CN113011493A - Electroencephalogram emotion classification method, device, medium and equipment based on multi-kernel width learning - Google Patents

Electroencephalogram emotion classification method, device, medium and equipment based on multi-kernel width learning Download PDF

Info

Publication number
CN113011493A
CN113011493A CN202110290152.XA CN202110290152A CN113011493A CN 113011493 A CN113011493 A CN 113011493A CN 202110290152 A CN202110290152 A CN 202110290152A CN 113011493 A CN113011493 A CN 113011493A
Authority
CN
China
Prior art keywords
kernel
electroencephalogram
feature
width learning
nodes
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.)
Granted
Application number
CN202110290152.XA
Other languages
Chinese (zh)
Other versions
CN113011493B (en
Inventor
康雪艳
晋建秀
张通
陈俊龙
刘竹琳
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.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
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 South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN202110290152.XA priority Critical patent/CN113011493B/en
Publication of CN113011493A publication Critical patent/CN113011493A/en
Application granted granted Critical
Publication of CN113011493B publication Critical patent/CN113011493B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
    • A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
    • 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
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

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

Abstract

The invention provides a method, a device, a medium and equipment for learning electroencephalogram emotion classification based on multi-kernel width; the method comprises the following steps: acquiring an electroencephalogram signal of a testee, and preprocessing the electroencephalogram signal; performing feature extraction on the preprocessed electroencephalogram signals by adopting a convolutional neural network to obtain a one-dimensional space vector V; and inputting the one-dimensional space characteristic V into a multi-kernel width learning system to obtain the emotion type of the electroencephalogram signal. The method integrates the advantages of a convolutional neural network and a multi-kernel width learning system, and the multi-kernel function mapping can express the characteristic data of the electroencephalogram signals more accurately and reasonably in a combined space, so that the classification precision is improved.

Description

Electroencephalogram emotion classification method, device, medium and equipment based on multi-kernel width learning
Technical Field
The invention relates to the technical field of emotion recognition, in particular to an electroencephalogram emotion classification model based on multi-kernel width learning.
Background
Emotion is a complex state that combines emotion, thought and behavior, and is a psychophysiological response of a person to internal or external stimuli. Emotion calculation is widely used. According to the research, the generation or activity of emotion is closely related to the activity of cerebral cortex, and brain electrical energy reflects various electrical activities of the brain and functional states of the brain, thus reflecting the emotional state of human. Although the existing emotion recognition methods have achieved good results, there are still places where improvements are needed. How to build a more effective emotion recognition calculation model is still a technical problem in the field of emotion recognition.
Furthermore, the method is simple. The recognition of emotion through electroencephalogram signals does reduce a large number of human interference problems, has great significance in make internal disorder or usurp research, but brings many problems at the same time. The electroencephalogram signal has the characteristics of dynamics, non-stationarity, large noise interference and the like, is more complex and contains larger information amount compared with other external characteristics, and can simultaneously acquire other interference signals such as electrooculogram and myoelectricity besides the electroencephalogram signal. Therefore, how to extract easily-recognized emotional features from the electroencephalogram signals is a difficult point of research. The traditional machine learning algorithm is widely applied to electroencephalogram emotion recognition and achieves certain progress. A disadvantage of this approach is that researchers must make extensive efforts to find and design various mood-related features from the originating noise signal.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention aims to provide a method, a device, a medium and equipment for learning electroencephalogram emotion classification based on multi-kernel width; the invention integrates the advantages of the convolutional neural network and the multi-kernel width learning system, and the multi-kernel function mapping can more accurately and reasonably express the characteristic data of the electroencephalogram signals in the combined space, thereby improving the classification precision.
In order to achieve the purpose, the invention is realized by the following technical scheme: a multi-kernel width learning-based electroencephalogram emotion classification method is characterized by comprising the following steps: the method comprises the following steps:
acquiring an electroencephalogram signal of a testee, and preprocessing the electroencephalogram signal;
performing feature extraction on the preprocessed electroencephalogram signals by adopting a convolutional neural network to obtain a one-dimensional space vector V;
inputting the one-dimensional spatial feature V into a multi-kernel width learning system to obtain the emotion type of the electroencephalogram signal; the operation method of the multi-core width learning system comprises the following steps:
s1, mapping the input one-dimensional space feature V to a feature node Z randomly1,Z2,…,ZnAnd obtaining mapping characteristics:
Figure BDA0002982109180000021
wherein the content of the first and second substances,
Figure BDA0002982109180000022
and
Figure BDA0002982109180000023
is randomly generated;
Figure BDA0002982109180000024
is a linear feature mapping function; f (V) ═ F1,F2,…,FM]Is an approximate multi-kernel feature of all M kernel functions; flThe method of finding is as follows:
Figure BDA0002982109180000025
wherein f isl(vi) Is a fourier approximation of one of the l-th set of kernel functions, the d-dimensional approximation is represented as follows:
Figure BDA0002982109180000026
wherein the weight is
Figure BDA0002982109180000027
Obtained from a given probability density p (w), bj,lFrom a given uniform distribution of U [0,2 π]Obtaining the result;
s2, setting all characteristic nodes as Zn＝[Z1Z2…Zn]Random mapping of feature nodes to enhanced nodes H1,H2,…,HmThen enhance node HjComprises the following steps:
Figure BDA0002982109180000028
wherein
Figure BDA0002982109180000029
And
Figure BDA00029821091800000210
is randomly generated, ζ is a non-linear mapping function;
s3, setting all enhanced nodes as Hm＝[H1H2…Hm](ii) a According to all characteristic nodes ZnAnd all enhanced nodes HmAnd calculating the output Y of the multi-core width learning system:
Figure BDA0002982109180000031
wherein, Wm＝[Zn|Hm]+Y is the connection weight to be solved; wherein pseudo-inverse [ Z ]n|Hm]+From ridge regression approximation:
Figure BDA0002982109180000032
i represents a unit vector matrix, and lambda is a hyperparameter;
and obtaining the emotion type corresponding to the electroencephalogram signal according to the output Y of the multi-kernel width learning system.
Preferably, in the step S1, fl(vi) Is a fourier approximation of one of the l-th set of kernel functions, meaning: f. ofl(vi) Is a fourier approximation of a gaussian or laplacian kernel or a power exponent kernel in the ith set of kernels.
Preferably, the preprocessing of the electroencephalogram signal refers to: the method comprises the following steps:
a1, dividing the acquired electroencephalogram signal into a plurality of data segments with equal duration and no overlap;
a2, decomposing each data segment into four frequency bands of theta, alpha, beta and gamma by using a Butterworth filter respectively, and extracting differential entropy characteristics by using a window; normalizing the differential entropy characteristics;
a3, converting the differential entropy eigenvectors into a compact two-dimensional matrix (h multiplied by w) according to the electrode distribution; where h and w are the height and width of the two-dimensional matrix, equal to the maximum number of electrodes used vertically and horizontally, respectively; each data segment is represented as a 4D structure Sj∈Rh×w×d×4J 1, 2.., N, where N is the total number of samples and d represents the number of frequency bands;
after preprocessing, each data segment is 4D EEG data with a fixed length, represented as follows:
Sj＝{fj1,fj2,fj3,fj4}
wherein f isjtAnd t is 1,2,3,4 is a data frame.
Preferably, the feature extraction is performed on the preprocessed electroencephalogram signal by using a convolutional neural network to obtain a one-dimensional space vector V, which means that: and inputting each 4D EEG data into a convolutional neural network, and extracting frequency and space information from each section of signal through the convolutional neural network to obtain a one-dimensional space vector V.
Preferably, the convolution kernel sizes of the first three convolutional layers of the convolutional neural network are all 4 × 4, and the feature map numbers of the three convolutional layers are 64, 128 and 256, respectively; the feature map number of the fourth convolutional layer is 64, the convolutional kernel size is 1 × 1, and the feature map of the previous convolutional layer is fused; using zero padding in each convolutional layer; the convolutional neural network does not use a pooling layer, and the activation function uses ReLu; the output of the fourth convolutional layer is flattened and input to a 512-cell fully connected layer.
The utility model provides a study brain wave mood sorter based on multinuclear width which characterized in that: the method comprises the following steps:
the preprocessing module is used for acquiring an electroencephalogram signal of a testee and preprocessing the electroencephalogram signal;
the feature extraction module is used for extracting features of the preprocessed electroencephalogram signals by adopting a convolutional neural network to obtain a one-dimensional space vector V;
the emotion classification module is used for inputting the one-dimensional spatial feature V into the multi-kernel width learning system to obtain the emotion type of the electroencephalogram signal;
the operation method of the multi-core width learning system comprises the following steps:
s1, mapping the input one-dimensional space feature V to a feature node Z randomly1,Z2,…,ZnAnd obtaining mapping characteristics:
Figure BDA0002982109180000041
wherein the content of the first and second substances,
Figure BDA0002982109180000042
and
Figure BDA0002982109180000043
is randomly generated;
Figure BDA0002982109180000044
is a linear feature mapping function; f (V) ═ F1,F2,…,FM]Is all M kernel functionsApproximate multi-core features of (a); flThe method of finding is as follows:
Figure BDA0002982109180000051
wherein f (v)i) Is a fourier approximation of one of the l-th set of kernel functions, the d-dimensional approximation is represented as follows:
Figure BDA0002982109180000052
wherein the weight is
Figure BDA0002982109180000053
Obtained from a given probability density p (w), bj,lFrom a given uniform distribution of U [0,2 π]Obtaining the result;
s2, setting all characteristic nodes as Zn＝[Z1Z2…Zn]Random mapping of feature nodes to enhanced nodes H1,H2,…,HmThen enhance node HjComprises the following steps:
Figure BDA0002982109180000054
wherein
Figure BDA0002982109180000055
And
Figure BDA0002982109180000056
is randomly generated, ζ is a non-linear mapping function;
s3, setting all enhanced nodes as Hm＝[H1H2…Hm](ii) a According to all characteristic nodes ZnAnd all enhanced nodes HmAnd calculating the output Y of the multi-core width learning system:
Figure BDA0002982109180000057
wherein, Wm＝[Zn|Hm]+Y is the connection weight to be solved; wherein pseudo-inverse [ Z ]n|Hm]+From ridge regression approximation:
Figure BDA0002982109180000058
i represents a unit vector matrix, and lambda is a hyperparameter;
and obtaining the emotion type corresponding to the electroencephalogram signal according to the output Y of the multi-kernel width learning system.
A storage medium, wherein the storage medium stores a computer program which, when executed by a processor, causes the processor to perform the above-described multi-kernel width-based learning electroencephalogram emotion classification method.
A computing device comprises a processor and a memory for storing a program executable by the processor, and is characterized in that when the processor executes the program stored in the memory, the multi-kernel width-based learning electroencephalogram emotion classification method is realized.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. the invention provides a method for combining a convolutional neural network and a multi-kernel width learning system for emotion recognition, which integrates the advantages of the convolutional neural network and the multi-kernel width learning system; the advantage of the convolutional neural network on the extraction of the spatial features of the sample is utilized, so that the spatial features of the adjacent electrodes of the electroencephalogram signal are extracted to the maximum extent; the multi-kernel width learning system enables the model to have stronger representation capability by using a Fourier kernel approximation method;
2. compared with the original width learning system, the multi-kernel width learning system increases the randomness of EEG data signal characteristics through nonlinear multi-kernel mapping; through multi-kernel function mapping, the electroencephalogram signal characteristic data can be more accurately and reasonably expressed in the combined space, so that the classification precision is improved;
3. the method comprises the steps of simply preprocessing an original electroencephalogram signal to obtain 4-dimensional preprocessed data, then, excavating space, frequency and time characteristics of the electroencephalogram signal through the data by a convolutional neural network, and extracting characteristics which are easy to classify.
Drawings
FIG. 1 is a flow chart of a method for learning electroencephalogram emotion classification based on multi-kernel width according to the present invention;
FIG. 2 is a schematic diagram of the result of comparing the accuracy of a DEAP data set in the method for learning electroencephalogram emotion based on multi-kernel width and other existing methods;
FIG. 3 is a schematic diagram of the result of comparing the accuracy of the SEED data set based on the multi-kernel width learning electroencephalogram emotion classification method and other existing methods.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
Example one
The flow of the method for learning electroencephalogram emotion based on multi-kernel width is shown in fig. 1, and the method comprises the following steps:
acquiring an electroencephalogram signal of a testee, and preprocessing the electroencephalogram signal;
performing feature extraction on the preprocessed electroencephalogram signals by adopting a convolutional neural network to obtain a one-dimensional space vector V;
and inputting the one-dimensional space characteristic V into a multi-kernel width learning system to obtain the emotion type of the electroencephalogram signal.
Specifically, preprocessing the electroencephalogram signal means: the method comprises the following steps:
a1, dividing the acquired electroencephalogram signal into a plurality of data segments with equal duration (such as 2 s/segment) and no overlap;
a2, decomposing each data segment into four frequency bands of theta, alpha, beta and gamma by using a Butterworth filter, and extracting a Differential Entropy (DE) feature by using a window (for example, the size is 0.5 s); to obtain the desired result, we normalized the differential entropy signature using Z-score;
a3, in order to maintain the space information between the adjacent channels, the differential entropy eigenvector is further converted into a compact two-dimensional matrix (h multiplied by w) according to the electrode distribution; where h and w are the height and width of the two-dimensional matrix, equal to the maximum number of electrodes used vertically and horizontally, respectively; in this patent, we set h-w-9, the unused channel fill is zero; thus, each data segment is represented as a 4D structure Sj∈Rh×w×d×4J is 1, 2.., N, where N is the total number of samples and d represents the number of frequency bands, and is 4;
after preprocessing, each data segment is 4D EEG data with a fixed length, represented as follows:
Sj＝{fj1,fj2,fj3,fj4}
wherein f isjtAnd t is 1,2,3,4 is a data frame.
The method is characterized in that the preprocessed electroencephalogram signal is subjected to feature extraction by adopting a convolutional neural network to obtain a one-dimensional space vector V, and the method comprises the following steps: each 4D EEG data is input into a Convolutional Neural Network (CNN), which extracts frequency and spatial information from each segment of the signal, obtaining a one-dimensional space vector V.
The sizes of convolution kernels of the first three convolution layers of the convolution neural network are 4 multiplied by 4, and the number of feature maps of the three convolution layers is 64, 128 and 256 respectively; the feature map number of the fourth convolutional layer is 64, the convolutional kernel size is 1 × 1, and the feature map of the previous convolutional layer is fused; to avoid losing edge information of the input data frame, we use zero padding in each convolutional layer; in order to retain all information of the electroencephalogram, a pooling layer is not used in the convolutional neural network, and the ReLu is used as an activation function; the output of the fourth convolutional layer is flattened and input to a full connection layer (FC) of 512 cells. To input the features into the multi-kernel width learning system, a depth cascade operation is required to combine the feature maps into a one-dimensional space vector V, so that each 4D EEG data is input into the CNN, and finally the one-dimensional space vector V is obtained.
The operation method of the multi-core width learning system comprises the following steps:
s1, inputtingIs mapped to a feature node Z randomly1,Z2,…,ZnAnd obtaining mapping characteristics:
Figure BDA0002982109180000081
wherein the content of the first and second substances,
Figure BDA0002982109180000082
and
Figure BDA0002982109180000083
is randomly generated;
Figure BDA0002982109180000084
is a linear feature mapping function; f (V) ═ F1,F2,…,FM]Is an approximate multi-kernel feature of all M kernel functions; flThe method of finding is as follows:
Figure BDA0002982109180000085
wherein f isl(vi) Is a Fourier approximation of one of the first set of kernel functions, specifically fl(vi) Is a Fourier approximation of a Gaussian or Laplace or Power exponential kernel in the ith set of kernels; each kernel function can be approximated by a fourier, d-dimensional approximation as follows:
Figure BDA0002982109180000086
wherein the weight is
Figure BDA0002982109180000087
Obtained from a given probability density p (w), bj,lFrom a given uniform distribution of U [0,2 π]Obtaining the result;
in addition, for the nuclear approximation, in addition to fourier approximation, a nuclear approximation algorithm based on Orthogonal Random Features (ORF) may be used;
s2, setting all characteristic nodes as Zn＝[Z1Z2…Zn]Random mapping of feature nodes to enhanced nodes H1,H2,…,HmThen enhance node HjComprises the following steps:
Figure BDA0002982109180000091
wherein
Figure BDA0002982109180000092
And
Figure BDA0002982109180000093
is randomly generated, ζ is a non-linear mapping function;
s3, setting all enhanced nodes as Hm＝[H1H2…Hm](ii) a According to all characteristic nodes ZnAnd all enhanced nodes HmAnd calculating the output Y of the multi-core width learning system:
Figure BDA0002982109180000094
wherein, Wm＝[Zn|Hm]+Y is the connection weight to be solved; wherein pseudo-inverse [ Z ]n|Hm]+From ridge regression approximation:
Figure BDA0002982109180000095
i represents a unit vector matrix, lambda is a hyperparameter and can be set by the user in the experimental process, and in the patent, the lambda is set to be 2-12；
And obtaining the emotion type corresponding to the electroencephalogram signal according to the output Y of the multi-kernel width learning system.
In order to implement the method for learning electroencephalogram emotion based on multi-kernel width, the embodiment provides a device for learning electroencephalogram emotion based on multi-kernel width, which is characterized in that: the method comprises the following steps:
the preprocessing module is used for acquiring an electroencephalogram signal of a testee and preprocessing the electroencephalogram signal;
the feature extraction module is used for extracting features of the preprocessed electroencephalogram signals by adopting a convolutional neural network to obtain a one-dimensional space vector V;
and the emotion classification module is used for inputting the one-dimensional spatial feature V into the multi-kernel width learning system to obtain the emotion type of the electroencephalogram signal.
In order to verify the real classification performance of the invention in classification research of electroencephalogram signals, experimental verification and comparison are carried out. Two experiments were performed on the DEAP dataset. The first experiment is that the multi-kernel width learning-based electroencephalogram emotion classification method is used for emotion classification; the second experiment is that the multi-kernel width learning system in the multi-kernel width learning-based electroencephalogram emotion classification method is replaced by the traditional width learning system, and then emotion classification is carried out. The experimental result shows that the emotion classification method adopting the multi-kernel width learning system can improve the identification accuracy by nearly 1%, which shows that the multi-kernel function can improve the identification accuracy of the electroencephalogram signal.
The electroencephalogram emotion classification method based on multi-kernel width learning is compared with other latest methods, for example, SVM, DNN, CNN, 3D-CNN, CRNN, 4D-CRNN and the like, and the comparison results of DEAP and SEED data sets are respectively shown in FIG. 2 and FIG. 3. The latter two methods (Model-BLS and Model-MKBLS) are the methods of the present invention; as can be seen from FIGS. 2 and 3, the method of the present invention works well with accuracy of the arousal level (arousal) and the titer (value) of 97.03% and 96.88%, respectively, for the DEAP data set. The accuracy for the SEED data set reached 96.84. The performance of the algorithm is obviously superior to that of the other five algorithms.
Example two
The present embodiment is a storage medium, wherein the storage medium stores a computer program, and the computer program, when executed by a processor, causes the processor to execute the method for learning electroencephalogram emotion classification based on multi-kernel width according to the first embodiment.
EXAMPLE III
The computing device comprises a processor and a memory for storing a processor executable program, wherein when the processor executes the program stored in the memory, the method for learning electroencephalogram emotion classification based on multi-kernel width is implemented.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.

Claims (8)

1. A multi-kernel width learning-based electroencephalogram emotion classification method is characterized by comprising the following steps: the method comprises the following steps:
acquiring an electroencephalogram signal of a testee, and preprocessing the electroencephalogram signal;
performing feature extraction on the preprocessed electroencephalogram signals by adopting a convolutional neural network to obtain a one-dimensional space vector V;
inputting the one-dimensional spatial feature V into a multi-kernel width learning system to obtain the emotion type of the electroencephalogram signal; the operation method of the multi-core width learning system comprises the following steps:
s1, mapping the input one-dimensional space feature V to a feature node Z randomly1,Z2,…,ZnAnd obtaining mapping characteristics:
Figure FDA0002982109170000011
wherein the content of the first and second substances,
Figure FDA0002982109170000012
and
Figure FDA0002982109170000013
is randomly generated;
Figure FDA0002982109170000014
is a linear feature mapping function; f (V) ═ F1,F2,…,FM]Is an approximate multi-kernel feature of all M kernel functions; flThe method of finding is as follows:
Figure FDA0002982109170000015
wherein f isl(vi) Is a fourier approximation of one of the l-th set of kernel functions, the d-dimensional approximation is represented as follows:
Figure FDA0002982109170000016
wherein the weight is
Figure FDA0002982109170000017
Obtained from a given probability density p (w), bj,lFrom a given uniform distribution of U [0,2 π]Obtaining the result;
s2, setting all characteristic nodes as Zn＝[Z1Z2…Zn]Random mapping of feature nodes to enhanced nodes H1,H2,…,HmThen enhance node HjComprises the following steps:
Figure FDA0002982109170000018
wherein
Figure FDA0002982109170000021
And
Figure FDA0002982109170000022
is randomly generated, ζ is a non-linear mapping function;
s3, setting all enhanced nodes as Hm＝[H1H2…Hm](ii) a According to all characteristic nodes ZnAnd all enhanced nodes HmAnd calculating the output Y of the multi-core width learning system:
Figure FDA0002982109170000023
wherein, Wm＝[Zn|Hm]+Y is the connection weight to be solved; wherein pseudo-inverse [ Z ]n|Hm]+From ridge regression approximation:
Figure FDA0002982109170000024
i represents a unit vector matrix, and lambda is a hyperparameter;
and obtaining the emotion type corresponding to the electroencephalogram signal according to the output Y of the multi-kernel width learning system.
2. The multi-kernel width learning-based electroencephalogram emotion classification method as claimed in claim 1, wherein: in the step S1, fl(vi) Is a fourier approximation of one of the l-th set of kernel functions, meaning: f. ofl(vi) Is a fourier approximation of a gaussian or laplacian kernel or a power exponent kernel in the ith set of kernels.
3. The multi-kernel width learning-based electroencephalogram emotion classification method as claimed in claim 1, wherein: the preprocessing of the electroencephalogram signals refers to: the method comprises the following steps:
a1, dividing the acquired electroencephalogram signal into a plurality of data segments with equal duration and no overlap;
a2, decomposing each data segment into four frequency bands of theta, alpha, beta and gamma by using a Butterworth filter respectively, and extracting differential entropy characteristics by using a window; normalizing the differential entropy characteristics;
a3, converting the differential entropy eigenvectors into a compact two-dimensional matrix (h multiplied by w) according to the electrode distribution; where h and w are the height and width of the two-dimensional matrix, equal to the maximum number of electrodes used vertically and horizontally, respectively; each data segment is represented as a 4D structure Sj∈Rh×w×d×4J 1, 2.., N, where N is the total number of samples and d represents the number of frequency bands;
after preprocessing, each data segment is 4D EEG data with a fixed length, represented as follows:
Sj＝{fj1,fj2,fj3,fj4}
wherein f isjtAnd t is 1,2,3,4 is a data frame.
4. The multi-kernel width learning-based electroencephalogram emotion classification method as claimed in claim 3, wherein: the method is characterized in that the preprocessed electroencephalogram signal is subjected to feature extraction by adopting a convolutional neural network to obtain a one-dimensional space vector V, and the method comprises the following steps: and inputting each 4D EEG data into a convolutional neural network, and extracting frequency and space information from each section of signal through the convolutional neural network to obtain a one-dimensional space vector V.
5. The multi-kernel width learning-based electroencephalogram emotion classification method as claimed in claim 4, wherein: the sizes of convolution kernels of the first three convolution layers of the convolution neural network are 4 multiplied by 4, and the number of feature maps of the three convolution layers is 64, 128 and 256 respectively; the feature map number of the fourth convolutional layer is 64, the convolutional kernel size is 1 × 1, and the feature map of the previous convolutional layer is fused; using zero padding in each convolutional layer; the convolutional neural network does not use a pooling layer, and the activation function uses ReLu; the output of the fourth convolutional layer is flattened and input to a 512-cell fully connected layer.
6. The utility model provides a study brain wave mood sorter based on multinuclear width which characterized in that: the method comprises the following steps:
the preprocessing module is used for acquiring an electroencephalogram signal of a testee and preprocessing the electroencephalogram signal;
the feature extraction module is used for extracting features of the preprocessed electroencephalogram signals by adopting a convolutional neural network to obtain a one-dimensional space vector V;
the emotion classification module is used for inputting the one-dimensional spatial feature V into the multi-kernel width learning system to obtain the emotion type of the electroencephalogram signal;
the operation method of the multi-core width learning system comprises the following steps:
s1, mapping the input one-dimensional space feature V to a feature node Z randomly1,Z2,…,ZnAnd obtaining mapping characteristics:
Figure FDA0002982109170000041
wherein the content of the first and second substances,
Figure FDA0002982109170000042
and
Figure FDA0002982109170000043
is randomly generated;
Figure FDA0002982109170000044
is a linear feature mapping function; f (V) ═ F1,F2,…,FM]Is an approximate multi-kernel feature of all M kernel functions; flThe method of finding is as follows:
Figure FDA0002982109170000045
wherein f isl(vi) Is a fourier approximation of one of the l-th set of kernel functions, the d-dimensional approximation is represented as follows:
Figure FDA0002982109170000046
wherein the weight is
Figure FDA0002982109170000047
Obtained from a given probability density p (w), bj,lFrom a given uniform distribution of U [0,2 π]Obtaining the result;
s2, setting all characteristic nodes as Zn＝[Z1Z2…Zn]Random mapping of feature nodes to enhanced nodes H1,H2,…,HmThen enhance node HjComprises the following steps:
Figure FDA0002982109170000048
wherein
Figure FDA0002982109170000049
And
Figure FDA00029821091700000410
is randomly generated, ζ is a non-linear mapping function;
s3, setting all enhanced nodes as Hm＝[H1H2…Hm](ii) a According to all characteristic nodes ZnAnd all enhanced nodes HmAnd calculating the output Y of the multi-core width learning system:
Figure FDA00029821091700000411
Figure FDA0002982109170000051
wherein, Wm＝[Zn|Hm]+Y is the connection weight to be solved; whereinPseudo inverse [ Z ]n|Hm]+From ridge regression approximation:
Figure FDA0002982109170000052
i represents a unit vector matrix, and lambda is a hyperparameter;
and obtaining the emotion type corresponding to the electroencephalogram signal according to the output Y of the multi-kernel width learning system.
7. A storage medium storing a computer program, which when executed by a processor causes the processor to perform the method of learning electroencephalogram emotion based on multi-kernel width as recited in any one of claims 1 to 5.
8. A computing device comprising a processor and a memory for storing processor-executable programs, wherein the processor, when executing the programs stored in the memory, implements the multi-kernel width-based learning electroencephalogram emotion classification method of any of claims 1-5.
CN202110290152.XA 2021-03-18 2021-03-18 Electroencephalogram emotion classification method, device, medium and equipment based on multi-kernel width learning Active CN113011493B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110290152.XA CN113011493B (en) 2021-03-18 2021-03-18 Electroencephalogram emotion classification method, device, medium and equipment based on multi-kernel width learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110290152.XA CN113011493B (en) 2021-03-18 2021-03-18 Electroencephalogram emotion classification method, device, medium and equipment based on multi-kernel width learning

Publications (2)

Publication Number Publication Date
CN113011493A true CN113011493A (en) 2021-06-22
CN113011493B CN113011493B (en) 2023-03-21

Family

ID=76409621

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110290152.XA Active CN113011493B (en) 2021-03-18 2021-03-18 Electroencephalogram emotion classification method, device, medium and equipment based on multi-kernel width learning

Country Status (1)

Country Link
CN (1) CN113011493B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113569924A (en) * 2021-07-09 2021-10-29 常州大学 Emotion identification classification method based on support vector machine multi-core cooperation
CN113657479A (en) * 2021-08-12 2021-11-16 广东省人民医院 Novel multi-scale depth-width combined pathological picture classification method, system and medium
CN113855020A (en) * 2021-09-18 2021-12-31 中国信息通信研究院 Method and device for emotion recognition, computer equipment and storage medium
CN114818786A (en) * 2022-04-06 2022-07-29 五邑大学 Channel screening method, emotion recognition method, system and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108520239A (en) * 2018-04-10 2018-09-11 哈尔滨理工大学 A kind of Method of EEG signals classification and system
CN110946576A (en) * 2019-12-31 2020-04-03 西安科技大学 Visual evoked potential emotion recognition method based on width learning
CN111401443A (en) * 2020-03-16 2020-07-10 重庆大学 Width learning system based on multi-feature extraction
CN111461176A (en) * 2020-03-09 2020-07-28 华南理工大学 Multi-mode fusion method, device, medium and equipment based on normalized mutual information
CN111931813A (en) * 2020-07-02 2020-11-13 北京理工大学 CNN-based width learning classification method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108520239A (en) * 2018-04-10 2018-09-11 哈尔滨理工大学 A kind of Method of EEG signals classification and system
CN110946576A (en) * 2019-12-31 2020-04-03 西安科技大学 Visual evoked potential emotion recognition method based on width learning
CN111461176A (en) * 2020-03-09 2020-07-28 华南理工大学 Multi-mode fusion method, device, medium and equipment based on normalized mutual information
CN111401443A (en) * 2020-03-16 2020-07-10 重庆大学 Width learning system based on multi-feature extraction
CN111931813A (en) * 2020-07-02 2020-11-13 北京理工大学 CNN-based width learning classification method

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
FANGYAO SHEN等: "EEG-based emotion recognition using 4D convolutional recurrent neural network", 《COGNITIVE NEURODYNAMICS》 *
XUE-HAN WANG等: "EEG Emotion Recognition Using Dynamical Graph Convolutional Neural Networks and Broad Learning System", 《2018 IEEE INTERNATIONAL CONFERENCE ON BIOINFORMATICS AND BIOMEDICINE (BIBM)》 *
ZHULIN LIU等: "Multi-Kernel Broad Learning systems Based on Random Features:A Novel Expansion for Nonlinear Feature Nodes", 《2019 IEEE INTERNATIONAL CONFERENCE ON SYSTEMS, MAN AND CYBERNETICS (SMC)》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113569924A (en) * 2021-07-09 2021-10-29 常州大学 Emotion identification classification method based on support vector machine multi-core cooperation
CN113569924B (en) * 2021-07-09 2024-03-12 常州大学 Emotion identification classification method based on support vector machine multi-core cooperation
CN113657479A (en) * 2021-08-12 2021-11-16 广东省人民医院 Novel multi-scale depth-width combined pathological picture classification method, system and medium
CN113855020A (en) * 2021-09-18 2021-12-31 中国信息通信研究院 Method and device for emotion recognition, computer equipment and storage medium
CN114818786A (en) * 2022-04-06 2022-07-29 五邑大学 Channel screening method, emotion recognition method, system and storage medium
CN114818786B (en) * 2022-04-06 2024-03-01 五邑大学 Channel screening method, emotion recognition system and storage medium

Also Published As

Publication number Publication date
CN113011493B (en) 2023-03-21

Similar Documents

Publication Publication Date Title
CN113011493B (en) Electroencephalogram emotion classification method, device, medium and equipment based on multi-kernel width learning
CN111012336B (en) Parallel convolutional network motor imagery electroencephalogram classification method based on spatio-temporal feature fusion
Patro et al. An efficient optimized feature selection with machine learning approach for ECG biometric recognition
CN109165556B (en) Identity recognition method based on GRNN
Sadiq et al. Toward the development of versatile brain–computer interfaces
Triastcyn et al. Generating artificial data for private deep learning
CN112101152A (en) Electroencephalogram emotion recognition method and system, computer equipment and wearable equipment
Taqi et al. Classification and discrimination of focal and non-focal EEG signals based on deep neural network
CN112528819B (en) P300 electroencephalogram signal classification method based on convolutional neural network
WO2022183966A1 (en) Electroencephalogram signal classification method and apparatus, device, storage medium and program product
Li et al. Spatial–temporal discriminative restricted Boltzmann machine for event-related potential detection and analysis
Jinliang et al. EEG emotion recognition based on granger causality and capsnet neural network
Miah et al. Movie Oriented Positive Negative Emotion Classification from EEG Signal using Wavelet transformation and Machine learning Approaches
Li et al. A novel motor imagery EEG recognition method based on deep learning
CN112990008B (en) Emotion recognition method and system based on three-dimensional characteristic diagram and convolutional neural network
El-Saadawy et al. Electrocardiogram (ECG) classification based on dynamic beats segmentation
CN116919422A (en) Multi-feature emotion electroencephalogram recognition model establishment method and device based on graph convolution
Wilson et al. Deep riemannian networks for eeg decoding
CN114224354B (en) Arrhythmia classification method, arrhythmia classification device, and readable storage medium
Du et al. EEG-based epileptic seizure detection model using CNN feature optimization
Moreno-Revelo et al. Comparison among physiological signals for biometric identification
CN115631371A (en) Extraction method of electroencephalogram signal core network
Pereira et al. Factor analysis for finding invariant neural descriptors of human emotions
Tseng et al. Sparse Matrix for ECG Identification with Two‐Lead Features
Liu et al. Application of convolutional neural network in automatic classification of arrhythmia

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