CN113988123A - Electroencephalogram fatigue prediction method based on self-weighted increment RVFL network - Google Patents

Electroencephalogram fatigue prediction method based on self-weighted increment RVFL network Download PDF

Info

Publication number
CN113988123A
CN113988123A CN202111212547.4A CN202111212547A CN113988123A CN 113988123 A CN113988123 A CN 113988123A CN 202111212547 A CN202111212547 A CN 202111212547A CN 113988123 A CN113988123 A CN 113988123A
Authority
CN
China
Prior art keywords
formula
rvfl
electroencephalogram
theta
network
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111212547.4A
Other languages
Chinese (zh)
Inventor
陈柯锭
张怿恺
彭勇
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.)
Hangzhou Dianzi University
Original Assignee
Hangzhou Dianzi University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hangzhou Dianzi University filed Critical Hangzhou Dianzi University
Priority to CN202111212547.4A priority Critical patent/CN113988123A/en
Publication of CN113988123A publication Critical patent/CN113988123A/en
Pending legal-status Critical Current

Links

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
    • G06F2218/12—Classification; Matching
    • 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
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

The invention provides an electroencephalogram fatigue prediction method based on a self-weighted increment RVFL network. The invention comprises the following steps: 1. and a plurality of testees respectively acquire electroencephalogram data under the simulated driving system. 2. And (3) preprocessing and extracting the characteristics of all the electroencephalogram data obtained in the step (1). 3. A self-weighted RVFL network is built for off-line training of a base fatigue prediction model. 4. Solving the output weight and characteristic weight distribution 5, collecting the on-line electroencephalogram data, and performing the same steps as the step 2. 6. And converting the fatigue prediction model after the off-line training into an incremental on-line fatigue prediction model. 7. And solving the output weight and the characteristic weight distribution. 8. And predicting and learning the on-line electroencephalogram data. The prediction accuracy of the model is improved through the modes of regression prediction, incremental learning and self-weighted variables.

Description

Electroencephalogram fatigue prediction method based on self-weighted increment RVFL network
Technical Field
The invention belongs to the technical field of electroencephalogram signal processing, and particularly relates to an electroencephalogram fatigue prediction method based on a self-weighted incremental RVFL network.
Background
With the increasing development of the traffic industry of China, the research of more practical and objective driving fatigue detection has important significance for improving the active traffic safety. The comprehensive domestic and foreign fatigue detection method mainly comprises subjective evaluation and objective detection. The subjective evaluation method is mainly used for judging whether the patient is in a fatigue state or not by recording subjective questionnaires such as a Pearson fatigue scale, a Stanford sleep scale and the like. The field of objective detection is mainly divided into: detection based on vehicle behavior characteristics, detection based on driver characteristic behavior characteristics, detection based on physiological electrical characteristics. The electroencephalogram signal in the objective detection method is used as the representation of the central nervous signal activity, the method has the characteristic of high detection accuracy, and the method is a gold standard for fatigue judgment.
Most of the current electroencephalogram fatigue models are batch processing modes, the streaming attribute of electroencephalogram data is not considered, and the performance is reduced due to the change of the tested fatigue state after the electroencephalogram fatigue models are used for a period of time; and the traditional fatigue prediction model is discrete in prediction result, and the fatigue prediction is only simple in a plurality of states, so that the tested fatigue degree cannot be accurately displayed. Therefore, the invention provides an electroencephalogram fatigue prediction method based on a self-weighted increment RVFL network on the basis of a Random Vector function Link network (RVFL). The method introduces an incremental learning mode so that the model can be added with acquired electroencephalogram data at any time to update the model, the prediction of the model is a regression mode, a continuous fatigue value can more accurately display the tested fatigue degree, in addition, the method also introduces an self-weighting variable to learn the importance distribution of each neuron, the influence of RVFL network randomization mapping on the model performance is reduced, and the model is ensured to have higher prediction accuracy.
Disclosure of Invention
The invention aims to provide an electroencephalogram fatigue prediction method based on a self-weighted increment RVFL network. By the method, regression prediction of fatigue and incremental learning of the model can be realized.
The method comprises the following specific steps:
step 1, carrying out electroencephalogram data acquisition on a plurality of testees in the process of gradual fatigue.
And 2, preprocessing and extracting characteristics of all the electroencephalogram data obtained in the step 1. Each processed group of data is taken as a sample matrix X. Each sample matrix X corresponds to a label vector y; the label vector y corresponds to the degree of fatigue of the subject.
And 3, establishing a self-weighted RVFL network for offline training.
Step 3-1, establishing a self-weighted RVFL network objective function as shown in formula (1):
Figure BDA0003309423260000021
in the formula,
Figure BDA0003309423260000022
is a data representation activated by an RVFL network hidden layer, wherein n represents the number of samples, and d represents a feature dimension;
Figure BDA0003309423260000023
is the output weight vector of the RVFL network;
Figure BDA0003309423260000024
a fatigue value vector representing training data;
Figure BDA0003309423260000025
representing a feature weight distribution vector; theta is a diagonal matrix and the ith diagonal element is thetai(ii) a λ is the regularization coefficient;
Figure BDA0003309423260000026
the square of the 2-norm of the vector is represented by
Figure BDA0003309423260000027
And step 4, solving the formula (1) to obtain updated formulas of theta and beta.
And 5, acquiring on-line electroencephalogram data, and performing electroencephalogram data preprocessing and feature extraction in the same way as the step 2.
And 6, establishing a self-weighted incremental RVFL network for online training.
Step 6-1, establishing a self-weighted incremental RVFL network objective function for online training as shown in (2);
Figure BDA0003309423260000028
in the formula, DnAnd ynRespectively representing n sample data and corresponding tags. Thetan+1And betan+1The partial table represents the feature weight distribution and the output weight obtained by training n +1 samples.
And 7, solving the formula (2) to obtain the theta after incremental learningn+1And betan+1The update formula of (2).
And 8, predicting the fatigue value of the real-time electroencephalogram data according to the learned theta and beta parameters, and performing incremental learning on the data if a real label of the real-time electroencephalogram data can be obtained after prediction.
Preferably, in step 1, the subject is gradually fatigued by continuing the simulated driving on a straight and monotonous road using the simulated driving platform.
Preferably, the label vector y in step 2 is determined by a time length proportion of the eyelid closure to a certain degree per unit time in the electroencephalogram data acquisition process of the subject. Typically, the elements in the tag vector y range from 0 to 1.
Preferably, the pretreatment process in step 2 is as follows:
step 2-1, down-sampling the electroencephalogram data to 200Hz, and performing band-pass filtering on the electroencephalogram data to a range of 1-50 Hz; according to the 5-frequency band method, the method is divided into five frequency bands of Delta, Theta, Alpha, Beta and Gamma
Step 2-2, respectively carrying out short-time Fourier transform with 4 seconds of time window and no overlap on the electroencephalogram data of the 5 frequency bands, extracting differential entropy characteristics h (X) as shown in formula (3),
h(X)＝-∫xf(x)ln(f(x))dx， (3)
in the formula (3), X is an input sample matrix, and X is an element in the input sample matrix; f (x) is a probability density function. The updated differential entropy characteristic h (X) is shown as formula (4):
Figure BDA0003309423260000031
in the formula (4), σ is a standard deviation of the probability density function; μ is the expectation of the probability density function.
Preferably, 17 leads are adopted for the electroencephalogram data acquisition, and 5 frequency bands are selected; the 5 frequency bands are respectively 1-4Hz, 4-8Hz, 8-14Hz, 14-31Hz and 31-50 Hz.
Preferably, the specific solving process for Θ and β in step 4 is as follows:
step 4-1, fixing theta and updating beta
In this case, (1) becomes:
Figure BDA0003309423260000032
an updated formula that derives equation (5) with respect to β and let the derivative be 0 to β:
Figure BDA0003309423260000033
step 4-2, fixing beta and updating theta
In this case, (1) becomes:
Figure BDA0003309423260000034
(7) the objective function of equation is written as:
Figure BDA0003309423260000035
wherein, denotes a matrix dot product; let the symmetric matrix
Figure BDA0003309423260000036
Vector b 2diag (D)0y0βT) And (7) the formula becomes:
Figure BDA0003309423260000037
(9) the formula is a convex quadratic programming problem, and theta is obtained according to the augmented Lagrange method solution (ALM).
Preferably, step 7 is performed on the theta after incremental learningn+1And betan+1The specific solving process of (2) is as follows:
step 7-1. fixing thetan+1Updating betan+1；
In this case, the formula (2) is:
Figure BDA0003309423260000041
according to the formula (6), the following components are obtained:
Figure BDA0003309423260000042
order to
Figure BDA0003309423260000043
Then the following simplification occurs:
Figure BDA0003309423260000044
according to the formula Woodbury:
Figure BDA0003309423260000045
(11) the right half of the formula is simplified as follows:
Figure BDA0003309423260000046
based on the formulae (11), (13) and (14), beta is obtainedn+1Incremental update ofFormula (II):
Figure BDA0003309423260000047
step 7-2. fixing betan+1Update thetan+1
In this case, the formula (2) is:
Figure BDA0003309423260000048
(16) the objective function of equation:
Figure BDA0003309423260000051
order to
Figure BDA0003309423260000052
Then there are:
Figure BDA0003309423260000053
and
Figure BDA0003309423260000054
similarly, the updated theta is obtained according to the ALM methodn+1。
The invention has the beneficial effects that:
1. the method adopts an incremental learning mode, fully utilizes the streaming attribute of the electroencephalogram data, updates the model while predicting the fatigue value, and enables the model to better conform to the tested fatigue state change, thereby improving the model prediction precision.
2. The invention is regression fatigue prediction, and has the advantages of more intuition and more accuracy compared with discrete fatigue prediction.
3. The invention introduces the adaptive weight variable, can abstractly learn the importance of each neuron in the RVFL network, reduces the influence of random mapping on the model performance, and improves the performance of the whole model.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic diagram of an embedding process of the self-weighting variable Θ;
figure 3 is a graph of the root mean square error of the model RVFL, delta RVFL, self-weighted RVFL, and self-weighted delta RVFL over 23 trials.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
The invention solves two important problems of stream type attribute and importance characteristic mining of the electroencephalogram signal in fatigue detection based on the following starting points:
in fatigue detection, electroencephalogram signals are streaming data, the data sequentially reach a model, the internal distribution of the tested data may be changed continuously in the process, and if the model can be updated continuously in the prediction process so as to adapt to the internal distribution of the tested continuously-changed data, the prediction accuracy of the model can be effectively improved. Since each neuron in the RVFL network still has a characteristic importance difference on data in training, the performance of the model can be effectively improved by introducing the self-weighting variable to learn the importance of each neuron.
The electroencephalogram fatigue prediction method based on the self-weighted increment RVFL network is shown in figures 1 and 2, and the RVFL, the increment RVFL, the self-weighted RVFL and the root mean square error comparison graph of the self-weighted increment RVFL on 23 tested persons are shown in figure 3, so that the model is obviously improved compared with other three models, and the problem that the performance of the increment RVFL and the self-weighted RVFL on electroencephalogram data is unstable compared with the traditional RVFL is solved.
The following describes our model in detail with reference to the accompanying drawings.
Step 1, acquiring electroencephalogram data of the induced fatigue state of a subject by using a simulated driving system (in the embodiment, a simulated driving scene containing obvious induced fatigue of the subject is used).
The method comprises the steps of carrying out M times of electroencephalogram data acquisition on N subjects in the same simulated driving environment to obtain N.M groups of electroencephalogram data, wherein the data volume of each group of data is d x N, d is the dimensionality of each group of data, and N is the number of electroencephalogram data samples which are acquired at a single time and are related to time. The set of data includes electroencephalogram data for a plurality of time instants obtained in one acquisition. Each set of data is taken as a sample matrix X. Each sample matrix X corresponds to a label y; the label y corresponds to the fatigue degree of the testee, and is determined by the percentage of the time length of the closed eyes of the testee in the electroencephalogram data acquisition process to the total acquisition time length. Longer duration of eye closure indicates more fatigue in the subject.
Unifying the corresponding tested task scenes of the tested subjects in the testing task, wherein each tested subject completes the testing of all the tested task scenes. In this embodiment, the task scene to be tested is a straight and monotonous road, and the experiment time is performed after lunch, so that the subject is more fatigued. The content of these different batches of experiments is the same, but their status at different dates may be different; this reflects that the essential features of the brain electricity will not change greatly with the change of time or experimental scenes. These differences in time, or in the scene during the experiment, are the differentiation components.
And 2, preprocessing and extracting characteristics of all the electroencephalogram data obtained in the step 1. The method is carried out on the basis of 17-lead and 5-frequency bands (Delta (1-4Hz), Theta (4-8Hz), Alpha (8-14Hz), Beta (14-31Hz) and Gamma (31-50Hz)) and by extracting differential entropy characteristics. In practical application, the number of leads depends on the electroencephalogram cap worn by a subject during data acquisition; the division of frequency bands also follows a physiologically meaningful 5-band division; the most common features of electroencephalographic signals are power spectral density and differential entropy. The electroencephalogram signal of a human being is very weak, which means that the electroencephalogram signal is easy to interfere, and the acquired result is difficult to directly carry out experiments, so that the requirements on the preprocessing of the electroencephalogram signal are provided:
the pretreatment process is as follows:
and 2-1, down-sampling the electroencephalogram data to 200Hz, and performing band-pass filtering on the electroencephalogram data to a range of 1-50 Hz. According to the 5-frequency band method, the method is divided into five frequency bands of Delta, Theta, Alpha, Beta and Gamma.
And 2-2, respectively taking the electroencephalogram data of the 5 frequency bands as sample matrixes, respectively carrying out short-time Fourier transform with 4 seconds of time window and no overlap, and extracting differential entropy characteristics. The differential entropy signature h (x) is defined as:
h(X)＝-∫xf(x)ln(f(x))dx (20)
in the formula (20), X is an input sample matrix, and X is an element in the input sample matrix; f (x) is a probability density function. The updated differential entropy characteristic h (x) is shown as formula (21):
Figure BDA0003309423260000071
in the formula (21), σ is a standard deviation of the probability density function; μ is the expectation of the probability density function.
It can be seen that the differential entropy signature is essentially a logarithmic form of the power spectral density signature, i.e.
Figure BDA0003309423260000072
The preprocessing of the electroencephalogram signals aims to improve the signal-to-noise ratio, so that the preprocessing effect of data is improved, and interference is reduced.
And 3, establishing a self-weighted RVFL network to off-line train the existing electroencephalogram data so as to obtain a basic electroencephalogram fatigue prediction model. The importance of each neuron in the RVFL network is measured by introducing an autoweighting variable Θ ═ diag (θ), and the influence of randomization on the model performance is reduced. Wherein, thetaiRepresenting the importance distribution of the ith neuron and having a constraint thetaT1. Embedding the self-weighting variables into the RVFL network results in a self-weighted RVFL network as shown in equation (22).
Figure BDA0003309423260000073
In the formula,
Figure BDA0003309423260000074
is a data representation activated by an RVFL network hidden layer, wherein n represents the number of samples, and d represents a feature dimension;
Figure BDA0003309423260000075
is the output weight vector of the RVFL network;
Figure BDA0003309423260000076
a fatigue value vector representing training data;
Figure BDA0003309423260000077
representing a feature weight distribution vector; theta is a diagonal matrix and the ith diagonal element is thetai(ii) a λ is the regularization coefficient;
Figure BDA0003309423260000078
the square of the 2-norm of the vector is expressed by a specific calculation method
Figure BDA0003309423260000079
And 4, optimizing the model based on a Lagrangian method and solving an update formula of theta and beta.
Step 4-1, fixing theta and updating beta
At this time, the expression (22) may be changed to:
Figure BDA0003309423260000081
an updated formula for β can be obtained by deriving equation (23) with respect to β and letting the derivative be 0:
Figure BDA0003309423260000082
step 4-2, fixing beta and updating theta
At this time, the expression (22) may be changed to:
Figure BDA0003309423260000083
(25) the objective function of equation (la) can be written as:
Figure BDA0003309423260000084
where, o represents the matrix dot multiplication. Let the symmetric matrix
Figure BDA0003309423260000085
Vector b 2diag (D)0y0βT) (25) the formula may be changed as follows:
Figure BDA0003309423260000086
(27) the formula is a convex quadratic programming problem, and theta can be obtained according to augmented Lagrange method solving (ALM).
And 5, acquiring online electroencephalogram data, and performing preprocessing and feature extraction in the same way as in the step 2.
And 6, changing the basic model after offline training into an online model, namely changing the self-weighted RVFL network into a self-weighted incremental RVFL network, as shown in a formula (28):
Figure BDA0003309423260000087
in the formula, DnAnd ynRespectively representing n sample data and corresponding tags. Thetan+1And betan+1The partial table represents the feature weight distribution and the output weight obtained by training n +1 samples.
And 7, solving the updating formulas of the variables theta and beta.
Step 7-1. fixing thetan+1Updating betan+1
At this time, the expression (28) may be changed to:
Figure BDA0003309423260000091
according to the formula (24), it is possible to obtain:
Figure BDA0003309423260000092
order to
Figure BDA0003309423260000093
Then the following simplification occurs:
Figure BDA0003309423260000094
from the Woodbury formula, one can obtain:
Figure BDA0003309423260000095
(30) the right half of the formula can be simplified as follows:
Figure BDA0003309423260000096
based on the formulae (30), (32) and (33), beta can be obtainedn+1Incremental update of (c):
Figure BDA0003309423260000097
step 7-2. fixing betan+1Update thetan+1
At this time, the expression (28) may be changed to:
Figure BDA0003309423260000098
(35) the objective function of equation:
Figure BDA0003309423260000101
order to
Figure BDA0003309423260000102
Then there are:
Figure BDA0003309423260000103
and
Figure BDA0003309423260000104
similarly, the updated Θ can be obtained according to the ALM methodn+1。
And 8, predicting the fatigue value of the real-time electroencephalogram data according to the learned parameters such as theta, beta and the like, and if the real label of the real-time electroencephalogram data can be obtained after prediction, performing incremental learning on the data.

Claims (7)

1. An electroencephalogram fatigue prediction method based on a self-weighted increment RVFL network is characterized by comprising the following steps:
step 1, carrying out electroencephalogram data acquisition on a plurality of testees in a gradual fatigue process;
step 2, preprocessing and feature extraction are carried out on all the electroencephalogram data obtained in the step 1; each processed group of data is used as a sample matrix X; each sample matrix X corresponds to a label vector y; the label vector y corresponds to the fatigue degree of the testee;
step 3, establishing a self-weighted RVFL network for offline training;
step 3-1, establishing a self-weighted RVFL network objective function as shown in formula (1):
Figure FDA0003309423250000011
in the formula,
Figure FDA0003309423250000012
is a data representation activated by an RVFL network hidden layer, wherein n represents the number of samples, and d represents a feature dimension;
Figure FDA0003309423250000013
is the output weight vector of the RVFL network;
Figure FDA0003309423250000014
a fatigue value vector representing training data;
Figure FDA0003309423250000015
representing a feature weight distribution vector; theta is a diagonal matrix and the ith diagonal element is thetai(ii) a λ is the regularization coefficient;
Figure FDA0003309423250000016
the square of the 2-norm of the vector is represented by
Figure FDA0003309423250000017
Step 4, solving the formula (1) to obtain updated formulas of theta and beta;
step 5, collecting on-line electroencephalogram data, and preprocessing and feature extraction of the electroencephalogram data;
step 6, establishing a self-weighted incremental RVFL network for on-line training;
step 6-1, establishing a self-weighted incremental RVFL network objective function for online training as shown in (2);
Figure FDA0003309423250000018
in the formula, DnAnd ynRespectively representing n sample data and corresponding labels; thetan+1And betan+1The sublist represents the feature weight distribution and output weight obtained by training n +1 samples;
and 7, solving the formula (2) to obtain the theta after incremental learningn+1And betan+1The update formula of (2);
and 8, predicting the fatigue value of the real-time electroencephalogram data according to the learned theta and beta parameters, and performing incremental learning on the data if a real label of the real-time electroencephalogram data can be obtained after prediction.
2. The electroencephalogram fatigue prediction method based on the self-weighted incremental RVFL network as claimed in claim 1, characterized in that: in step 1, the subject is gradually fatigued by continuously simulating driving on a straight and monotonous road by using the simulated driving platform.
3. The electroencephalogram fatigue prediction method based on the self-weighted incremental RVFL network as claimed in claim 1, characterized in that: in the step 2, the label vector y is determined by the proportion of the duration of the eyelid closure to a certain degree in a unit time of the testee in the process of acquiring the electroencephalogram data; the range of elements in the label vector y is 0-1.
4. The electroencephalogram fatigue prediction method based on the self-weighted incremental RVFL network as claimed in claim 1, characterized in that: the pretreatment process in step 2 is as follows;
step 2-1, down-sampling the electroencephalogram data to 200Hz, and performing band-pass filtering on the electroencephalogram data to a range of 1-50 Hz; according to the 5-frequency band method, the method is divided into five frequency bands of Delta, Theta, Alpha, Beta and Gamma
Step 2-2, respectively carrying out short-time Fourier transform with 4 seconds of time window and no overlap on the electroencephalogram data of the 5 frequency bands, extracting differential entropy characteristics h (X) as shown in formula (3),
h(X)＝-∫xf(x)ln(f(x))dx， (3)
in the formula (3), x is an input sample matrix, and x is an element in the input sample matrix; (x) is a probability density function; the updated differential entropy characteristic h (X) is shown as formula (4):
Figure FDA0003309423250000021
in the formula (4), σ is a standard deviation of the probability density function; μ is the expectation of the probability density function.
5. The electroencephalogram fatigue prediction method based on the self-weighted incremental RVFL network as claimed in claim 1, characterized in that: 17 leads are adopted for the electroencephalogram data acquisition, and 5 frequency bands are selected; the 5 frequency bands are respectively 1-4Hz, 4-8Hz, 8-14Hz, 14-31Hz and 31-50 Hz.
6. The electroencephalogram fatigue prediction method based on the self-weighted incremental RVFL network as claimed in claim 1, characterized in that: the concrete solving process of theta and beta in the step 4 is as follows;
step 4-1, fixing theta and updating beta
In this case, (1) becomes:
Figure FDA0003309423250000022
an updated formula that derives equation (5) with respect to β and let the derivative be 0 to β:
Figure FDA0003309423250000023
step 4-2, fixing beta and updating theta
In this case, equation (6) is:
Figure FDA0003309423250000031
(7) the objective function of equation is written as:
Figure FDA0003309423250000032
wherein,
Figure FDA0003309423250000033
representing a matrix dot product; let the symmetric matrix
Figure FDA0003309423250000034
Vector b 2diag (D)0y0βT) And (7) the formula becomes:
Figure FDA0003309423250000035
(9) the formula is a convex quadratic programming problem, and theta is obtained according to the augmented Lagrange method solution (ALM).
7. The electroencephalogram fatigue prediction method based on the self-weighted incremental RVFL network as claimed in claim 1, characterized in that: theta after incremental learning in step 7n+1And betan+1The specific solving process of (2) is as follows:
step 7-1, fixing thetan+1Updating betan+1
In this case, the formula (2) is:
Figure FDA0003309423250000036
according to the formula (6):
Figure FDA0003309423250000037
order to
Figure FDA0003309423250000038
Then the following simplification occurs:
Figure FDA0003309423250000039
according to the Woodbury formula, the following results are obtained:
Figure FDA00033094232500000310
(11) the right half of the formula is simplified as follows:
Figure FDA0003309423250000041
based on the formulae (11), (13) and (14), beta is obtainedn+1Incremental update of (c):
Figure FDA0003309423250000042
step 7-2, immobilizing betan+1Update thetan+1
In this case, the formula (2) is:
Figure FDA0003309423250000043
(16) the objective function of equation:
Figure FDA0003309423250000044
order to
Figure FDA0003309423250000045
Then there are:
Figure FDA0003309423250000046
and
Figure FDA0003309423250000047
similarly, the updated theta is obtained according to the ALM methodn+1。
CN202111212547.4A 2021-10-19 2021-10-19 Electroencephalogram fatigue prediction method based on self-weighted increment RVFL network Pending CN113988123A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111212547.4A CN113988123A (en) 2021-10-19 2021-10-19 Electroencephalogram fatigue prediction method based on self-weighted increment RVFL network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111212547.4A CN113988123A (en) 2021-10-19 2021-10-19 Electroencephalogram fatigue prediction method based on self-weighted increment RVFL network

Publications (1)

Publication Number Publication Date
CN113988123A true CN113988123A (en) 2022-01-28

Family

ID=79739212

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111212547.4A Pending CN113988123A (en) 2021-10-19 2021-10-19 Electroencephalogram fatigue prediction method based on self-weighted increment RVFL network

Country Status (1)

Country Link
CN (1) CN113988123A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116746931A (en) * 2023-06-15 2023-09-15 中南大学 Incremental driver bad state detection method based on brain electricity
CN117290781A (en) * 2023-10-24 2023-12-26 中汽研汽车检验中心(宁波)有限公司 Driver KSS grade self-evaluation training method for DDAW system test
CN117717309A (en) * 2023-12-25 2024-03-19 上海雄博精密仪器股份有限公司 Method, device and storage medium for detecting human eye higher-order aberration

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110946576A (en) * 2019-12-31 2020-04-03 西安科技大学 Visual evoked potential emotion recognition method based on width learning
WO2020151144A1 (en) * 2019-01-24 2020-07-30 五邑大学 Generalized consistency-based fatigue classification method for constructing brain function network and relevant vector machine
CN112773378A (en) * 2021-01-20 2021-05-11 杭州电子科技大学 Electroencephalogram emotion recognition method for feature weight adaptive learning

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020151144A1 (en) * 2019-01-24 2020-07-30 五邑大学 Generalized consistency-based fatigue classification method for constructing brain function network and relevant vector machine
CN110946576A (en) * 2019-12-31 2020-04-03 西安科技大学 Visual evoked potential emotion recognition method based on width learning
CN112773378A (en) * 2021-01-20 2021-05-11 杭州电子科技大学 Electroencephalogram emotion recognition method for feature weight adaptive learning

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
YONG PENG等: "A joint optimization framework to semi-supervised RVFL and ELM networks for efficient data classification", APPLIED SOFT COMPUTING JOURNAL, 31 December 2020 (2020-12-31), pages 1 - 10 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116746931A (en) * 2023-06-15 2023-09-15 中南大学 Incremental driver bad state detection method based on brain electricity
CN116746931B (en) * 2023-06-15 2024-03-19 中南大学 Incremental driver bad state detection method based on brain electricity
CN117290781A (en) * 2023-10-24 2023-12-26 中汽研汽车检验中心(宁波)有限公司 Driver KSS grade self-evaluation training method for DDAW system test
CN117717309A (en) * 2023-12-25 2024-03-19 上海雄博精密仪器股份有限公司 Method, device and storage medium for detecting human eye higher-order aberration

Similar Documents

Publication Publication Date Title
CN113988123A (en) Electroencephalogram fatigue prediction method based on self-weighted increment RVFL network
CN112656427B (en) Electroencephalogram emotion recognition method based on dimension model
CN114052735A (en) Electroencephalogram emotion recognition method and system based on depth field self-adaption
CN111714118B (en) Brain cognition model fusion method based on ensemble learning
CN108960182A (en) A kind of P300 event related potential classifying identification method based on deep learning
CN113662545B (en) Personality assessment method based on emotion electroencephalogram signals and multitask learning
CN110090017A (en) A kind of EEG signals source localization method based on LSTM
CN105868532B (en) A kind of method and system of intelligent evaluation heart aging degree
CN109255366A (en) A kind of affective state regulating system for on-line study
CN111772629B (en) Brain cognitive skill transplanting method
CN112185493A (en) Personality preference diagnosis device and project recommendation system based on same
CN112438741A (en) Driving state detection method and system based on electroencephalogram feature transfer learning
CN114366103B (en) Attention assessment method and device and electronic equipment
CN113974627B (en) Emotion recognition method based on brain-computer generated confrontation
CN117883082A (en) Abnormal emotion recognition method, system, equipment and medium
Lopes et al. Ensemble deep neural network for automatic classification of EEG independent components
Cao et al. Recognition of cognitive load with a stacking network ensemble of denoising autoencoders and abstracted neurophysiological features
CN109271894A (en) A kind of product image recognition methods based on EEG signals and fuzzy reasoning
CN117171658A (en) Cognitive load judging method based on brain intelligent technology
CN116662736A (en) Human body state assessment method based on deep learning hybrid model
Xiaohong et al. A Hybrid Macro and Micro Method for Consumer Emotion and Behavior Research
CN116369945A (en) Electroencephalogram cognitive recognition method based on 4D pulse neural network
CN113143275B (en) Electroencephalogram fatigue detection method for quantitative evaluation of sample and characteristic quality in combined manner
Guo et al. Research on classification and recognition of emotional EEG signal based on extreme learning machine
Ma Neurophysiological analysis of the genesis mechanism of EEG during the interictal and ictal periods using a multiple neural masses model

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