputting the electroencephalogram signals collected by the signal collection unit and extracting the characteristics of the electroencephalogram signals by adopting a common space mode algorithm; the characteristic classification unit is used for inputting the electroencephalogram signal characteristics extracted by the characteristic extraction unit, performing pattern recognition classification on the extracted electroencephalogram signal characteristics by adopting a support vector machine, classifying the electroencephalogram signal characteristics into a target task and a non-target task, and feeding back a classification result to an MI-BCI system; the MI-BCI system receives the electroencephalogram signals collected by the signal collecting unit, and drives the actuator to act according to the feedback signals of the feature classification unit in the following mode: when the electroencephalogram signals are classified into target tasks, the MI-BCI system sends out signals corresponding to the electroencephalogram signals, and drives the actuator to drive the affected limb to do corresponding actions; when the electroencephalogram signals are identified as non-target tasks, the MI-BCI system does not drive the actuator to act.
2. The system for inhibiting MIs-triggering of the MI-BCI system as claimed in claim 1, wherein the signal acquisition unit adopts a 64-lead electroencephalogram acquisition system manufactured by Neuroscan company to acquire 60-lead 0.5-100Hz electroencephalogram signals through a silver or silver chloride alloy electrode cap.
3. The system for suppressing MIs-triggering of an MI-BCI system of claim 1, wherein the lead profile of the electrode cap is in accordance with international standard 10 or 20 electrode system; wherein the reference electrode is attached to the tip of the nose and the ground electrode is attached to the forehead.
4. The system for suppressing MIs-triggering of an MI-BCI system of claim 1, wherein the signal acquisition unit samples at a frequency of 1000Hz and filters out 50Hz power frequency interference.
5. The system for inhibiting MIs-triggering of an MI-BCI system according to claim 1, wherein the electrical stimulation unit applies stimulation to limbs through the self-adhesive electrocardio-electrode by adopting a biphasic pulse current with the pulse width of 100-200 μ s, and the stimulation frequency is 30-32 Hz.
6. The system for suppressing MIs-triggering of an MI-BCI system as claimed in claim 1, further comprising an electroencephalogram signal preprocessing unit, wherein the electroencephalogram signal preprocessing unit performs filtering preprocessing on the electroencephalogram signal from the signal acquisition unit and outputs the processed signal to the feature extraction unit, performs spatial filtering on the acquired data by using a co-average reference method, and down-samples the signal to 200 Hz.
7. The system for suppressing MIs-triggering of an MI-BCI system as claimed in claim 1, wherein the feature extraction unit includes an 8-13Hz band pass filter, a 13-30Hz band pass filter and a 30-32Hz band pass filter, and the feature extraction unit obtains electroencephalogram data corresponding to three frequency bands by respectively subjecting the acquired electroencephalogram signals to band pass filtering of the three frequency bands, and performs feature extraction on the electroencephalogram data of each frequency band.
8. The system for suppressing MI-BCI system false triggering as claimed in claim 7, wherein the feature extraction unit calculates CSP projection matrices for EEG components of each frequency band, and further extracts spatial features of each EEG component separately.
9. The system for suppressing MIs-triggering of an MI-BCI system of claim 1, wherein training and testing samples of the feature extraction unit and the feature classification unit are obtained by experimenting with healthy subjects; during the experiment, an electrical stimulation signal is applied to the limb on one side to be tested; the electrical stimulation adopts a two-phase pulse current with the pulse width of 200 mus, and the stimulation is applied through two self-adhesive electrocardio electrodes, the stimulation frequency is 31Hz, and the stimulation intensity is adjusted until the fingers of a user slightly vibrate to generate stable and clear SSSEP; setting three interference tasks as test tasks, namely limb movement imagination, limb movement execution and mental calculation tasks; collecting 4s electroencephalogram data in a task execution period, and sequentially performing 8-13Hz, 13-30Hz and 30-32Hz band-pass filtering processing to be used as a training or testing sample.
10. A method for training and testing a system for suppressing false triggering of an MI-BCI system according to any of claims 1 to 9, comprising the steps of:
step 1, firstly, preprocessing an original signal to obtain X, selecting task period data of each sample, and performing band-pass filtering on corresponding characteristic frequency bands to obtain XiWherein, i is 1,2 and 3 respectively corresponding to alpha, beta and SSSEP frequency bands;
step 2, for each frequency band, dividing a training set Xtrain_iAnd test set Xtest_i；
Step 3, constructing a CSP filter based on the training set sample to obtain a projection matrix Wi(ii) a From WiObtaining Z after spatial filteringi＝Wi TXtrain_iIs provided with Zip(p ═ 1,2, …,2m) represents the filtered signal ZiAnd m rows before and after the center, the feature calculation formula of a single trial is as follows:
m with proper size is selected to obtain a characteristic vector f under a certain frequency bandtrain_iThen training feature ftrain＝[ftrain_1,ftrain_2,ftrain_3]A combination of three frequency band features;
step 4, for each frequency band, sending the test set into a projection matrix W constructed by the training set samplesiTo obtain ftest_iThe test sample is characterized by ftest＝[ftest_1，ftest_2，ftest_3]；
Step 5, the characteristic vector f obtained in the step 3 is processedtrainTraining a support vector machine as training data; the feature vector f obtained in the step 4testAnd sending the test data into a classifier constructed by the training set features to obtain a prediction result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110966440.2A CN113807402A (en) | 2021-08-23 | 2021-08-23 | System for inhibiting MIs-triggering of MI-BCI system and training and testing method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110966440.2A CN113807402A (en) | 2021-08-23 | 2021-08-23 | System for inhibiting MIs-triggering of MI-BCI system and training and testing method thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113807402A true CN113807402A (en) | 2021-12-17 |
Family
ID=78893862
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110966440.2A Pending CN113807402A (en) | 2021-08-23 | 2021-08-23 | System for inhibiting MIs-triggering of MI-BCI system and training and testing method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113807402A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114305450A (en) * | 2022-01-07 | 2022-04-12 | 天津大学 | Method for recognizing lower limb multi-joint motor imagery |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101711709A (en) * | 2009-12-07 | 2010-05-26 | 杭州电子科技大学 | Method for controlling electrically powered artificial hands by utilizing electro-coulogram and electroencephalogram information |
| CN106502404A (en) * | 2016-10-24 | 2017-03-15 | 天津大学 | A kind of new brain-machine interface method and system based on stable state somatosensory evoked potential |
| CN106502405A (en) * | 2016-10-24 | 2017-03-15 | 天津大学 | Based on the compound limbs Imaginary Movement multi-mode brain-computer interface method and system of synchronicity |
| CN110993056A (en) * | 2019-10-29 | 2020-04-10 | 浙江迈联医疗科技有限公司 | Hybrid active rehabilitation method and device based on mirror image neurons and brain-computer interface |
| CN111544854A (en) * | 2020-04-30 | 2020-08-18 | 天津大学 | Cerebral apoplexy motor rehabilitation method based on brain myoelectric signal deep learning fusion |
-
2021
- 2021-08-23 CN CN202110966440.2A patent/CN113807402A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101711709A (en) * | 2009-12-07 | 2010-05-26 | 杭州电子科技大学 | Method for controlling electrically powered artificial hands by utilizing electro-coulogram and electroencephalogram information |
| CN106502404A (en) * | 2016-10-24 | 2017-03-15 | 天津大学 | A kind of new brain-machine interface method and system based on stable state somatosensory evoked potential |
| CN106502405A (en) * | 2016-10-24 | 2017-03-15 | 天津大学 | Based on the compound limbs Imaginary Movement multi-mode brain-computer interface method and system of synchronicity |
| CN110993056A (en) * | 2019-10-29 | 2020-04-10 | 浙江迈联医疗科技有限公司 | Hybrid active rehabilitation method and device based on mirror image neurons and brain-computer interface |
| CN111544854A (en) * | 2020-04-30 | 2020-08-18 | 天津大学 | Cerebral apoplexy motor rehabilitation method based on brain myoelectric signal deep learning fusion |
Non-Patent Citations (1)
| Title |
|---|
| 谢松云;段绪;冯怀北;孟雅;陈刚;: "基于意识任务的机器人脑控系统", 《西北工业大学学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114305450A (en) * | 2022-01-07 | 2022-04-12 | 天津大学 | Method for recognizing lower limb multi-joint motor imagery |
| CN114305450B (en) * | 2022-01-07 | 2023-09-19 | 天津大学 | Method for identifying lower limb multi-joint motor imagery |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110765920B (en) | Motor imagery classification method based on convolutional neural network | |
| Rak et al. | Brain-computer interface as measurement and control system the review paper | |
| Parini et al. | A robust and self-paced BCI system based on a four class SSVEP paradigm: algorithms and protocols for a high-transfer-rate direct brain communication | |
| Wang et al. | Common spatial pattern method for channel selelction in motor imagery based brain-computer interface | |
| Pfurtscheller et al. | 15 years of BCI research at Graz University of Technology: current projects | |
| Pun et al. | Brain-computer interaction research at the Computer Vision and Multimedia Laboratory, University of Geneva | |
| CN111110230A (en) | Motor imagery electroencephalogram feature enhancement method and system | |
| CN112617863B (en) | Hybrid online computer-computer interface method for identifying lateral direction of left and right foot movement intention | |
| Bansal et al. | EEG-based brain-computer interfacing (BCI) | |
| Birch et al. | Brain-computer interface research at the Neil Squire Foundation | |
| CN110262658B (en) | Brain-computer interface character input system based on enhanced attention and implementation method | |
| CN107562191A (en) | The online brain-machine interface method of fine Imaginary Movement based on composite character | |
| CN113143293B (en) | Continuous speech envelope nerve entrainment extraction method based on electroencephalogram source imaging | |
| CN109657560A (en) | Mechanical arm controls online brain-computer interface system and implementation method | |
| Ferrante et al. | Data-efficient hand motor imagery decoding in EEG-BCI by using Morlet wavelets & common spatial pattern algorithms | |
| Bastos-Filho | Introduction to non-invasive EEG-Based brain-computer interfaces for assistive technologies | |
| Hasan et al. | Supervised classification of eeg signals with score threshold regulation for pseudo-online asynchronous detection of gait intention | |
| Geng et al. | A fusion algorithm for EEG signal processing based on motor imagery brain-computer interface | |
| CN113807402A (en) | System for inhibiting MIs-triggering of MI-BCI system and training and testing method thereof | |
| Li et al. | Preliminary study of online real-time control system for lower extremity exoskeletons based on EEG and sEMG fusion | |
| CN109078262B (en) | MI-BCI training method based on peripheral nerve electrical stimulation | |
| Kaur et al. | Brain computer interface: a review | |
| CN114305450B (en) | Method for identifying lower limb multi-joint motor imagery | |
| Goel | An overview of brain computer interface | |
| Khan et al. | EEG-based BCI systems for neurorehabilitation applications |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |