CN109009098B - Electroencephalogram signal feature identification method under motor imagery state - Google Patents

Electroencephalogram signal feature identification method under motor imagery state Download PDF

Info

Publication number
CN109009098B
CN109009098B CN201810792231.9A CN201810792231A CN109009098B CN 109009098 B CN109009098 B CN 109009098B CN 201810792231 A CN201810792231 A CN 201810792231A CN 109009098 B CN109009098 B CN 109009098B
Authority
CN
China
Prior art keywords
classification
electroencephalogram
electrode
electroencephalogram signal
energy
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.)
Active
Application number
CN201810792231.9A
Other languages
Chinese (zh)
Other versions
CN109009098A (en
Inventor
袁艳丽
关天民
吕斌
陈志华
童小英
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dalian Jiaotong University
Original Assignee
Dalian Jiaotong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Dalian Jiaotong University filed Critical Dalian Jiaotong University
Priority to CN201810792231.9A priority Critical patent/CN109009098B/en
Publication of CN109009098A publication Critical patent/CN109009098A/en
Application granted granted Critical
Publication of CN109009098B publication Critical patent/CN109009098B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Physics & Mathematics (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Surgery (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Biomedical Technology (AREA)
  • Veterinary Medicine (AREA)
  • Pathology (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Evolutionary Computation (AREA)
  • Physiology (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Fuzzy Systems (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Psychology (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

The invention discloses an electroencephalogram signal feature recognition method under a motor imagery state, which comprises the following steps: s1: acquiring electroencephalogram signal data information of a subject under the condition of imagination movement; s2: calculating the energy spectrum of the electroencephalogram signal electrode by adopting a Welch method; s3: setting personalized information of an optimal electrode of the electroencephalogram signal, and selecting the electrode with the highest classification accuracy for different subjects; s4: extracting characteristic values of left and right imagination of the electroencephalogram signals by using a synchronization/desynchronization method; s5: and performing feature classification on the extracted electroencephalogram signals by adopting an optimal classification function, and optimizing a classification process by adopting a three-stage classification method based on a support vector machine. The method carries out feature classification on the electroencephalogram signals by calculating the energy spectrum and the feature value information of the electrodes of the electroencephalogram signals, wherein the motor imagery electroencephalogram signal feature extraction and classification technology can be used in the field of neural rehabilitation.

Description

Electroencephalogram signal feature identification method under motor imagery state
Technical Field
The invention relates to the technical field of signal processing, in particular to an electroencephalogram signal feature identification method under a motor imagery state.
Background
The motor imagery electroencephalogram feature extraction and classification technology can be used in the field of nerve rehabilitation, provides a new way for communicating with the outside for paralyzed or severely dyskinesia people, especially people with complete brain function but no movement (such as patients with diseases of multiple sclerosis, amyotrophic lateral sclerosis and the like), bypasses damaged neurons, and regains control over limbs or artificial limbs. The technology can also be used in the fields of traffic, military, leisure and entertainment and the like, and has wide application prospect. At present, the main feature extraction method of electroencephalogram signals is a time domain method, a multidimensional statistical analysis method, a frequency domain analysis method and a time domain analysis method, which are the main means for carrying out electroencephalogram analysis in the early stage, directly extract waveform features from the time domain, and obtain some important electroencephalogram time domain features such as amplitude, mean value, variance, skewness and the like through the analysis of the geometric properties of electroencephalogram waveforms; the multidimensional statistical analysis method is to extract the independent components of the source signal from the observed mixed signal, separate the noise signals of electrocardio, electro-oculogram and power frequency interference, etc. from the EEG signal, and extract the topographic map and power spectrum of the independent components as features. However, the most valuable information is often difficult to obtain by analyzing non-stationary signals by using statistical characteristics; the electroencephalogram signals with the amplitude changing along with the time are converted into a spectrogram with the electroencephalogram power changing along with the frequency, and the signal rule is revealed from the aspect of frequency domain. Can reflect the relative strength of frequency components and show the energy distribution characteristics. The method has good frequency resolution, but not good resolution in time domain, and is suitable for analysis of stationary signals. When non-stationary signals are processed, if a better classification effect is required, the electroencephalogram signals are required to be divided into data segments which are short enough to meet the requirement of stationarity.
In addition, in the prior art, the study of motor imagery electroencephalogram signals is mainly in a laboratory stage at present, and the classification accuracy is one of important evaluation indexes. The following problems mainly exist at present:
1. the way in which the signal is induced needs to be improved
The induction of the brain electrical signals requires an additional stimulation device and depends on a certain sensory pathway of a user, so the application range is limited to a certain extent. Furthermore, the contradiction that the subject is not suitable for long-term continuous use, or the subject is adaptive to relevant events or fatigued physically, so that the relevant potential is changed remains to be solved. The electroencephalogram signal of the spontaneous system completely comes from the spontaneous electroencephalogram of the subject, and external stimulation is not needed, so that the sensory nerve action of the subject is not needed, but the system has high requirements on a signal processing method and low recognition accuracy at present due to the fact that the system is non-stable, easy to be influenced by environment, emotion and the like and has insignificant characteristic representation.
2. The signal acquisition mode needs to be improved
Electroencephalogram signals belong to complex non-stationary signals, are very weak and are very easily interfered by the outside, so that the problems of designing a more reasonable experimental scheme, improving the signal-to-noise ratio and the like need to be solved. At present, electroencephalogram signals are mainly acquired through an implanted electrode and an external electrode, the external electrode can adopt the forms of sticking electrode plates one by one or wearing an electrode cap and the like, the number of used electrodes is usually large, and acquisition of electroencephalogram data cannot be completed by a subject independently. In addition, the best electrode position of each testee is different, so that individuation and accuracy of the electrode position cannot be realized at present, and the signal testing accuracy is influenced.
3. The EEG signal pattern recognition method needs to be improved
Non-stationary signals usually contain low and high frequency component signals, and features used for classification are often contained in local time-frequency information, and different time instants contain different frequency information, wherein implicit signal components often influence each other. The existing method mainly has the problems that information of non-stationary signals is difficult to analyze, the information is easy to lose, time-frequency domain resolution cannot be taken into account, analysis speed and precision cannot be taken into account, and the like.
4 the speed and accuracy of the analysis needs to be improved
Brain mental activities cannot be accurately analyzed, and the more kinds of tasks, the lower the analysis accuracy. For example, the accuracy rate of two types of thought tasks can reach more than 80%, and the accuracy rate of three types of thought tasks can only reach about 70%. This is related to the subject, the signal acquisition, and each link involved in the signal analysis process, such as whether the subject is familiar with the data acquisition requirements and acquisition process, whether there are abnormalities in physical and emotional aspects, whether the electrode position is suitable for the subject, whether the signal preprocessing, feature extraction, and feature classification algorithms are appropriate and effective, and so on.
5 system interference rejection needs to be improved
When leaving the laboratory environment, the system is interfered by the external environment and self conditions, the spontaneous electroencephalogram signals generated by the testee are constantly changed and are easily influenced by artifacts, the system not only needs to accurately judge the task execution and idle states, but also needs to quickly and accurately extract signal characteristics, the current brain-computer interface technology cannot meet the requirements, and the testee cannot flexibly and freely control peripheral equipment through a brain-computer interface in a real environment.
The above problems lead to BCI technology still being in the laboratory stage at present, requiring multiple technical intensive studies spanning multiple fields, such as signal processing algorithms, signal acquisition levels, subject selection and training, etc.
Disclosure of Invention
According to the problems in the prior art, the invention discloses an electroencephalogram signal feature identification method in a motor imagery state, which comprises the following steps:
s1: acquiring electroencephalogram signal data information of a subject under the condition of imagination movement;
s2: calculating the energy spectrum of the electroencephalogram signal electrode by adopting a Welch method;
s3: setting personalized information of an optimal electrode of the electroencephalogram signal, and selecting the electrode with the highest classification accuracy for different subjects;
s4: extracting characteristic values of left and right imagination of the electroencephalogram signals by using a synchronization/desynchronization method;
s5: and performing feature classification on the extracted electroencephalogram signals by adopting an optimal classification function, and optimizing a classification process by adopting a three-stage classification method based on a support vector machine.
Further, in S2, a specific algorithm for calculating an energy spectrum of the electroencephalogram signal electrode by using a Welch method is as follows:
setting the total length of a time sequence signal F (N) of the electroencephalogram signal as N, dividing the time sequence signal into L sections, setting the length of each section as M, overlapping a part of data between each section, and overlapping 1/3 data, then:
Figure BDA0001735188670000031
applying a window function ω (n) to each data segment, the average power spectrum of each segment of data
Figure BDA0001735188670000032
Figure BDA0001735188670000033
Wherein: u is a normalization factor, and the calculation formula is as follows:
Figure BDA0001735188670000034
further, the personalized information for setting the optimal electrode adopts the following mode:
s31: taking the calculated energy of the EEG signal electrode as a feature vector;
s32: calculating Euclidean distances among the feature vectors;
s33: and analyzing the Euclidean distance value, and selecting the electrode energy corresponding to the minimum Euclidean distance as a classification feature vector.
Further, S4 uses a synchronization/desynchronization method to extract the feature values of the left and right imagination of the electroencephalogram signal, specifically adopting the following method:
from the average power spectrum of the brain electrical signal
Figure BDA0001735188670000035
Extracting energy difference of left and right imagination electroencephalogram signals by the signals through a synchronization/desynchronization algorithm ERD/ERS, wherein the energy difference of different periods of time forms a motor imagery feature vector;
the ERD/ERS adopts the following mode:
taking a certain length of time period before an event occurs as a reference time period, taking the energy of a corresponding frequency band in the reference time period as a reference, and calculating the percentage of frequency band energy change caused by motor imagery, wherein the calculation formula is as follows:
Figure BDA0001735188670000041
wherein A represents the energy of a specific wave band when the left hand and the right hand are imagined to move, R is the energy of a wave band corresponding to A in the reference period, and when ERD is a positive number, the energy of the corresponding specific wave band in the time period is increased, namely, the event correlation synchronization; if the negative number indicates that the energy of the corresponding specific band is reduced in the time period, the event correlation desynchronization is carried out.
Further, in S5, feature classification is performed on the electroencephalogram signals based on the optimal classification function, the electrode position with the highest classification accuracy of each electroencephalogram signal sample is determined according to the optimal electrode selected in S3, the electroencephalogram signals are extracted based on the position, and the feature vectors obtained in S2 and S4 are used as input vectors of the classification function to perform classification.
By adopting the technical scheme, the electroencephalogram signal feature identification method under the motor imagery state provided by the invention can be used for carrying out feature classification on the electroencephalogram signals by calculating the energy spectrum and the feature value information of the electrodes of the electroencephalogram signals, wherein the motor imagery electroencephalogram signal feature extraction and classification technology can be used in the field of neural rehabilitation and provides a new way for communicating with the outside for paralyzed or severely dyskinesia people, especially people with complete brain functions but no movement, so that the method bypasses damaged neurons and obtains the control on limbs or artificial limbs again.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
In order to make the technical solutions and advantages of the present invention clearer, the following describes the technical solutions in the embodiments of the present invention clearly and completely with reference to the drawings in the embodiments of the present invention:
as shown in fig. 1, the electroencephalogram signal feature identification method in the motor imagery state specifically includes the following steps:
s1: and acquiring electroencephalogram signal data information of the subject under the condition of imagination movement.
S2: and calculating the energy spectrum of the electroencephalogram signal electrode by adopting a Welch method.
Setting the total length of a time sequence signal F (N) of the electroencephalogram signal as N, dividing the time sequence signal into L sections, setting the length of each section as M, overlapping a part of data between each section, and overlapping 1/3 data, then:
Figure BDA0001735188670000051
applying a window function ω (n) to each data segment, the average power spectrum of each segment of data
Figure BDA0001735188670000052
Figure BDA0001735188670000053
Wherein: u is a normalization factor, and the calculation formula is as follows:
Figure BDA0001735188670000054
s3: setting personalized information of an optimal electrode of the electroencephalogram signal, and selecting the electrode with the highest classification accuracy for different subjects;
s31: taking the calculated energy of the EEG signal electrode as a feature vector;
s32: calculating Euclidean distances among the feature vectors;
the euclidean distance is defined as follows:
let two n-dimensional vectors xi＝(xi1,xi2,,xin)TAnd xj＝(xj1,xj2,,xjn)TTwo objects are represented, respectively, with their euclidean distances:
Figure BDA0001735188670000055
s33: and analyzing the Euclidean distance value, and selecting the electrode energy corresponding to the minimum Euclidean distance as a classification feature vector.
S4: and extracting the characteristic values of the left imagination and the right imagination of the electroencephalogram signal by using a synchronization/desynchronization method.
From the average power spectrum of the brain electrical signal
Figure BDA0001735188670000056
Extracting energy difference of left and right imagination electroencephalogram signals by the signals through a synchronization/desynchronization algorithm ERD/ERS, wherein the energy difference of different periods of time forms a motor imagery feature vector;
the ERD/ERS adopts the following mode:
taking a certain length of time period before an event occurs as a reference time period, taking the energy of a corresponding frequency band in the reference time period as a reference, and calculating the percentage of frequency band energy change caused by motor imagery, wherein the calculation formula is as follows:
Figure BDA0001735188670000057
wherein A represents the energy of a specific wave band when the left hand and the right hand are imagined to move, R is the energy of a wave band corresponding to A in the reference period, and when ERD is a positive number, the energy of the corresponding specific wave band in the time period is increased, namely, the event correlation synchronization; if the negative number indicates that the energy of the corresponding specific band is reduced in the time period, the event correlation desynchronization is carried out.
S5: and performing feature classification on the extracted electroencephalogram signals by adopting an optimal classification function, and optimizing a classification process by adopting a three-stage classification method based on a support vector machine. And (4) carrying out feature classification on the electroencephalogram signals based on an optimal classification function, determining the electrode position with the highest classification accuracy of each electroencephalogram signal sample according to the optimal electrode selected in the S3, extracting the electroencephalogram signals based on the positions, and classifying the feature vectors obtained in the S2 and the S4 as input vectors of the classification function.
The classification of the motor imagery electroencephalogram signals belongs to small sample and nonlinear classification, and left and right two types of imagery results are separated through operation. Based on the classification characteristics of the motor imagery, the method selects a support vector machine method to perform characteristic classification. The support vector machine method is a typical two-class classification method, is established on the basis of a statistical learning theory and a structural risk minimization principle, and has a plurality of specific advantages in solving the problems of small sample, nonlinearity and high-dimensional pattern recognition.
Further, the following method is adopted for calculating the optimal function of the electroencephalogram signals and solving the optimal classification surface of the electroencephalogram signals:
given a training sample set { (x)i,yi) 1,2,. n }, where x isiAs an observed value, yiIs the corresponding class number and takes the value of 0 or 1. The expression of the linear discriminant function is g (x) ═ ω · x + b, where x is the sample and ω and b are the calculation parameters. In d-dimensional space, sample x ═ x1,x2,...xd) The distance from the classification plane is expressed as d ═ ωTX + b/| ω | |, where
Figure BDA0001735188670000061
The optimal classification surface equation is ω · x + b ═ 0, normalization processing is performed, i.e., g (x) ═ 1, and the classification surface capable of correctly classifying all samples must satisfy:
yi[(ω·xi)+b]when-1 is not less than 0, i is 1,2,. n (3.8), the classification interval is 2/| ω |, and the sample for which this expression holds is the support vector.
The optimal classification plane needs to be satisfied with (3.8) and to minimize (3.9).
Figure BDA0001735188670000062
Therefore, defining a Lagrange function translates the problem into minimizing the Lagrange function for w and b:
Figure BDA0001735188670000063
in the formula, alphaiIs the Lagrange coefficient corresponding to each sample.
The objective function (3.11) is established using Lagrange multiplier to convert to a dual problem:
Figure BDA0001735188670000071
assuming α is the optimal solution, then:
Figure BDA0001735188670000072
in addition, a classified threshold value b is obtained from any one of the support vectors, and b can be obtained by the following formula:
Figure BDA0001735188670000073
in the formula: (x)i,xj) Is any support vector.
Through the above calculation, the expression of the optimal classification function is obtained as follows:
Figure BDA0001735188670000074
the electroencephalogram signal presents the nonlinear characteristic, so the nonlinearity is converted into the linear problem in another high-dimensional space through the conversion from the nonlinearity to the linearity, and the optimal classification surface is obtained in the conversion space.
At this time, the expression of the optimal classification function is:
Figure BDA0001735188670000075
wherein n is the number of support vectors, K (x)iX) is a radial basis kernel function, and the expression is K (x)i·x)＝exp(-γ||xi-x||2)。
Further, the optimal parameters of the optimal classification surface are sought
The main parameters of the SVM model are a penalty factor C and a kernel function parameter gamma, and then, a cross validation method is adopted to carry out parameter optimization on the test data through a grid search method, wherein the search range is 2-5-25The step size is 0.5. The 270 sets of experimental data were divided into two parts, one part as the training set and the other part as the test set. 180 of the groups of data were randomly selected as the training set and another portion of 90 groups of data were selected as the test set. After the test is finished, the test samples are divided again, when all the test samples are tested, the group of parameters which enable the classification accuracy of the test group to be highest is selected as the optimal parameters of the SVM model, and if the parameters which enable the highest accuracy to be obtained are multiple, the group with the minimum C is selected as the optimal parameters. If there are a plurality of gamma corresponding to C, the first group of combination parameters of C and gamma is selected and searched as the optimal parameter. The method adopts the parameters of C-8 and gamma-0.5, and the parameters are applied to the optimal classification function.
And classifying based on the optimal classification function, determining the electrode position with the highest classification accuracy of each sample according to the optimal electrode selected in the S3, extracting electroencephalogram signals based on the position, and classifying by taking the obtained feature vectors in the S2 and the S4 as input vectors of the classification function.
The first two signal test moments are selected as reference time points. In experiments, it is considered that the imagination is valid only if a certain imagination lasts for a period of time (at least 1 second or more). In order to avoid the interference caused by the wrong imagination of the user, the feature vector participating in the operation each time comprises three pieces of signal information.
And (5) calculating the motor imagery signals at the three time points by adopting the method in S5, comparing and analyzing the classification result, and outputting the result with the most repetition times as a final classification result.
By optimizing electrode signals and classifying based on an improved support vector machine method, the accuracy is obviously improved compared with that of a traditional method for directly outputting results, and the average classification accuracy is improved by 24% in 270 test samples.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.

Claims (4)

1. An electroencephalogram signal feature recognition method under a motor imagery state is characterized by comprising the following steps:
s1: acquiring electroencephalogram signal data information of a subject under the condition of imagination movement;
s2: calculating the energy spectrum of the electroencephalogram signal electrode by adopting a Welch method;
s3: setting personalized information of an optimal electrode of the electroencephalogram signal, and selecting the electrode with the highest classification accuracy for different subjects;
s4: extracting characteristic values of left and right imagination of the electroencephalogram signals by using a synchronization/desynchronization method;
s5: performing feature classification on the extracted electroencephalogram signals by adopting an optimal classification function, and optimizing a classification process by adopting a three-stage classification method based on a support vector machine;
the personalized information for setting the optimal electrode adopts the following mode:
s31: taking the calculated energy of the EEG signal electrode as a feature vector;
s32: calculating Euclidean distances among the feature vectors;
s33: and analyzing the Euclidean distance value, and selecting the electrode energy corresponding to the minimum Euclidean distance as a classification feature vector.
2. The electroencephalogram signal feature recognition method under the motor imagery state according to claim 1, further characterized by: the specific algorithm for calculating the energy spectrum of the electroencephalogram signal electrode by adopting the Welch method in the S2 is as follows:
setting the total length of a time sequence signal F (N) of the electroencephalogram signal as N, dividing the time sequence signal into L sections, setting the length of each section as M, overlapping a part of data between each section, and overlapping 1/3 data, then:
Figure FDA0002951496560000011
applying a window function ω (n) to each data segment, the average power spectrum of each segment of data
Figure FDA0002951496560000012
Figure FDA0002951496560000013
Wherein: u is a normalization factor, and the calculation formula is as follows:
Figure FDA0002951496560000014
3. the electroencephalogram signal feature recognition method under the motor imagery state according to claim 1, further characterized by: in the step S4, the following method is specifically adopted to extract the feature values of the left and right imagination of the electroencephalogram signal by using a synchronization/desynchronization method:
from the average power spectrum of the brain electrical signal
Figure FDA0002951496560000021
Extracting energy difference of left and right imagination electroencephalogram signals by the signals through a synchronization/desynchronization algorithm ERD/ERS, wherein the energy difference of different periods of time forms a motor imagery feature vector;
the ERD/ERS adopts the following mode:
taking a certain length of time period before an event occurs as a reference time period, taking the energy of a corresponding frequency band in the reference time period as a reference, and calculating the percentage of frequency band energy change caused by motor imagery, wherein the calculation formula is as follows:
Figure FDA0002951496560000022
wherein A represents the energy of a specific wave band when the left hand and the right hand are imagined to move, R is the energy of a wave band corresponding to A in the reference period, and when ERD is a positive number, the energy of the corresponding specific wave band in the time period is increased, namely, the event correlation synchronization; if the negative number indicates that the energy of the corresponding specific band is reduced in the time period, the event correlation desynchronization is carried out.
4. The electroencephalogram signal feature recognition method under the motor imagery state according to claim 1, further characterized by: and in S5, feature classification is carried out on the electroencephalogram signals based on the optimal classification function, the electrode position with the highest classification accuracy of each electroencephalogram signal sample is determined according to the optimal electrode selected in S3, the electroencephalogram signals are extracted based on the positions, and the feature vectors obtained in S2 and S4 are used as input vectors of the classification function for classification.
CN201810792231.9A 2018-07-18 2018-07-18 Electroencephalogram signal feature identification method under motor imagery state Active CN109009098B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810792231.9A CN109009098B (en) 2018-07-18 2018-07-18 Electroencephalogram signal feature identification method under motor imagery state

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810792231.9A CN109009098B (en) 2018-07-18 2018-07-18 Electroencephalogram signal feature identification method under motor imagery state

Publications (2)

Publication Number Publication Date
CN109009098A CN109009098A (en) 2018-12-18
CN109009098B true CN109009098B (en) 2021-05-28

Family

ID=64642269

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810792231.9A Active CN109009098B (en) 2018-07-18 2018-07-18 Electroencephalogram signal feature identification method under motor imagery state

Country Status (1)

Country Link
CN (1) CN109009098B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110674725B (en) * 2019-09-20 2022-06-03 电子科技大学 Equipment signal type identification method based on multi-dimensional feature vector combination of detection signals
CN111543988B (en) * 2020-05-25 2021-06-08 五邑大学 Adaptive cognitive activity recognition method and device and storage medium
CN111950455B (en) * 2020-08-12 2022-03-22 重庆邮电大学 Motion imagery electroencephalogram characteristic identification method based on LFFCNN-GRU algorithm model
CN113855023B (en) * 2021-10-26 2023-07-04 深圳大学 Iterative tracing-based lower limb movement BCI electrode selection method and system

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011123059A1 (en) * 2010-03-31 2011-10-06 Agency For Science, Technology And Research Brain- computer interface system and method
CN102722728A (en) * 2012-06-11 2012-10-10 杭州电子科技大学 Motion image electroencephalogram classification method based on channel weighting supporting vector

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011123059A1 (en) * 2010-03-31 2011-10-06 Agency For Science, Technology And Research Brain- computer interface system and method
CN102722728A (en) * 2012-06-11 2012-10-10 杭州电子科技大学 Motion image electroencephalogram classification method based on channel weighting supporting vector

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
基于BCI技术的手部运动功能康复研究;郭硕达;《中国优秀硕士学位论文全文数据 医药卫生科技辑》;20160315(第03期);第33-69页 *
用于BCI设计的运动想象脑电信号分析方法;冯焕清等;《2007中国生物医学工程联合学术年会》;20070430;第1428-1432页 *
融合小波和神经网络的发动机转子***故障诊断;康志亮等;《农机化研究》;20110131(第1期);第84-91页 *

Also Published As

Publication number Publication date
CN109009098A (en) 2018-12-18

Similar Documents

Publication Publication Date Title
Wang et al. Detection analysis of epileptic EEG using a novel random forest model combined with grid search optimization
CN106886792B (en) Electroencephalogram emotion recognition method for constructing multi-classifier fusion model based on layering mechanism
CN109009098B (en) Electroencephalogram signal feature identification method under motor imagery state
Liu et al. Retracted: Human emotion recognition based on galvanic skin response signal feature selection and svm
CN110070105B (en) Electroencephalogram emotion recognition method and system based on meta-learning example rapid screening
Rivet et al. Theoretical analysis of xDAWN algorithm: application to an efficient sensor selection in a P300 BCI
CN111476158A (en) Multi-channel physiological signal somatosensory gesture recognition method based on PSO-PCA-SVM
Haider et al. Performance enhancement in P300 ERP single trial by machine learning adaptive denoising mechanism
Yudhana et al. Human emotion recognition based on EEG signal using fast fourier transform and K-Nearest neighbor
Rahman et al. A comprehensive survey of the feature extraction methods in the EEG research
Ramos-Aguilar et al. Analysis of EEG signal processing techniques based on spectrograms
CN110543831A (en) brain print identification method based on convolutional neural network
Nasehi et al. A novel effective feature selection algorithm based on S-PCA and wavelet transform features in EEG signal classification
Ma et al. A classification algorithm of an SSVEP brain-Computer interface based on CCA fusion wavelet coefficients
Dzitac et al. Identification of ERD using fuzzy inference systems for brain-computer interface
CN110338760A (en) A kind of three classification method of schizophrenia based on electroencephalogram frequency domain data
Azami et al. Automatic signal segmentation based on singular spectrum analysis and imperialist competitive algorithm
Kundu et al. P300 detection using ensemble of SVM for brain-computer interface application
Gao et al. An ICA/HHT hybrid approach for automatic ocular artifact correction
Cecotti et al. Suboptimal sensor subset evaluation in a p300 brain-computer interface
Baziyad et al. A study and performance analysis of three paradigms of wavelet coefficients combinations in three-class motor imagery based BCI
CN110516711B (en) Training set quality evaluation method of MI-BCI system and optimization method of single training sample
Álvarez-Meza et al. Feature relevance analysis supporting automatic motor imagery discrimination in EEG based BCI systems
Vahabi et al. Enhancing P300 wave of BCI systems via negentropy in adaptive wavelet denoising
Lu Human emotion recognition based on multi-channel EEG signals using LSTM neural network

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
GR01 Patent grant
GR01 Patent grant