CN113855023A

CN113855023A - Lower limb movement BCI electrode selection method and system based on iteration tracing

Info

Publication number: CN113855023A
Application number: CN202111249220.4A
Authority: CN
Inventors: 彭小波; 宋霖; 刘俊宏
Original assignee: Shenzhen University
Current assignee: Shenzhen University
Priority date: 2021-10-26
Filing date: 2021-10-26
Publication date: 2021-12-31
Anticipated expiration: 2041-10-26
Also published as: CN113855023B

Abstract

The invention discloses a lower limb movement BCI electrode selection method and system based on iteration tracing, wherein the method comprises the following steps: s1: selecting a tracing initial electrode; s2: setting a plurality of tested electrodes, calculating and selecting a traceability time point, and tracing each tested initial electrode to obtain a traceability characteristic distribution map; s3: calculating a superposition average graph; s4: setting a pixel threshold value and processing the superposed average graph to obtain a tracing result graph; s5: superposing the tracing result graph and the electrode-brain graph to obtain an electrode selected by the feature distribution of the tracing; s6: and S5, taking the electrode obtained by tracing in the step S5 as an initial electrode for next tracing to obtain an electrode selected by a tracing result until the electrode is the same as the electrode selected by the last tracing result, ending tracing, and taking the electrode selected by the tracing result as the optimal electrode of the BCI for lower limb movement. The optimal electrode selected by the invention can extract the most distinguishing characteristic information, and the identification accuracy and the running speed of the BCI system are further improved.

Description

Lower limb movement BCI electrode selection method and system based on iteration tracing

Technical Field

The invention relates to the technical field of brain-computer interfaces, in particular to a lower limb movement BCI electrode selection method and system based on iteration tracing.

Background

With the aggravation of aging problems, various brain diseases such as cerebral apoplexy and the like are increased, which causes great inconvenience to the life of people. In addition, some accidents, such as brain injury caused by traffic accidents, can be difficult to avoid, resulting in partial or complete loss of labor. The above patients all have in common: brain function is sound but neuromuscular channels are impaired. Scientists have been keen on research in the field of brain-computer interface (BCI) in recent years in an attempt to solve this problem through BCI technology. Meanwhile, the development of the BCI technology is greatly promoted by the rapid development of the computer technology, the continuous maturity of the artificial intelligence technology and the demand in the fields of transportation, medical treatment and the like. The BCI technology is a system that does not depend on human muscle tissue and a conventional neural pathway, and creates a direct pathway between a human brain and an external device through a technical means, thereby realizing communication control between the brain and an external environment. From a macroscopic view, the research of the BCI technology is beneficial to the vigorous development of the rehabilitation medical field and the public transportation industry, and meanwhile, the research is helpful for the high-speed development of the artificial intelligence in China. On the application level, the BCI technology is widely applied to the fields of medical treatment, traffic, military, exoskeleton robots and the like, and is a development aid for various industries. However, BCI itself still has many technical problems to be solved, such as low recognition accuracy, poor real-time performance, and low response speed, and is still in the laboratory research stage at present, and is not popularized in the daily life of people. The problem of identification accuracy is a key problem of the BCI technology, and determines whether the BCI technology can be further put into practical use to a certain extent. In the process of processing signals by the BCI system, signal preprocessing, electrode selection, feature extraction and classification identification links all affect the identification accuracy of the BCI system.

The selection of the electrodes includes upper limb motor imagery electrode selection and lower limb motor imagery electrode selection. Because the motion cortical areas associated with the left and right hands in the upper limb motion have obvious discrimination in anatomy, the selected electrodes are more definite when the electrodes are selected; in the lower limb movement, the motor cortical areas associated with the left foot and the right foot are close to each other and partially overlapped in anatomy, so that the lower limb motor imagery classification difficulty is higher, and at present, a relatively perfect lower limb motor imagery electroencephalogram signal electrode selection method does not exist.

At present, most of researches on the optimal electrode for lower limb motor imagery still adopt the same method as that of the upper limb motor imagery, namely, electrodes in motion areas, which are usually C3, C4 and Cz electrodes, are directly selected. In the prior art, the publication numbers are: the CN101433460A Chinese invention patent discloses a lower limb imaginary action potential spatial filtering method in 5/20/2009, and the scheme establishes a method for performing imaginary action potential spatial filtering by wavelet packet independent component analysis aiming at imaginary action potentials induced by 3 key composite lower limb imaginary actions (imaginary standing, imaginary left-hand and same-side single-leg cooperative action, imaginary right-hand and same-side single-leg cooperative action) through C3, C4 and Cz electrodes, and meets the spatial resolution requirement of composite lower limb imaginary action potential research. Although effective electroencephalogram characteristics can be extracted to a certain extent by extracting the lower limb motor imagery electroencephalogram signals of the C3, C4 and Cz electrodes, the extraction is not the optimal characteristics.

Disclosure of Invention

The invention provides a lower limb movement BCI electrode selection method and system based on iteration tracing, aiming at overcoming the defect that the existing lower limb movement imagination electroencephalogram electrode selection method cannot obtain the optimal electrode.

The primary objective of the present invention is to solve the above technical problems, and the technical solution of the present invention is as follows:

a lower limb movement BCI electrode selection method based on iteration tracing comprises the following steps:

s1: selecting a tracing initial electrode;

s2: setting a plurality of tested electrodes, calculating a time point with the maximum amplitude in each tested time domain, taking the time point as a traceability time point, and performing traceability on each tested initial electrode to obtain a traceability feature distribution map of each tested electrode;

s3: calculating a superposition average graph by using the tested traceability feature distribution graph;

s4: setting a pixel threshold value, and processing the superposed average graph by using the pixel threshold value to obtain a tracing result graph;

s5: superposing the tracing result graph and the electrode-brain graph to obtain an electrode-brain graph containing tracing distribution, wherein the electrodes surrounded by a tracing result area are the electrodes selected by the tracing characteristic distribution;

s6: and (5) taking the electrode obtained by tracing in the step (S5) as an initial electrode for next tracing, performing next tracing to obtain an electrode selected by a tracing result until the electrode selected by the tracing result is the same as the electrode selected by the last tracing result, ending tracing, and taking the electrode selected by the tracing result as the optimal electrode of the BCI for lower limb movement.

Further, the specific process of step S1 is: respectively extracting middle 17-lead and whole-brain-lead electroencephalogram signals for all the tested subjects, carrying out short-time Fourier transform, then carrying out baseline correction on video domain signals subjected to short-time Fourier transform, respectively obtaining time-frequency graphs of the whole-brain electrodes and the middle 17-lead electrodes, wherein the electrodes with obvious ERD phenomenon in a motor imagery time period in the time-frequency graphs are initial electrodes.

Further, the time for baseline correction was 1s before the task.

Further, by setting a uniform amplitude threshold value in the traceability feature distribution map of step S2, only the region having an amplitude higher than the maximum amplitude value multiplied by the amplitude threshold value is displayed in each of the traceability feature distribution maps to be tested.

Further, in step S2, by drawing time domain contour maps of different leads, a time point with the maximum amplitude is determined as the time point of the electroencephalogram tracing.

Further, the specific process of calculating the superposition average map by using the tested traceability feature distribution map in step S3 is as follows:

setting a blank template of the whole brain area, carrying out difference processing on the traceability feature distribution map of each tested object and the blank template, and superposing and averaging all images subjected to difference processing to obtain a superposed average map.

Further, setting a pixel threshold, and processing the superposed average graph by using the pixel threshold to obtain a tracing result graph, wherein the specific process is as follows:

setting a threshold pixel value, extracting the superposition average image by using the pixel threshold value, and subtracting the extracted image by using a white template to obtain a tracing result image.

Further, the white template is an image template having a pixel value of 0.

Further, the pixel threshold is the same every time tracing.

The invention provides a lower limb movement BCI electrode selection system based on iteration tracing, which comprises: the memory comprises a lower limb movement BCI electrode selection method program based on iteration tracing, and the lower limb movement BCI electrode selection method program based on iteration tracing realizes the following steps when being executed by the processor:

s1: selecting a tracing initial electrode;

Compared with the prior art, the technical scheme of the invention has the beneficial effects that:

according to the invention, through an iteration traceability mode, the electrode selected by the previous iteration traceability result is used as the initial electrode of the next traceability until the electrodes selected by the two traceability results are the same, the traceability is finished, and the electrode selected by the current traceability result is used as the optimal electrode of the BCI for the lower limb movement.

Drawings

Fig. 1 is a flowchart of a lower limb movement BCI electrode selection method based on iterative tracing.

FIG. 2 is a drawing of an intermediate 17 guide according to an embodiment of the present invention.

FIG. 3 is a time-frequency diagram of a whole brain electrode according to an embodiment of the present invention.

FIG. 4 is a time-frequency diagram of the middle 17 conducting electrode in the embodiment of the present invention.

Fig. 5 is a tracing result diagram of motor imagery according to an embodiment of the present invention.

FIG. 6 is a time domain profile of a whole brain lead according to an embodiment of the present invention.

Fig. 7 is a first-time tracing characteristic distribution diagram of 20 tested devices according to an embodiment of the present invention.

FIG. 8 is a diagram of a blank template according to an embodiment of the present invention.

Fig. 9 is a graph illustrating the difference between the first tracing result of 20 tested pieces and the blank template according to the embodiment of the present invention.

Fig. 10 is a pixel distribution diagram of a superimposed average of 20 images after trial subtraction according to an embodiment of the present invention.

Fig. 11 is a superimposed average diagram of 20 images after trial and error according to the embodiment of the present invention.

FIG. 12 is a graph of the average of the overlap after setting the pixel threshold according to the embodiment of the present invention.

Fig. 13 is a diagram illustrating a first tracing result according to the embodiment of the present invention.

Fig. 14 is an electrode-brain diagram according to an embodiment of the present invention.

Fig. 15 is an electrode-brain diagram with an traceable distribution according to an embodiment of the present invention.

FIG. 16 is a schematic diagram of a first trace distribution of selected electrodes in accordance with an embodiment of the present invention.

Fig. 17 is a diagram illustrating a second tracing result according to the embodiment of the present invention.

FIG. 18 is a schematic diagram of a second trace distribution of selected electrodes in accordance with an embodiment of the present invention.

Fig. 19 is a diagram illustrating a third tracing result according to the embodiment of the present invention.

Fig. 20 is a schematic diagram of a third trace distribution of selected electrodes according to an embodiment of the invention.

Detailed Description

In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described herein, and therefore the scope of the present invention is not limited by the specific embodiments disclosed below.

The noun explains:

the test is as follows: refers to the subject undergoing the test or examination in a psychological test or psychological test that produces or displays the psychological phenomenon or behavioral trait being observed.

Example 1

As shown in fig. 1, a method for selecting BCI electrodes for lower limb movement based on iterative tracing includes the following steps:

s1: selecting a tracing initial electrode;

it should be noted that the iterative tracing is to select a plurality of electrodes right above the previous tracing average result to be tested for re-tracing, and after a plurality of iterations, the result area is verified and locked as the final tracing result area.

When iterative tracing is carried out, a tracing initial electrode is selected firstly, the quality of an electrode electroencephalogram signal of a middle area of a brain is often higher than that of an electrode of a peripheral area, the interference of noise such as head movement is small, and time-frequency analysis can reveal the oscillation activity related to an event by measuring the power change of different frequency bands on each electrode. In the embodiment of the invention, all tested electroencephalograms of the same task are averaged, and time-frequency analysis is carried out to obtain time-frequency spectrograms of all the tested electroencephalograms. The electroencephalogram signals of the middle 17 leads (as shown in fig. 2) and the whole brain leads are respectively extracted and subjected to short-time Fourier transform (STFT), and the electroencephalogram signals can be converted into time-frequency domain display from time sequence. Meanwhile, in order to further improve the accuracy of the signals, a baseline correction method is adopted to reduce the influence of signal drift, and the baseline correction time is 1s before a task.

As shown in fig. 3 and 4, which are time-frequency diagrams of the whole brain electrode and the middle 17 conductive electrode, it can be seen in the time-frequency diagrams that the ERD phenomenon of the middle 17 conductive electrode is more obvious than that of the whole brain electrode in the motor imagery time period, so the middle 17 conductive electrode is selected as the initial electrode for tracing.

it should be noted that the electroencephalogram signal characteristics at different times are different, so the electroencephalogram tracing results at different time points also have differences, for example, the tracing results at the resting state and the motor imagery time have a relatively large difference, the difference is reflected in the distribution of the electroencephalogram tracing characteristic diagram, the resting state tracing result distribution is often irregular, and the tracing results during lower limb movement or lower limb movement imagery are generally concentrated in the middle area of the brain, as shown in fig. 5. The color represents the current density value of the source, and the brighter the color (the larger the gray value), the more active the brain electrical signal activity at the position.

The distribution of point sources at different time and the amplitude of the source are different in the motor imagery period, the time with the maximum amplitude is determined as the time point of electroencephalogram source tracing by drawing time domain contour maps of different leads, the electroencephalogram signal is strongest at the time, and the source tracing result is more obvious. It should be noted that, since the response time of different subjects to the stimulus, that is, the time point at which the motor imagery starts, differs, the time corresponding to the maximum amplitude value also differs. As shown in fig. 6, the amplitude is maximum when t is 0.302s for a certain tested time domain profile.

In the embodiment of the present invention, 20 tested electrodes are set, a time point with the maximum amplitude in each tested time domain is respectively calculated, 17 middle conductive electrodes are obtained for each tested electrode, and each initial electrode is obtained for tracing, so as to obtain a feature distribution map of tracing of each tested electrode, as shown in fig. 7; meanwhile, by setting the uniform amplitude threshold value to be 10% in the tracing characteristic distribution map, only the area with the amplitude higher than the maximum amplitude value multiplied by 10% is displayed in each tested tracing characteristic distribution map.

it should be noted that after 20 tested traceability feature distribution maps are obtained, a superposition average map is obtained by superposition and averaging, and the specific steps are as follows:

setting a blank template of the whole brain area, carrying out difference processing on the traceability feature distribution map of each tested object and the blank template, and superposing and averaging all images subjected to difference processing to obtain a superposed average map. The blank template is to reduce the influence of the brain template during image processing, as shown in fig. 8, the source amplitudes of the blank template are all 0, the traceable feature distribution map of each tested object is subjected to subtraction with the blank template map, the subtracted image is subjected to graying processing, and the obtained image is as shown in fig. 9.

in a specific embodiment, in order to effectively reduce the feature dimension, fully extract the features and properly reduce the calculation amount of the algorithm, the number of electrodes is generally not more than 5 when the electrodes are selected in the previous period. Since 64 electrodes are distributed on the brain surface in a uniform area, 5 electrodes occupy about 7.81% of the total area of the brain surface. Fig. 10 is a pixel distribution diagram of a 20-image-after-difference-sum average diagram, in which pixel values are integrated from high to low, and the pixel distribution accounts for 7.81% of the total distribution when the pixel values are higher than 173. Then, the pixel threshold is obtained by subtracting the pixel value 173 from each pixel point in fig. 11, and the distribution of the result after subtracting the pixel value is more concentrated. The entire superimposed average map obtained by subtracting the pixel threshold is multiplied by 10, and as shown in fig. 12, the superimposed average map obtained by setting the pixel threshold is obtained. Further, in order to make the background white, the image is subtracted from a white template (the white template is an image template having a pixel value of 0), and the result is shown in fig. 13. The feature distributions selected by the pixel threshold are all parts with larger gray values, i.e. parts with larger current density of the brain and more active brain, and as can be seen from fig. 13, they are all intensively distributed in the middle area of the brain.

in one embodiment, an electrode-brain diagram is shown in FIG. 14, and an electrode-brain diagram with an tracing distribution is shown in FIG. 15. The electrodes enclosed by the traceability result area are the final electrodes, and as shown in fig. 16, the electrodes are 6 electrodes FC1, FC2, C1, Cz, C2 and CP2 in the box. The result is the selected electrode according to the feature distribution of the first tracing, and is simultaneously used as the initial electrode in the second tracing.

It should be noted that after the first tracing is finished, the electrode selected by the first tracing result is used as an initial electrode for the second tracing, and meanwhile, the pixel threshold is also consistent with that during the first tracing, where the pixel threshold is 173, as shown in fig. 17, a second tracing result graph is obtained, as shown in fig. 18, the second tracing distributes the selected electrodes, and the second tracing result is more concentrated in the middle area than the first tracing result distribution. The second trace-to-source result graph is superimposed with the electrode-brain graph to obtain the graph 18. And selecting the electrodes FC1, FC2, C1, Cz and C2 as initial electrodes in the third tracing according to the tracing result.

In the 3 rd tracing step, the same as the previous 2 times, the pixel value threshold value is 173, and the obtained final tracing result is shown in fig. 19, is more concentrated than the 2 nd tracing result in distribution, and is overlapped with the electrode-brain graph to obtain a graph 20. And selecting electrodes FC1, FC2, C1, Cz and C2 according to the tracing result, finding that the electrodes are the same as the electrodes obtained by the 2 nd tracing, ending the tracing, and ending the iteration.

In this embodiment, through the above tracing for 3 times, the optimal electrode selected as the lower limb motor imagery experiment according to the iteration result is: FC1, FC2, C1, Cz, C2.

The invention provides a lower limb movement BCI electrode selection system based on iteration tracing, which comprises: the memory comprises a lower limb movement BCI electrode selection method program based on iteration tracing, and the lower limb movement BCI electrode selection method program based on iteration tracing realizes the following steps when being executed by the processor: s1: selecting a tracing initial electrode;

Verification and analysis

In the embodiment of the invention, an SVM model classifier is established for classification recognition, wherein four-domain fusion features are extracted based on FC1, FC2, C1, Cz and C2 electrodes, and a PSO-SVM is adopted for classification recognition. The final recognition results were obtained by averaging the 10 classification recognition results of the test set, as shown in table 1. Comparative analysis shows that: (1) the longitudinal results show that the multi-domain fusion feature provided by the embodiment improves the classification accuracy compared with a single CSP feature when the same electrode and the same classification method are used. (2) The transverse result shows that when the same electrode and the same feature extraction method is used, compared with the traditional SVM algorithm, the SVM algorithm after particle swarm optimization improves the identification accuracy. (3) When the same feature extraction and classification algorithm is used, the FC1, FC2, C1, Cz and C2 electrodes have higher identification accuracy than the C3, C4 and Cz electrodes. (4) Compared with the whole table, the multi-domain fusion features are extracted based on FC1, FC2, C1, Cz and C2 electrodes, and the highest accuracy rate of 88.4274% is obtained by adopting a PSO-SVM classification recognition method, which is obviously higher than that of other methods.

TABLE 1 Classification and identification results table for different methods

As can be seen from table 1, the accuracy of extracting multi-domain fusion features based on FC1, FC2, C1, Cz, C2 electrodes and C3, C4, Cz electrodes and performing classification and identification by using PSO-SVM is 88.4274% and 87.8205%, respectively. Therefore, the recognition accuracy of the FC1, FC2, C1, Cz and C2 electrodes selected by iterative tracing is higher than that of the conventional electrodes.

Through designing an electroencephalogram experiment and acquiring electroencephalogram data, selecting five electrodes including FC1, FC2, C1, Cz and C2 based on an iterative tracing method, preprocessing the experimental data, extracting four-domain fusion characteristics of time domain, frequency domain, time frequency and space domain aiming at the five electrodes and conventional motion area electrodes C3, C4 and Cz respectively, and finally classifying and identifying the extracted characteristics by adopting a PSO-SVM classification method, the identification precision of the five electrodes selected through iterative tracing is proved to be high, namely the electrodes more suitable for the electroencephalogram experiment of lower limb motor imagery are FC1, FC2, C1, Cz and C2.

It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A lower limb movement BCI electrode selection method based on iteration tracing is characterized by comprising the following steps:

s1: selecting a tracing initial electrode;

2. The iterative traceability-based lower limb movement BCI electrode selection method as claimed in claim 1, wherein the specific process of step S1 is as follows: respectively extracting middle 17-lead and whole-brain-lead electroencephalogram signals for all the tested subjects, carrying out short-time Fourier transform, then carrying out baseline correction on video domain signals subjected to short-time Fourier transform, respectively obtaining time-frequency graphs of the whole-brain electrodes and the middle 17-lead electrodes, wherein the electrodes with obvious ERD phenomenon in a motor imagery time period in the time-frequency graphs are initial electrodes.

3. The iterative traceability-based lower limb movement BCI electrode selection method as claimed in claim 2, wherein the time for baseline correction is 1s before the task.

4. The iterative traceability-based lower limb movement BCI electrode selection method as claimed in claim 1, wherein, by setting a uniform amplitude threshold value in the traceability characteristic distribution map of step S2, only the region with the amplitude higher than the maximum amplitude value multiplied by the amplitude threshold value is displayed in each tested traceability characteristic distribution map.

5. The BCI electrode selection method based on iterative tracing of lower limb movement of claim 1, wherein in step S2, the time point with the maximum amplitude is determined as the time point of electroencephalogram tracing by drawing time domain contour maps of different leads.

6. The iterative-traceability-based lower limb movement BCI electrode selection method as claimed in claim 1, wherein the specific process of calculating the superposition average map by using the tested traceability feature distribution map in step S3 is as follows:

7. The iterative traceability-based lower limb movement BCI electrode selection method according to claim 1, wherein a pixel threshold is set, the pixel threshold is used for processing the superposition average graph to obtain a traceability result graph, and the specific process is as follows:

8. The iterative traceability-based lower limb movement BCI electrode selection method according to claim 7, wherein the white template is an image template with a pixel value of 0.

9. The iterative traceability-based lower limb movement BCI electrode selection method as claimed in claim 1, wherein the pixel threshold value is the same every time of traceability.

10. A lower limb movement BCI electrode selection system based on iterative tracing is characterized by comprising: the memory comprises a lower limb movement BCI electrode selection method program based on iteration tracing, and the lower limb movement BCI electrode selection method program based on iteration tracing realizes the following steps when being executed by the processor: s1: selecting a tracing initial electrode;