CN109993132B - A method and system for generating pattern recognition based on EEG signals - Google Patents
A method and system for generating pattern recognition based on EEG signals Download PDFInfo
- Publication number
- CN109993132B CN109993132B CN201910269588.3A CN201910269588A CN109993132B CN 109993132 B CN109993132 B CN 109993132B CN 201910269588 A CN201910269588 A CN 201910269588A CN 109993132 B CN109993132 B CN 109993132B
- Authority
- CN
- China
- Prior art keywords
- signal
- eeg
- eeg signal
- original
- hilbert
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 84
- 238000003909 pattern recognition Methods 0.000 title claims description 22
- 238000012706 support-vector machine Methods 0.000 claims abstract description 63
- 238000007781 pre-processing Methods 0.000 claims abstract description 33
- 230000008569 process Effects 0.000 claims abstract description 24
- 210000004556 brain Anatomy 0.000 claims abstract description 23
- 238000011960 computer-aided design Methods 0.000 claims abstract description 13
- 239000013598 vector Substances 0.000 claims description 77
- 238000001228 spectrum Methods 0.000 claims description 46
- 238000012880 independent component analysis Methods 0.000 claims description 33
- 238000012549 training Methods 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 17
- 238000000354 decomposition reaction Methods 0.000 claims description 16
- 238000012360 testing method Methods 0.000 claims description 12
- 238000009826 distribution Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 3
- 230000003993 interaction Effects 0.000 abstract description 12
- 230000000007 visual effect Effects 0.000 abstract description 8
- 230000002452 interceptive effect Effects 0.000 abstract description 6
- 230000000638 stimulation Effects 0.000 description 21
- 238000002474 experimental method Methods 0.000 description 12
- 238000001914 filtration Methods 0.000 description 8
- 239000011159 matrix material Substances 0.000 description 8
- 230000008859 change Effects 0.000 description 6
- 238000012935 Averaging Methods 0.000 description 4
- 108091006146 Channels Proteins 0.000 description 4
- 238000004070 electrodeposition Methods 0.000 description 4
- 241000699666 Mus <mouse, genus> Species 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000004424 eye movement Effects 0.000 description 3
- 230000010354 integration Effects 0.000 description 3
- 238000010183 spectrum analysis Methods 0.000 description 3
- 230000001131 transforming effect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 241000699670 Mus sp. Species 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- VZCCETWTMQHEPK-QNEBEIHSSA-N gamma-linolenic acid Chemical compound CCCCC\C=C/C\C=C/C\C=C/CCCCC(O)=O VZCCETWTMQHEPK-QNEBEIHSSA-N 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- 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
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
本发明公开了一种基于脑电信号的图形识别生成方法及系统，涉及人机交互技术领域。该方法包括：构建支持向量机分类器；采集基础图形视觉刺激诱发的待识别脑电数据；对待识别脑电数据进行预处理；提取脑电信号特征；将提取脑电信号特征导入构建的支持向量机分类器，得到判别结果；根据判别结果通过计算机辅助设计，生成基础图形。该方法构建了脑电特征的分类器，实现了脑电信号中图形含义的解读，并将该信息以实际图形的方式再现。该方法既可以解放双手，又能直观地将脑中图形进行呈现，实现了自然的交互过程；还省去了传统画图交互软件的学习成本和学习过程，提高了在绘图和表达上的效率。
The invention discloses a method and a system for generating graphic recognition based on electroencephalogram signals, and relates to the technical field of human-computer interaction. The method includes: constructing a support vector machine classifier; collecting EEG data to be identified induced by basic graphic visual stimuli; preprocessing the EEG data to be identified; extracting EEG signal features; importing the extracted EEG signal features into the constructed support vector machine classifier to obtain the discriminant result; according to the discriminant result, the basic graph is generated through computer-aided design. The method constructs a classifier of EEG features, realizes the interpretation of the graphic meaning in the EEG signal, and reproduces the information in the form of actual graphics. This method not only frees hands, but also intuitively presents the graphics in the brain, realizing a natural interactive process; it also saves the learning cost and learning process of traditional drawing interactive software, and improves the efficiency of drawing and expression.
Description
Technical Field
The invention relates to the technical field of human-computer interaction, in particular to a pattern recognition generation method and system based on electroencephalogram signals.
Background
The existing interaction process of the man-machine interaction mode comprises the steps of transmitting a graph in the brain of a user to a system, presenting the graph by the system, inputting the graph in the brain of the user in the interaction process, outputting the graph presented by the system, and relating the man-machine interaction mode between input and output. The existing interactive interface of the interactive process includes: keyboards and mice, pens, touch devices, gestures, and voice.
However, in the interaction process of the keyboard and the mouse, the corresponding song operation instructions and buttons need to be selected, which is an interaction mode with higher learning cost and lower operation efficiency; the pen, this kind of interactive mode needs the interaction personnel self to have certain drawing ability on the one hand, just can express the figure in the interaction personnel brain, on the other hand has the problem that efficiency is lower on the expression of interactive result.
Therefore, the conventional method for presenting the graphics in the brain of the user has the problem of low operation efficiency.
Disclosure of Invention
The invention aims to provide a pattern recognition generation method and system based on electroencephalogram signals, and solves the problem of low operation efficiency in a mode of presenting patterns in a brain of a user.
In order to achieve the purpose, the invention provides the following scheme:
a pattern recognition generation method based on electroencephalogram signals comprises the following steps:
constructing a support vector machine classifier;
acquiring an electroencephalogram signal to be identified;
preprocessing the electroencephalogram signal to be identified to obtain a preprocessed electroencephalogram signal to be identified;
extracting the feature vector of the preprocessed electroencephalogram signal to be recognized;
importing the feature vector into a constructed support vector machine classifier to obtain a judgment result;
generating a basic graph based on the electroencephalogram signal according to the judgment result;
the specific process for constructing the support vector machine classifier comprises the following steps:
acquiring an original electroencephalogram signal acquired by non-invasive electroencephalogram acquisition equipment;
preprocessing the acquired original electroencephalogram signal to obtain a preprocessed original electroencephalogram signal;
decomposing the preprocessed original electroencephalogram signal into a plurality of electroencephalogram independent components;
extracting an initial feature vector of the electroencephalogram signal through the electroencephalogram independent component;
and constructing and adjusting a support vector machine classifier according to the initial feature vector.
Optionally, the preprocessing the acquired original electroencephalogram signal includes: and identifying and eliminating artifact interference in the original electroencephalogram signal.
Optionally, the identifying and removing the artifact interference in the original electroencephalogram signal specifically includes:
decomposing the original electroencephalogram signal into a plurality of independent source signals by adopting an independent component analysis method;
and eliminating the artifact signals in the independent source signals according to the artifact characteristics.
Optionally, the decomposing the preprocessed original electroencephalogram signal into a plurality of electroencephalogram independent components specifically includes: decomposing the preprocessed original electroencephalogram signal into a plurality of electroencephalogram independent components by adopting an independent component analysis method;
the extracting of the initial feature vector of the electroencephalogram signal through the electroencephalogram independent component specifically comprises the following steps: and extracting the initial characteristic vector of the electroencephalogram signal by using Hilbert-Huang transform.
Optionally, the extracting the initial feature vector of the electroencephalogram signal by using hilbert yellow transform specifically includes:
decomposing the unsteady-state time-series signal into r signals including instantaneous frequencies through empirical mode decomposition, wherein the signals including the instantaneous frequencies are natural mode functions;
performing Hilbert transform on each natural mode function to obtain instantaneous amplitude, phase and instantaneous frequency of the signal comprising the instantaneous frequency;
all the time-frequency distributions including the instantaneous amplitudes of the instantaneous frequency signals are Hilbert spectrums, and Hilbert marginal spectrums are obtained by integrating the Hilbert spectrums;
and calculating the energy of different frequency bands of the electroencephalogram independent components through the Hilbert marginal spectrum to obtain the initial characteristic vector of the electroencephalogram signal.
Optionally, the constructing and adjusting a support vector machine classifier according to the initial feature vector specifically includes:
storing the feature vector of the electroencephalogram signal in a data set;
dividing the data set into a training set and a test set;
constructing the support vector machine classifier through the training set;
and adjusting the constructed support vector machine classifier through the test set.
Optionally, the generating a basic graph based on the electroencephalogram signal according to the discrimination result specifically includes:
and drawing a basic graph based on the electroencephalogram signal through computer aided design according to the judgment result.
A pattern recognition generation system based on electroencephalogram signals, comprising:
the support vector machine classifier building module is used for building a support vector machine classifier;
the electroencephalogram signal acquisition module is used for acquiring an electroencephalogram signal to be identified;
the electroencephalogram signal preprocessing module is used for preprocessing the electroencephalogram signal to be identified to obtain a preprocessed electroencephalogram signal to be identified;
the electroencephalogram signal feature vector extraction module is used for extracting the feature vector of the preprocessed electroencephalogram signal to be identified;
the result acquisition module is used for importing the feature vector into the constructed support vector machine classifier to obtain a judgment result;
the graph generating module is used for generating a basic graph based on the electroencephalogram signal according to the judgment result;
the support vector machine classifier building module specifically comprises:
the original electroencephalogram signal acquisition unit is used for acquiring original electroencephalogram signals acquired by non-invasive electroencephalogram acquisition equipment;
the original electroencephalogram signal preprocessing unit is used for preprocessing the acquired original electroencephalogram signal to obtain a preprocessed original electroencephalogram signal;
the decomposition preprocessing electroencephalogram signal unit is used for decomposing the preprocessed original electroencephalogram signals into a plurality of electroencephalogram independent components;
the initial feature vector extracting unit is used for extracting the initial feature vector of the electroencephalogram signal through the electroencephalogram independent component;
and the construction adjusting unit is used for constructing and adjusting the support vector machine classifier according to the initial feature vector.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a pattern recognition generation method and a system based on an electroencephalogram signal, wherein the pattern recognition generation method based on the electroencephalogram signal comprises the following steps: constructing a support vector machine classifier; acquiring electroencephalogram data to be identified induced by basic graphic visual stimulation; preprocessing electroencephalogram data to be recognized to obtain preprocessed electroencephalogram signals to be recognized; extracting electroencephalogram signal features; leading the extracted electroencephalogram signal characteristics into a constructed support vector machine classifier to obtain a judgment result; and generating a basic graph through computer aided design according to the judgment result. The method constructs a classifier of the electroencephalogram characteristics, reads the graphic meanings in the electroencephalogram signals, and reproduces the information in the actual graphic mode. The method has the following advantages: extracting the electroencephalogram characteristics of a basic graph by using Hilbert-Huang transformation, and fully reserving time-frequency information of an electroencephalogram signal; and secondly, calling the CAD to convert the graph discrimination result of the classifier into graph information, so that the graph reproduction in the human brain is more intuitively realized, and a closed loop for feeding back the design result is formed, thereby being better applied to practice. The whole process and result can more visually reflect the accuracy of electroencephalogram identification from the generation of electroencephalogram signals to basic graphs, and a good foundation is laid for the further development of electroencephalogram intention identification and reproduction technologies in the future.
Meanwhile, compared with the existing Computer Aided Design (CAD) technology for drawing by means of a mouse, a keyboard, a touch pen and the like, the pattern recognition generation method based on the electroencephalogram signals can not only liberate hands, but also visually present patterns in the brain, and realize a natural interaction process; and complicated operations such as mouse moving, instruction selecting and the like are omitted, the learning cost and the learning process of the traditional drawing interaction software are reduced, and the efficiency on drawing and expression is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
Fig. 1 is a flowchart of a pattern recognition generation method based on electroencephalogram signals according to embodiment 1 of the present invention;
fig. 2 is a system structure diagram of a pattern recognition generating system based on electroencephalogram signals provided in embodiment 2 of the present invention;
fig. 3 is a flowchart of a pattern recognition generation method based on electroencephalogram signals according to embodiment 3 of the present invention;
fig. 4 is a directed acyclic graph of the classification method provided in embodiment 3 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example 1
The embodiment provides a pattern recognition generation method based on electroencephalogram signals. Fig. 1 is a flowchart of a pattern recognition generation method based on electroencephalogram signals according to embodiment 1 of the present invention. Referring to fig. 1, a pattern recognition generation method based on electroencephalogram signals includes:
obtaining an original electroencephalogram signal collected by a non-invasive electroencephalogram collecting device. The specific process of acquiring the original electroencephalogram signal is as follows:
the raw brain electrical signals are acquired using a non-invasive brain electrical acquisition device.
The original electroencephalogram signal acquisition experiment comprises the following steps:
experimental equipment: brain electrical signal collection equipment.
External stimulation: steady-state visual stimulus for a base graphic, comprising: stimulation frequency and stimulation duration.
The experimental process comprises the following steps: (1) informing experimenters to collect attention items of experiments; (2) an experimenter wears the electroencephalogram cap to detect the connection efficiency of the electroencephalogram cap; (3) starting the experiment, wherein an experimenter watches a screen and watches a visual stimulation video; (4) recording the electroencephalogram signals of the experimenters in real time; (5) and (5) ending the experiment after the electroencephalogram signal is collected.
And preprocessing the acquired original electroencephalogram signals to obtain preprocessed original electroencephalogram signals.
Preprocessing the acquired original brain electrical signals by using a matrix factory or a matrix laboratory (Matlab), including: and (3) identifying artifact interference in the original electroencephalogram signal and removing the artifact interference.
The method for recognizing artifact interference in the original electroencephalogram signal and eliminating the artifact interference comprises the following specific steps:
an Independent Component Analysis (ICA) method is used to decompose the original brain electrical signal into multiple Independent source signals.
And eliminating the artifact signals in the independent source signals according to the artifact characteristics.
ICA is a principal component decomposition method, and the basic idea of ICA is to separate the individual components from a set of mixed observed signals. Theoretically, interference signals generated by the heartbeat, the eye movement, the myoelectricity artifact and other interference sources in the electroencephalogram signals are all generated by mutually independent information sources, so that the signals are decomposed into n independent source signals through the ICA, artifact interference is identified according to artifact characteristics, and independent component sources considered as artifacts are removed by the ICA. Wherein n is a positive integer.
Preprocessing the acquired original electroencephalogram signal by using Matlab, and before identifying artifact interference in the original electroencephalogram signal and eliminating the artifact interference, further comprising: (1) leading in an electroencephalogram signal; (2) positioning an electrode position of the acquisition equipment; (3) re-referencing; (4) filtering; (5) segmenting the electroencephalogram signal according to the stimulation event; (6) reject bad lead signals/delete bad segments; (7) and (4) superposing and averaging the electroencephalogram signal segments under the same type of stimulation. Wherein, filtering selection filters 50HZ power frequency interference.
And decomposing the preprocessed original electroencephalogram signal into a plurality of electroencephalogram independent components.
Decomposing the preprocessed original electroencephalogram signal into a plurality of electroencephalogram independent components, and specifically comprising the following steps: the method adopts an independent component analysis method to decompose the preprocessed original electroencephalogram signal into a plurality of electroencephalogram independent components, and comprises the following specific processes: after the independent component sources containing the artifacts are removed, the residual independent component sources not containing the artifacts are decomposed by ICA again to form g electroencephalogram independent components, minimum mutual information exists among the g electroencephalogram independent components, and each independent component is a time sequence signal. Wherein g is a positive integer.
And extracting the initial characteristic vector of the original electroencephalogram signal through the electroencephalogram independent components.
The method comprises the following steps of extracting initial characteristic vectors of original electroencephalogram signals through electroencephalogram independent components, and specifically comprises the following steps: and extracting the initial characteristic vector of the original electroencephalogram signal by using Hilbert-Huang transform.
The method for extracting the initial characteristic vector of the original electroencephalogram signal by using Hilbert-Huang transform comprises the following steps: the unsteady-state time-series signals are decomposed into r signals including instantaneous frequencies through empirical mode decomposition, and the signals including the instantaneous frequencies are natural mode functions. Wherein r is a positive integer.
And performing Hilbert transform on each natural mode function to obtain instantaneous amplitude, phase and instantaneous frequency of the signal comprising the instantaneous frequency.
All time-frequency distributions including the instantaneous amplitude of the instantaneous frequency signal are Hilbert spectra, and Hilbert marginal spectra are obtained by integrating the Hilbert spectra.
And calculating the energy of different frequency bands of the electroencephalogram independent components through the Hilbert marginal spectrum to obtain the initial characteristic vector of the electroencephalogram signal.
The specific process of extracting the initial characteristic vector of the electroencephalogram signal through the electroencephalogram independent components comprises the following steps: as the electroencephalogram signals belong to nonlinear characteristics and can change rapidly along with time, Hilbert-Huang Transform (HHT) is selected to extract characteristic vectors, the HHT can track the time-frequency change of the original signals, the detailed characteristic vectors of the electroencephalogram signals are reserved in any time-frequency range, and then the high-resolution time-frequency characteristic extraction can be realized. The HHT includes: empirical Mode Decomposition (EMD) and Hilbert Transform (Hilbert Transform).
In the first step, the time series signal is decomposed into r Intrinsic Mode Functions (IMFs) u by EMDr(t) of (d). After EMD decomposition, the instantaneous frequency of the time series signal has physical significance, i.e. the instantaneous frequency of the time series signal can be calculated.
In the second step, a Hilbert transform (Hilbert transform) is performed for each IMFs. The Hilbert transform is a spectral analysis method by which the instantaneous amplitude, phase and instantaneous frequency of each IMFs can be obtained.
In the formula (1), H [, ]]For the Hilbert transform operator, H [ u ]r(t)]Is the original signal ur(t) a signal value, H [ u ], obtained by Hilbert transformr(t)]For calculating the instantaneous amplitude; u. ofr(t) denotes the r-th IMFs; denotes a convolution operation; pi is the circumference ratio; p.v. represents the cauchy principal value integral; u. ofr(τ) represents the instantaneous signal of the r-th IMFs;τ represents a minute amount of time, τ being a constant; t represents time.
Instantaneous amplitude:
wherein, ar(t) represents the instantaneous amplitude of the r-th IMFs.
Phase position:
wherein phi isr(t) denotes the phase of the r-th IMFs, H { u }r(t) is a Hilbert transform of the IMFs. The purpose of obtaining the phase is to find the instantaneous frequency.
Instantaneous frequency:
wherein, ω isr(t) denotes the instantaneous frequency of the r-th IMFs; and pi is the circumferential ratio.
The distribution of the instantaneous amplitudes of all IMFs over time and frequency constitutes the Hilbert spectrum (Hilbert spectrum) H (ω, t), by fourier transforming the Hilbert transformed signal:
in the formula (5), ω represents frequency, Re represents a real part, e is a natural base number, and j is a complex unit.
The Hilbert spectrum represents the law of instantaneous amplitude variation with time and frequency over the entire frequency band. Integration of the Hilbert spectrum yields a Hilbert marginal spectrum (Hilbert marginal spectrum) which represents the variation of the signal amplitude with frequency over the entire frequency band. The Hilbert marginal spectrum has a higher frequency resolution. The Hilbert margin spectrum is expressed as:
in the formula (6), T represents the electroencephalogram signal acquisition time.
Computing each independent component IC of electroencephalogram signal by using Hilbert marginal spectrumg5 bands of energy. These 5 commonly used bands divided according to different frequencies include: delta 1-4Hz, theta 4-8Hz, alpha 8-12Hz, beta 12-30Hz, gamma 30-64 Hz. And further 5 n-dimensional electroencephalogram signal feature vectors corresponding to each basic graph can be obtained.
Constructing and adjusting a support vector machine classifier according to the initial feature vector, which specifically comprises the following steps:
storing the feature vector of the electroencephalogram signal in a data set, namely storing the feature vector z of the electroencephalogram signal in the data setsStored within the data set.
The data set is divided into a training set and a test set.
And constructing a support vector machine classifier through a training set.
And adjusting the constructed support vector machine classifier through the test set.
And 102, acquiring an electroencephalogram signal to be identified.
And collecting the electroencephalogram signals by using a non-invasive electroencephalogram collecting device.
The electroencephalogram signal acquisition experiment comprises the following steps:
experimental equipment: brain electrical signal collection equipment.
External stimulation: steady-state visual stimulus for a base graphic, comprising: stimulation frequency and stimulation duration.
The experimental process comprises the following steps: (1) informing experimenters to collect attention items of experiments; (2) an experimenter wears the electroencephalogram cap to detect the connection efficiency of the electroencephalogram cap; (3) starting the experiment, wherein an experimenter watches a screen and watches a visual stimulation video; (4) recording the electroencephalogram signals of the experimenters in real time; (5) and (5) ending the experiment after the electroencephalogram signal is collected.
And 103, preprocessing the electroencephalogram signal to be recognized to obtain a preprocessed electroencephalogram signal to be recognized.
The specific process of preprocessing the electroencephalogram signals to be identified by using a matrix factory or a matrix laboratory (matrix & laboratory, Matlab) is as follows: (1) leading in an electroencephalogram signal; (2) positioning an electrode position of the acquisition equipment; (3) re-referencing; (4) filtering; (5) segmenting the electroencephalogram signal according to the stimulation event; (6) reject bad lead signals/delete bad segments; (7) superposing and averaging electroencephalogram signal segments under the same type of stimulation; (8) and identifying and eliminating artifact interference.
Wherein, filtering selection filters 50HZ power frequency interference.
Artifact interference is identified using an Independent Component Analysis (ICA) method.
ICA is a principal component decomposition method that separates individual components from a set of mixed observed signals. Theoretically, interference signals generated by the heartbeat, the eye movement, the myoelectricity artifact and other interference sources in the electroencephalogram signals are all generated by mutually independent information sources, so that the signals are decomposed into n independent source signals through the ICA, artifact interference is identified according to artifact characteristics, and independent component sources considered as artifacts are removed by the ICA.
And 104, extracting the preprocessed feature vector of the electroencephalogram signal to be recognized.
After the independent component sources containing the artifacts are removed, the residual independent component sources not containing the artifacts are decomposed by ICA again to form g electroencephalogram independent components, minimum mutual information exists among the g electroencephalogram independent components, and each independent component is a time sequence signal. The electroencephalogram signals belong to nonlinear characteristics and can change rapidly along with time, so that Hilbert-Huang Transform (HHT) is selected to extract characteristic vectors, the HHT can track time-frequency change of original signals, detailed characteristic vectors of the electroencephalogram signals are reserved in any time-frequency range, and high-resolution time-frequency characteristic extraction can be achieved. The HHT includes: empirical Mode Decomposition (EMD) and Hilbert Transform (Hilbert Transform).
In the first step, the time series signal is decomposed into r Intrinsic Mode Functions (IMFs) by EMDs)ur(t) of (d). After EMD decomposition, the instantaneous frequency of the time series signal has physical significance, i.e. the instantaneous frequency of the time series signal can be calculated.
In the second step, a Hilbert transform is performed for each IMFs. The Hilbert transform is a spectral analysis method by which the instantaneous amplitude, phase and instantaneous frequency of each IMFs can be obtained.
In the formula (1), H [, ]]For the Hilbert transform operator, H [ u ]r(t)]Is the original signal ur(t) a signal value, H [ u ], obtained by Hilbert transformr(t)]For calculating the instantaneous amplitude; u. ofr(t) denotes the r-th IMFs; denotes a convolution operation; pi is the circumference ratio; p.v. represents the cauchy principal value integral; u. ofr(τ) represents the instantaneous signal of the r-th IMFs; τ represents a minute amount of time, τ being a constant; t represents time.
Instantaneous amplitude:
wherein, ar(t) represents the instantaneous amplitude of the r-th IMFs.
Phase position:
wherein phi isr(t) denotes the phase of the r-th IMFs, H { u }r(t) is a Hilbert transform of the IMFs. The purpose of obtaining the phase is to find the instantaneous frequency.
Instantaneous frequency:
wherein, ω isr(t) representsInstantaneous frequency of the r-th IMFs; and pi is the circumferential ratio.
The distribution of the instantaneous amplitudes of all IMFs over time and frequency constitutes the Hilbert spectrum (Hilbert spectrum) H (ω, t), by fourier transforming the Hilbert transformed signal:
in the formula (5), ω represents frequency, Re represents a real part, e is a natural base number, and j is a complex unit.
The Hilbert spectrum represents the law of instantaneous amplitude variation with time and frequency over the entire frequency band. Integration of the Hilbert spectrum yields a Hilbert marginal spectrum (Hilbert marginal spectrum) which represents the variation of the signal amplitude with frequency over the entire frequency band. The Hilbert marginal spectrum has a higher frequency resolution. The Hilbert margin spectrum is expressed as:
in the formula (6), T represents the electroencephalogram signal acquisition time.
Computing each independent component IC of electroencephalogram signal by using Hilbert marginal spectrumg5 bands of energy. These 5 commonly used bands divided according to different frequencies include: δ: 1-4Hz, θ: 4-8Hz, α: 8-12Hz, beta: 12-30Hz, γ: 30-64 Hz. And further 5 n-dimensional electroencephalogram signal feature vectors corresponding to each basic graph can be obtained.
And 105, importing the feature vector into the constructed support vector machine classifier to obtain a judgment result. Specifically, 5 n-dimensional electroencephalogram signal feature vectors are led into a constructed support vector machine classifier to obtain a judgment result.
And 106, generating a basic graph based on the electroencephalogram signal according to the judgment result.
Generating a basic graph based on the electroencephalogram signal according to the judgment result, and specifically comprising the following steps:
and drawing a basic graph based on the electroencephalogram signal through Computer Aided Design (CAD) according to the judgment result.
Example 2
The embodiment provides a pattern recognition generation system based on electroencephalogram signals. Fig. 2 is a system structure diagram of a pattern recognition generating system based on electroencephalogram signals provided in embodiment 2 of the present invention. Referring to fig. 2, a pattern recognition generating system based on electroencephalogram signals includes:
and a support vector machine classifier building module 201, configured to build a support vector machine classifier.
The support vector machine classifier building module specifically comprises:
and the original electroencephalogram signal acquiring unit is used for acquiring the original electroencephalogram signals acquired by the non-invasive electroencephalogram acquiring equipment.
And the original electroencephalogram signal preprocessing unit is used for preprocessing the acquired original electroencephalogram signal to obtain a preprocessed original electroencephalogram signal.
The original brain electrical signal preprocessing unit comprises: and identifying artifact interference and rejecting subunits.
The artifact interference identification and elimination subunit is used for identifying artifact interference in the original electroencephalogram signal and eliminating the artifact interference, and specifically comprises the following steps:
decomposing an original electroencephalogram signal into a plurality of independent source signals by adopting an independent component analysis method;
and eliminating the artifact signals in the independent source signals according to the artifact characteristics.
The decomposition preprocessing electroencephalogram signal unit is used for decomposing the preprocessed electroencephalogram signals into a plurality of electroencephalogram independent components and specifically comprises the following steps: and decomposing the preprocessed electroencephalogram signal into a plurality of electroencephalogram independent components by adopting an independent component analysis method.
The unit for extracting the initial feature vector is used for extracting the initial feature vector of the electroencephalogram signal through the electroencephalogram independent component, and specifically comprises the following steps:
decomposing the unsteady-state time series signals into r signals comprising instantaneous frequencies through empirical mode decomposition, wherein the signals comprising the instantaneous frequencies are natural mode functions;
and performing Hilbert transform on each natural mode function to obtain instantaneous amplitude, phase and instantaneous frequency of the signal comprising the instantaneous frequency.
The unit for extracting the initial feature vector further comprises:
all time-frequency distributions including the instantaneous amplitude of the instantaneous frequency signal are Hilbert spectrums, and Hilbert marginal spectrums are obtained by integrating the Hilbert spectrums;
and calculating the energy of different frequency bands of the electroencephalogram independent components through the Hilbert marginal spectrum to obtain the initial characteristic vector of the electroencephalogram signal.
The construction and adjustment unit is used for constructing and adjusting the support vector machine classifier according to the initial feature vector, and specifically comprises the following steps:
storing the feature vector of the electroencephalogram signal in a data set;
dividing the data set into a training set and a testing set;
constructing a support vector machine classifier through a training set;
and adjusting the constructed support vector machine classifier through the test set.
And the electroencephalogram signal acquisition module 202 is used for acquiring the electroencephalogram signal to be identified.
The electroencephalogram signal preprocessing module 203 is used for preprocessing the electroencephalogram signal to be identified to obtain the preprocessed electroencephalogram signal to be identified.
And the electroencephalogram signal feature vector extraction module 204 is used for extracting the feature vector of the preprocessed electroencephalogram signal to be identified.
And the result obtaining module 205 is configured to import the feature vector into the constructed support vector machine classifier to obtain a determination result.
The graph generating module 206 is configured to generate a basic graph based on the electroencephalogram signal according to the determination result, and specifically includes:
and drawing a basic graph based on the electroencephalogram signal through computer aided design according to the judgment result.
Example 3
The present embodiment provides a pattern recognition generation method based on electroencephalogram signals, and fig. 3 is a flowchart of the pattern recognition generation method based on electroencephalogram signals provided in embodiment 3 of the present invention. The graph identified and generated in the embodiment is as follows: circular, square and triangular. Referring to fig. 3, the method includes:
training data is collected. The training data is the original brain electrical signal.
And collecting the electroencephalogram signals by using a non-invasive electroencephalogram collecting device.
The electroencephalogram signal acquisition experiment comprises the following steps:
experimental equipment: brain electrical signal collection equipment. Preferably, EMOTIV EPOC +14 channel electroencephalogram acquisition equipment is adopted.
External stimulation: a steady-state visual stimulus for a base graphic, comprising: stimulation frequency and stimulation duration.
The experimental process comprises the following steps: (1) informing experimenters to collect attention items of experiments; (2) an experimenter wears the electroencephalogram cap to detect the connection efficiency of the electroencephalogram cap; (3) starting the experiment, wherein an experimenter watches a screen and watches a visual stimulation video; (4) recording the electroencephalogram signals of the experimenters in real time; (5) and (5) ending the experiment after the electroencephalogram signal is collected.
And preprocessing the training data to obtain preprocessed training data.
The specific process of preprocessing the original electroencephalogram signal by using Matlab is as follows: (1) leading in an electroencephalogram signal; (2) positioning an electrode position of the acquisition equipment; (3) re-referencing; (4) filtering; (5) segmenting the electroencephalogram signal according to the stimulation event; (6) reject bad lead signals/delete bad segments; (7) superposing and averaging electroencephalogram signal segments under the same type of stimulation; (8) and identifying and eliminating artifact interference.
Wherein, filtering selection filters 50HZ power frequency interference.
The ICA method is adopted for identifying the artifact interference.
ICA is a principal component decomposition method that separates individual components from a set of mixed observed signals. Theoretically, interference signals generated by the heartbeat, the eye movement, the myoelectricity artifact and other interference sources in the electroencephalogram signals are all generated by mutually independent information sources, so that the signals are decomposed into n independent source signals through the ICA, artifact interference is identified according to artifact characteristics, and independent component sources considered as artifacts are removed by the ICA.
The brain electrical acquisition device used has 14 channels, and therefore the independent component source has 14 (m is 14), i.e. the original brain electrical signal includes brain electrical signals X { X } from 14 channels1(t),x2(t),…,x14(t)}，X＝{x1(t),x2(t),…,x14(t) } m unknown independent source signals S ═ S1(t),s2(t),…,sm(t) }, i.e. the relationship of the electroencephalogram signal to each independent component is X ═ WS, where W is the weight matrix. Decomposing EEG signals from 14 channels by ICA to obtain independent component IC decomposed by ICAx，ICx＝W-1X，ICx＝{IC1(t),IC2(t),…,IC14(t) }, in which W-1Is an inverse matrix of W, ICx(ii) S; x represents the channel sequence number, x is less than or equal to 14. t represents time.
And extracting the feature vector of the preprocessed training data.
After the independent component sources containing the artifacts are removed, the residual independent component sources not containing the artifacts are decomposed by ICA again to form g electroencephalogram independent component ICsg，g＝14，ICg＝{IC1(t),IC2(t),…,IC14(t), there is minimum mutual information among the 14 electroencephalogram independent components, and each independent component is a time sequence signal. The electroencephalogram signals belong to nonlinear characteristics and can change rapidly along with time, so that Hilbert-Huang Transform (HHT) is selected to extract characteristic vectors, the HHT can track time-frequency change of original signals, detailed characteristic vectors of the electroencephalogram signals are reserved in any time-frequency range, and high-resolution time-frequency characteristic extraction can be achieved. The HHT includes: empirical Mode Decomposition (EMD) and Hilbert Transform (Hilbert Transform).
In the first step, the time series signal is decomposed into r Intrinsic Mode Functions (IMFs) u by EMDr(t) of (d). After EMD decomposition, the instantaneous frequency of the time series signal has physical significance, i.e. the instantaneous frequency of the time series signal can be calculated.
In the second step, a Hilbert transform is performed for each IMFs. The Hilbert transform is a spectral analysis method by which the instantaneous amplitude, phase and instantaneous frequency of each IMFs can be obtained.
In the formula (1), H [, ]]For the Hilbert transform operator, H [ u ]r(t)]Is the original signal ur(t) a signal value, H [ u ], obtained by Hilbert transformr(t)]For calculating the instantaneous amplitude; u. ofr(t) denotes the r-th IMFs; denotes a convolution operation; pi is the circumference ratio; p.v. represents the cauchy principal value integral; u. ofr(τ) represents the instantaneous signal of the r-th IMFs; τ represents a minute amount of time, τ being a constant; t represents time.
Instantaneous amplitude:
wherein, ar(t) represents the instantaneous amplitude of the r-th IMFs.
Phase position:
wherein phi isr(t) denotes the phase of the r-th IMFs, H { u }r(t) is a Hilbert transform of the IMFs. The purpose of obtaining the phase is to find the instantaneous frequency.
Instantaneous frequency:
wherein, ω isr(t) denotes the r-thInstantaneous frequencies of the IMFs; and pi is the circumferential ratio.
The distribution of the instantaneous amplitudes of all IMFs over time and frequency constitutes the Hilbert spectrum (Hilbert spectrum) H (ω, t), by fourier transforming the Hilbert transformed signal:
in the formula (5), ω represents frequency, Re represents a real part, e is a natural base number, and j is a complex unit.
The Hilbert spectrum represents the law of instantaneous amplitude variation with time and frequency over the entire frequency band. Integration of the Hilbert spectrum yields a Hilbert marginal spectrum (Hilbert marginal spectrum) which represents the variation of the signal amplitude with frequency over the entire frequency band. The Hilbert marginal spectrum has a higher frequency resolution. The Hilbert margin spectrum is expressed as:
in the formula (6), T represents the electroencephalogram signal acquisition time.
Computing each independent component IC of electroencephalogram signal by using Hilbert marginal spectrumg5 bands of energy. These 5 commonly used bands divided according to different frequencies include: δ: 1-4Hz, θ: 4-8Hz, α: 8-12Hz, beta: 12-30Hz, γ: 30-64 Hz. And further 70-dimensional electroencephalogram signal feature vectors corresponding to each basic graph can be obtained.
In the embodiment, the Hilbert-Huang transformation is adopted to extract the electroencephalogram characteristics of the basic graph, and the time-frequency information of the electroencephalogram signals is fully reserved.
And constructing and adjusting a support vector machine classifier according to the electroencephalogram signal feature vector, storing the electroencephalogram signal feature vector in a data set, and dividing the data set into a training set and a test set.
And (3) learning a discriminant model of a Support Vector Machine (SVM) by using a training set. The training set includes 80% of the data in the data set.
Data set is denoted as D { (z)1，y1)，(z2，y2)，...，(zs，ys) In which z issRepresenting a 70-dimensional electroencephalogram signal feature vector, ysA class label corresponding to the feature vector is indicated, and s is a feature vector sample number. The category label is the category to which the electroencephalogram signal features belong, and the category corresponding to the electroencephalogram signal features is input when the model is trained.
Because the generated graphs are not more than two types and belong to the multi-classification problem, one SVM classifier can classify two types, the support vector machine classifier constructed by the embodiment comprises a plurality of SVM classifiers, and therefore the multi-classification SVM algorithm is realized. It is first determined that classifying k classes requires constructing k (k-1)/2 classifiers. The electroencephalogram signal features in this example are of 3 categories: if the number k is 3 for a circle, a square, or a triangle, that is, 3 classifiers need to be constructed in this embodiment. These 3 classifiers are classifier 1: round-square support vector machine classifier, classifier 2: triangle-square support vector machine classifier and classifier 3: a circular-triangular support vector machine classifier.
SVM is a classifier, when zsWhen it belongs to the first class, ys1 is ═ 1; when z issWhen it belongs to the second class, ysIs-1. When training a support vector machine classifier, according to collected training data, namely the corresponding relation between an original electroencephalogram signal and 3 graphic categories, a training set can be divided into three categories of small data sets, then one classifier is trained through every two categories of small data sets, when the training classifier 1 is classified, a circle is a first category, and a square is a second category; when the training classifier 2 is used for classification, the triangle is of a first type, and the square is of a second type; when the classifier 3 is trained for classification, the circle is of the first class and the triangle is of the second class.
As known by SVM algorithm, constructing a classifier is a classification hyperplane problem which determines two classes, and is a typical constrained quadratic programming problem, wherein an optimal classification hyperplane is a plane which maximizes the geometric separation of two classes of vectors to be classified from the hyperplane. And because of the classification of EEG signalsSubject is linear inseparable problem, needs to introduce relaxation variable xisDelta, relaxation variable xisBeing non-negative, s is the feature vector sample number. That is, the objective function for finding the optimal classification hyperplane is expressed as:
the constraint conditions are as follows:
ys[wTK(zs,z′s)+b]≥1-ξs (8)
wherein w is a weight vector; b is the hyperplane intercept; ξ represents the relaxation variable; xisIs the relaxation variable of the s-th eigenvector; c is a penalty coefficient, and the larger C is, the larger the penalty of misclassification is; w is aTA transposed matrix representing the weight vector. And because the classification problem of the electroencephalogram signals is a linear inseparable problem, the samples are mapped into an implicit characteristic space and then solved. K (z)s,z′s) To be z issA kernel function mapped to a high dimensional space, i.e. an implicit feature space. Because the feature dimension of the brain electricity is more, the linear kernel function is adopted, and the classification effect is better.
The quadratic programming problem is solved by a Lagrange (Lagrange) method, and a decision function of the Lagrange method is represented as:
wherein, f (z)s) Is the result of the decision, asAre Lagrange's (Lagrange) coefficients, i.e. zsThe determined category to which the user belongs.
During data training, 3 classification hyperplanes are constructed as formulas (7) and (8), which correspond to 3 classifiers: classifier 1 (round-square support vector machine classifier), classifier 2 (triangle-square support vector machine classifier), and classifier 3 (round-triangle support vector machine classifier).
When classifying an unknown sample, this embodiment provides a specific implementation of a classifier classification method, and fig. 4 is a directed acyclic graph of the classification method provided in embodiment 3 of the present invention, where the classification method provided in this embodiment uses a directed acyclic graph. Referring to fig. 4, when the 3 classifiers are classified, which branch to continue classifying is determined according to the classification result starting from the top root node, that is, when the classifier 1 performs classification, when the classification result is a circle, the classification is performed left and down by the classifier 3; and when the classification result is square, classifying the square by the classifier 2 to the right and downwards until the final classification result is displayed at the bottom layer.
And verifying the discriminant model of the support vector machine by using the test set. The test set comprises 20% of data of the data set, the data in the test set is used as the input of a discriminant model of the support vector machine, and the penalty coefficient C is adjusted according to the output of the discriminant model so as to balance the relationship between the complexity and the misclassification rate of the support vector. When the penalty coefficient C is small, the penalty to the experience error is small, the complexity of the support vector machine is small, and the misclassification rate is high; when the value C is large, the penalty on the empirical error is large, the complexity of the support vector machine is large, and the misclassification rate is low.
And obtaining a final discrimination model. When the complexity of the support vector machine reaches the maximum value allowed by the feature space, the popularization capability of the SVM is the best, namely the misclassification rate is in the acceptance range at the moment, and the punishment coefficient C at the moment is determined. In addition, C depends to some extent on the amount of noise in the brain electrical signal data.
and acquiring the electroencephalogram signal to be identified.
And preprocessing the electroencephalogram signal to be recognized to obtain the preprocessed electroencephalogram signal to be recognized.
The specific process of preprocessing the electroencephalogram signal to be identified by using Matlab is as follows: (1) leading in an electroencephalogram signal; (2) positioning an electrode position of the acquisition equipment; (3) re-referencing; (4) filtering; (5) segmenting the electroencephalogram signal according to the stimulation event; (6) reject bad lead signals/delete bad segments; (7) superposing and averaging electroencephalogram signal segments under the same type of stimulation; (8) and identifying and eliminating artifact interference.
Wherein, filtering selection filters 50HZ power frequency interference.
Artifact interference is identified using an Independent Component Analysis (ICA) method.
And extracting the feature vector of the preprocessed electroencephalogram signal to be identified. And extracting the feature vector by using Hilbert-Huang Transform (HHT).
And importing the feature vector into the constructed support vector machine classifier to obtain a judgment result.
And drawing three basic graphs generated based on the electroencephalogram signals through Computer Aided Design (CAD) according to the judgment result.
In the embodiment, CAD is called, the graph discrimination result of the classifier is converted into the graph information, the cognitive intention reappearance is realized more intuitively, and a closed loop for feeding back the design result is formed, so that the method can be better applied to practice and aided design.
In the embodiment, the accuracy of electroencephalogram identification can be reflected more visually in the whole process and result from acquisition of electroencephalogram signals to generation of basic graphs, and a good foundation is laid for further development of electroencephalogram intention identification and reproduction technologies in the future.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (4)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910269588.3A CN109993132B (en) | 2019-04-04 | 2019-04-04 | A method and system for generating pattern recognition based on EEG signals |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910269588.3A CN109993132B (en) | 2019-04-04 | 2019-04-04 | A method and system for generating pattern recognition based on EEG signals |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109993132A CN109993132A (en) | 2019-07-09 |
| CN109993132B true CN109993132B (en) | 2021-07-13 |
Family
ID=67132168
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910269588.3A Expired - Fee Related CN109993132B (en) | 2019-04-04 | 2019-04-04 | A method and system for generating pattern recognition based on EEG signals |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109993132B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112668402A (en) * | 2020-12-09 | 2021-04-16 | 山东大学 | Brain wave analysis method based on Hilbert-Huang transform and support vector machine optimization |
| CN114469140A (en) * | 2021-07-19 | 2022-05-13 | 济南大学 | EEG feature extraction method and system based on synchronous extraction and transformation |
| CN118021323B (en) * | 2024-02-27 | 2024-07-05 | 华南理工大学 | A method for EEG signal recognition based on multi-domain feature fusion |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103699230A (en) * | 2014-01-14 | 2014-04-02 | 东南大学 | Digital interface interaction method on basis of icon electrocerebral control |
| CN104809434A (en) * | 2015-04-22 | 2015-07-29 | 哈尔滨工业大学 | Sleep staging method based on single-channel electroencephalogram signal ocular artifact removal |
| US20150347846A1 (en) * | 2014-06-02 | 2015-12-03 | Microsoft Corporation | Tracking using sensor data |
| CN106803081A (en) * | 2017-01-25 | 2017-06-06 | 东南大学 | A kind of brain electricity sorting technique based on Multi-classifers integrated |
| CN107361766A (en) * | 2017-07-17 | 2017-11-21 | 中国人民解放军信息工程大学 | A kind of mood EEG signal identification method based on EMD domains multidimensional information |
| CN108363493A (en) * | 2018-03-20 | 2018-08-03 | 山东建筑大学 | User characteristics method for establishing model, system and storage medium based on brain-computer interface |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101869477B (en) * | 2010-05-14 | 2011-09-14 | 北京工业大学 | Self-adaptive EEG signal ocular artifact automatic removal method |
| CN102835955B (en) * | 2012-09-08 | 2014-02-26 | 北京工业大学 | A method for automatic removal of electro-oculogram artifacts in EEG signals without threshold setting |
| CN103690163B (en) * | 2013-12-21 | 2015-08-05 | 哈尔滨工业大学 | Based on the automatic eye electrical interference minimizing technology that ICA and HHT merges |
| CN103942568B (en) * | 2014-04-22 | 2017-04-05 | 浙江大学 | A kind of sorting technique based on unsupervised feature selection |
| IL239191A0 (en) * | 2015-06-03 | 2015-11-30 | Amir B Geva | Image classification system |
| CN106419912A (en) * | 2016-10-20 | 2017-02-22 | 重庆邮电大学 | Multi-lead electroencephalogram signal ocular artifact removing method |
| CN106859641B (en) * | 2017-02-20 | 2019-08-20 | 华南理工大学 | A method for removing NMR artifacts in EEG signals based on automatic ICA |
-
2019
- 2019-04-04 CN CN201910269588.3A patent/CN109993132B/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103699230A (en) * | 2014-01-14 | 2014-04-02 | 东南大学 | Digital interface interaction method on basis of icon electrocerebral control |
| US20150347846A1 (en) * | 2014-06-02 | 2015-12-03 | Microsoft Corporation | Tracking using sensor data |
| CN104809434A (en) * | 2015-04-22 | 2015-07-29 | 哈尔滨工业大学 | Sleep staging method based on single-channel electroencephalogram signal ocular artifact removal |
| CN106803081A (en) * | 2017-01-25 | 2017-06-06 | 东南大学 | A kind of brain electricity sorting technique based on Multi-classifers integrated |
| CN107361766A (en) * | 2017-07-17 | 2017-11-21 | 中国人民解放军信息工程大学 | A kind of mood EEG signal identification method based on EMD domains multidimensional information |
| CN108363493A (en) * | 2018-03-20 | 2018-08-03 | 山东建筑大学 | User characteristics method for establishing model, system and storage medium based on brain-computer interface |
Non-Patent Citations (2)
| Title |
|---|
| 基于脑机接口的界面设计风格沟通方法研究;李钰等;《设计》;20171231(第19期);56-57 * |
| 多模态脑电模式识别方法及其应用;刘烨;《万方数据》;20171129;正文全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109993132A (en) | 2019-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Shukla et al. | Feature extraction and selection for emotion recognition from electrodermal activity | |
| Kumar et al. | Envisioned speech recognition using EEG sensors | |
| CN109299751B (en) | SSVEP EEG Classification Method Based on EMD Data Augmented Convolutional Neural Model | |
| Chen et al. | Effects of data augmentation method borderline-SMOTE on emotion recognition of EEG signals based on convolutional neural network | |
| CN109993132B (en) | A method and system for generating pattern recognition based on EEG signals | |
| Shahnaz et al. | Emotion recognition based on wavelet analysis of Empirical Mode Decomposed EEG signals responsive to music videos | |
| CN103294199B (en) | A kind of unvoiced information identifying system based on face's muscle signals | |
| CN106919956A (en) | Brain wave age forecasting system based on random forest | |
| CN105426842A (en) | Support vector machine based surface electromyogram signal multi-hand action identification method | |
| CN107361766A (en) | A kind of mood EEG signal identification method based on EMD domains multidimensional information | |
| CN103631941A (en) | Electroencephalogram-based target image retrieval system | |
| Shao et al. | Single-channel SEMG using wavelet deep belief networks for upper limb motion recognition | |
| Bueno-López et al. | Understanding instantaneous frequency detection: a discussion of Hilbert-Huang Transform versus Wavelet Transform | |
| Vayadande et al. | Facial emotion based song recommendation system | |
| US20230315203A1 (en) | Brain-Computer Interface Decoding Method and Apparatus Based on Point-Position Equivalent Augmentation | |
| Khan et al. | Automated eye movement classification based on EMG of EOM signals using FBSE-EWT technique | |
| Chen et al. | Design and implementation of human-computer interaction systems based on transfer support vector machine and EEG signal for depression patients’ emotion recognition | |
| CN115414051A (en) | Emotion classification and recognition method of electroencephalogram signal self-adaptive window | |
| CN116244633A (en) | A motor imagery EEG signal classification method and system | |
| CN111466930A (en) | A method and system for audio-visual evoked emotion recognition based on EEG signals | |
| CN113974627A (en) | An emotion recognition method based on brain-computer generative confrontation | |
| Ahmed et al. | Human emotion modeling based on salient global features of EEG signal | |
| Keutayeva et al. | Data constraints and performance optimization for transformer-based models in eeg-based brain-computer interfaces: A survey | |
| Islam et al. | Classification of electromyography signals using support vector machine | |
| Gupta et al. | Hand Gesture recognition and classification by discriminant and principal component analysis using machine learning techniques |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210713 |