CN109620223A - A kind of rehabilitation of stroke patients system brain-computer interface key technology method - Google Patents
A kind of rehabilitation of stroke patients system brain-computer interface key technology method Download PDFInfo
- Publication number
- CN109620223A CN109620223A CN201811492294.9A CN201811492294A CN109620223A CN 109620223 A CN109620223 A CN 109620223A CN 201811492294 A CN201811492294 A CN 201811492294A CN 109620223 A CN109620223 A CN 109620223A
- Authority
- CN
- China
- Prior art keywords
- network
- sample
- classification
- class
- motor imagery
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000005516 engineering process Methods 0.000 title claims abstract description 6
- 208000006011 Stroke Diseases 0.000 title abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 34
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 13
- 238000000605 extraction Methods 0.000 claims abstract description 13
- 230000000694 effects Effects 0.000 claims abstract description 8
- 238000001914 filtration Methods 0.000 claims abstract description 8
- 239000011159 matrix material Substances 0.000 claims description 34
- 230000006870 function Effects 0.000 claims description 15
- 238000010606 normalization Methods 0.000 claims description 15
- 230000009466 transformation Effects 0.000 claims description 13
- 239000013598 vector Substances 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 10
- 238000000354 decomposition reaction Methods 0.000 claims description 9
- 210000004556 brain Anatomy 0.000 claims description 5
- 238000013135 deep learning Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 238000012880 independent component analysis Methods 0.000 claims description 3
- 238000002560 therapeutic procedure Methods 0.000 claims description 3
- 230000002087 whitening effect Effects 0.000 claims description 3
- 238000011158 quantitative evaluation Methods 0.000 claims description 2
- 230000033001 locomotion Effects 0.000 abstract description 7
- 230000003340 mental effect Effects 0.000 abstract description 3
- 238000004519 manufacturing process Methods 0.000 abstract 2
- 230000015572 biosynthetic process Effects 0.000 abstract 1
- 230000002490 cerebral effect Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000007659 motor function Effects 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 206010008190 Cerebrovascular accident Diseases 0.000 description 4
- 238000011282 treatment Methods 0.000 description 4
- 206010061296 Motor dysfunction Diseases 0.000 description 3
- 230000004913 activation Effects 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 208000019622 heart disease Diseases 0.000 description 3
- 238000011084 recovery Methods 0.000 description 3
- 241000282414 Homo sapiens Species 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000003042 antagnostic effect Effects 0.000 description 2
- 210000004204 blood vessel Anatomy 0.000 description 2
- 230000034994 death Effects 0.000 description 2
- 231100000517 death Toxicity 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 230000004064 dysfunction Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 230000007433 nerve pathway Effects 0.000 description 2
- 230000036544 posture Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 208000024172 Cardiovascular disease Diseases 0.000 description 1
- 206010020772 Hypertension Diseases 0.000 description 1
- 208000016285 Movement disease Diseases 0.000 description 1
- 208000001738 Nervous System Trauma Diseases 0.000 description 1
- 206010033799 Paralysis Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 230000017531 blood circulation Effects 0.000 description 1
- 210000005013 brain tissue Anatomy 0.000 description 1
- 208000010877 cognitive disease Diseases 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000035475 disorder Diseases 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000000302 ischemic effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000017074 necrotic cell death Effects 0.000 description 1
- 208000028412 nervous system injury Diseases 0.000 description 1
- 230000036403 neuro physiology Effects 0.000 description 1
- 230000000474 nursing effect Effects 0.000 description 1
- 238000011369 optimal treatment Methods 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 230000037361 pathway Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000011514 reflex Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 210000004761 scalp Anatomy 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 208000027765 speech disease Diseases 0.000 description 1
- 230000002195 synergetic effect Effects 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 208000029257 vision disease Diseases 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/389—Electromyography [EMG]
-
- 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
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- 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
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Veterinary Medicine (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Psychiatry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Signal Processing (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present invention discloses a kind of based on the rehabilitation of stroke patients system brain-computer interface key technology method for generating confrontation network, comprising the following steps: step 1: EEG signals pretreatment obtains filtering out the EEG signals after noise；Step 2: improved OVR-CSP algorithm can carry out feature extraction to the multiclass Mental imagery EEG signal after noise is filtered out, and obtain the feature of the Mental imagery EEG signals of every one kind, formation one-dimensional characteristic data, while using variance as the input of classifier；Step 3: utilizing DCGANs network model, wherein convolutional neural networks are added to production model and carry out Further Feature Extraction and classify.Using technical solution of the present invention, convolutional network is re-introduced into production model and does unsupervised training to improve the learning effect for generating network, it realizes the exact classification final to patient motion imagination EEG signals, provides objective data supporting for Measuring scale assessing sufferer rehabilitation degree.
Description
Technical Field
The invention belongs to the technical field of neurology of rehabilitation therapy, and relates to a stroke rehabilitation system brain-computer interface key technical method based on a generation countermeasure network.
Background
Stroke (commonly known as Stroke) is a common cerebral blood circulation disorder disease in which cerebral tissue is damaged due to sudden rupture of cerebral blood vessels or blood failure to flow into the brain caused by blood vessel blockage. The damage of nerve pathways (conduction pathways) of stroke patients caused by ischemic necrosis of brain tissues is clinically manifested as various human dysfunctions: such as movement disorders, vision disorders, and speech and cognitive disorders. Because the morbidity is high, the Chinese medicinal composition has extremely high disability rate and seriously threatens the health of human beings. Heart disease and stroke statistics update in 2018 of the American Heart Association (AHA) showed that stroke death globally accounts for 11.8% of total deaths, second only to heart disease. While 155 cities and rural areas in 31 provinces of mainland China are investigated in households in China, the coarse rate of stroke is 345.1/10 ten thousand years, the morbidity and mortality of cardiovascular diseases are second to that of hypertension, and the heart disease brings heavy mental stress and huge economic burden to the society and families.
Currently, from the analysis of treatment means, stroke rehabilitation systems at home and abroad mainly focus on clinical rehabilitation training systems. The clinical rehabilitation training system mostly adopts the rehabilitation training based on the basis of neurophysiology, mainly utilizes the common movement, the synergistic action, the posture reflex and other nerve movement mechanisms according to the movement development control principle and the brain plasticity principle, judges the functional state and the potential capability of a patient through the clinical rehabilitation evaluation of a doctor, and then performs the corresponding rehabilitation training by 'taking medicines for the disease'. At present, many methods for assessing the motor function of the cerebral apoplexy clinically are available, such as a simplified Fugl-Meyer motor function assessment method, a Brunnstrom grade assessment method and the like. The scale assessment methods all depend on examination and observation of doctors, belong to manual assessment, are widely used clinically, but assessment results are easily affected by subjective factors of rehabilitation doctors, and scale grading indexes are more, so that the rehabilitation doctors need to participate in the whole process, and limited doctors often feel uneasy in the face of huge patient groups, and even the optimal treatment time is delayed. At present, the current state of China is that clinical rehabilitation resources (such as rehabilitation doctors, therapists, nursing staff, beds and the like) are increasingly tense, severe regional imbalance exists, and the medical community has no ability to completely repair the nervous system injury at present. How to solve the problem of the post-illness motor dysfunction of the apoplexy patient through active rehabilitation training is a research hotspot and difficulty point of the current medicine.
Therefore, the research on the brain-computer interface of the high-performance stroke rehabilitation system helps patients who cannot pass through the normal output channel to complete active rehabilitation training, so that the motor function is improved and normal activity is recovered, and the method has very important research significance and development prospect in the crossing field of neurology and information disciplines for stroke treatment in China.
The existing medical research shows that most of patients with cerebral apoplexy and paralysis of the nerve pathways of the brain and limbs are not damaged, so that BCI based on motor imagery can be used for reconstructing damaged cerebral apoplexy areas, namely BCI recovery function uses motor imagery and damaged motor control of nerve feedback to enhance the study of motor control network reconstruction.
A Common Spatial Pattern (CSP) is proved to be one of the most effective methods, other tested electroencephalogram signals are introduced into the CSP learning process by means of the idea of transfer learning, the estimation deviation of the covariance of the tested electroencephalogram signals is guaranteed to be good, and the method is widely applied to small training sample data. But as the training samples increase, the classification accuracy rate of the training samples increases slowly; and the practical application of the algorithm is limited along with the rise of time complexity.
In recent years, with the intensive research of deep learning, many electroencephalogram sample data can be introduced into a convolutional neural network framework for training processing, and a relatively large number of training samples are required. The number of the convolution layers is selected according to the sample size, and once the sample size data is too small, the recognition error rate is greatly improved. For multi-class motor imagery electroencephalogram signals, the sample size is generally small and medium, and the CNN algorithm is difficult to exert the advantages because the deep learning frame is simply introduced and the deep learning frame cannot be fully trained. GoodfellowIan proposes a generation countermeasure network (GANs) in a paper GenetiDeversamental Nets, and provides a new idea for solving the problem of unbalanced distribution of small samples caused by incapability of continuously collecting a large amount of patient data.
In conclusion, the brain-computer interface technology is utilized to identify the motor imagery electroencephalogram signals of the patient, the movement will of the patient can be translated into the control command to drive the rehabilitation device to act, the patient is helped to complete the active rehabilitation training, and the motor function recovery effect is improved. The algorithm for the technology is also increasingly updated, but the application of the algorithm is not mature.
Disclosure of Invention
The invention provides a stroke rehabilitation system brain-computer interface key technical method based on a generation countermeasure network, which aims at the field of rehabilitation training for researching patients with motor dysfunction, realizes an active rehabilitation treatment method for the patients with the dysfunction, accurately judges the motion postures and the limb motion positions of the patients, and provides the method for preprocessing an electroencephalogram signal by adopting wavelet packet transformation and combining a quick independent analysis method, mainly filtering an original motor imagery electroencephalogram signal, reducing various noises in the electroencephalogram signal as much as possible and improving the signal-to-noise ratio. Based on the improved generation confrontation network model DCGANs, the convolutional network is introduced into the generation model again to perform unsupervised training to improve the learning effect of the generation network, and the final accurate classification of the electroencephalogram signals of the motor imagery of the patient is realized.
In order to achieve the purpose, the invention adopts the following technical scheme:
a stroke rehabilitation system brain-computer interface key technical method based on a generation countermeasure network comprises the following steps:
step 1: preprocessing the electroencephalogram signal to obtain the electroencephalogram signal after noise is filtered;
step 2: the improved OVR-CSP algorithm can perform feature extraction on the multi-class motor imagery EEG signals after noise is filtered out to obtain the features of each class of motor imagery EEG signals to form one-dimensional feature data, and meanwhile, the variance is used as the input of a classifier;
and step 3: and (4) utilizing a DCGANs network model, wherein a convolutional neural network is added into the generative model to perform quadratic feature extraction and classification.
The step 1 specifically comprises the following steps: wavelet packet transformation and fast independent component analysis:
firstly, WPT is carried out on an original EEG, the number of decomposition layers of the WPT is determined, a proper wavelet basis function is selected according to the characteristics of the EEG and noise, finally, a frequency band where high-frequency noise is to be filtered is determined, and the corresponding frequency band is set to be zero;
and performing FastICA transformation, FastICA inverse transformation and FastICA inverse transformation on the signal subjected to high-frequency noise filtering to obtain the electroencephalogram signal subjected to noise filtering.
The step 2 specifically comprises the following steps: the OVR-CSP divides four types of motor imagery tasks into four new two types of classification problems to obtain four projection matrixes, and five groups of corresponding airspace characteristics can be obtained after projection;
the specific calculation process is as follows, setting Xi(i is 1,2,3,4) is N x T brain electrical signals of four types of tasks, N is the number of channels for collecting signals, T is the number of sampling points of each channel, T is more than or equal to N, and the basic principle of the OVR-CSP algorithm is to respectively calculate the normalized covariance matrix R of the four types of dataiNamely:
the resulting hybrid spatial covariance matrix R is:
wherein,and (3) carrying out characteristic value decomposition on R for an average covariance matrix of multiple experiments of four types of tasks to obtain:
R＝UVUT
wherein U is an eigenvector of R, and V is an eigenvalue matrix of R; and (3) performing descending arrangement on the eigenvalue matrix, and performing the same adjustment on the eigenvectors according to the sorted positions, so that the whitening matrix P is as follows:
P＝V-1/2UT
the OVR-CSP classifies one of the categories as one category and the remaining four categories as another category when calculating the projection matrix, namely:
the preprocessed EEG signal X is divided into two new classes, X is1，X1' projecting in the projection direction in the i-th mode to obtain:
Z1＝(U1′)TP1X1,Z1′＝(U1′)TP1X1′(i＝1,2,3,4)
the covariance matrix value of the matrix after projection of the four types of data is
Then, normalizing the eigenvector of the covariance matrix to obtain:
wherein n is the number of columns of the feature vector,
the feature vectors are used for classification learning.
The step 3 specifically comprises the following steps: a deep convolution generation countermeasure Network (DCGANs) for classification of motor imagery electroencephalogram signals introduces a Convolutional Neural Network (CNN) into a GANs structure for the first time, and the powerful feature extraction capability of a Convolutional layer is utilized to improve the effect of the GANs. The DCGANs architecture is named EEGANs network architecture. In the network structure of EEGGANs, a generated network G is a four-layer structure, four-dimensional reshaping from a random noise vector (1X 100) is realized by using fractional-distorted convolution (Deconv), and then the random noise vector is sent into a generator to be gradually upsampled to a pseudo sample Xfake(64X 1). In the traditional CNN network feature mapping structure, a Sigmoid function is adopted as an activation function of a convolutional network, a modified linear unit (ReLU) function is adopted in the invention to enable the trained network to have appropriate sparsity, and meanwhile, the problem that the traditional activation function possibly produces in the process of tuning back propagation parameters can be well solvedThe resulting gradient disappears, speeding up the convergence of the network. Each upsampled layer performs a transposed convolution operation with step size 2 using a convolution kernel of size 5 x 5. Its depth gradually decreases from 512 to 64 and then to 1 by RPCA. The last layer outputs a 64 x 64 tensor output via RPCA dimensionality reduction and compresses the value between-1 and 1 using the Tanh function.
The discrimination network D is also a 4-layer CNN with BN (except for an input layer), and in order to solve the problem that the network cannot carry out back propagation due to the disappearance of the gradient of the ReLU, a Leaky ReLU is adopted for activation. Finally, the discriminator needs to output the probability and finally classifies the probability by using a Softmax function.
The network training data needs normalization processing:
in the training process of DCGANs, Batch Normalization (BN) is generally adopted to force the data to be pulled back to a normal distribution with a mean value of 0 and a variance of 1, so as to avoid the disappearance of the gradient. But BN is sensitive to the size of the batch _ size, and if the batch _ size is too small, the calculated mean, variance are not sufficient to represent the entire data distribution.
IN order to solve the problems, the invention is inspired by three current normalization modes, namely BN (batch normalization), IN (example normalization) and LN (layer normalization), and provides an Adaptive normalization (Adaptive Norm) method to improve the generalization capability of the model and avoid the problem of local replacement of the whole body caused by adopting the same normalization method for solving different problems. Unlike reinforcement learning, the AN determines the appropriate normalization operation for each normalization layer in a deep network using differentiable learning.
Drawings
FIG. 1 technical route diagram proposed by the present invention
FIG. 2 is a schematic diagram of a motor imagery-based electroencephalogram signal feature extraction method provided by the invention.
FIG. 3 is a schematic diagram of wavelet packet decomposition of electroencephalogram signals.
FIG. 4 is a schematic diagram of preprocessing of electroencephalogram signals according to the present invention.
FIG. 5 illustrates a semi-supervised learning model EEGGANs network architecture proposed by the present invention.
Fig. 6 shows a network structure of the generation network G of the EEGGANs according to the present invention.
Detailed Description
As shown in fig. 2, the invention adopts a stroke rehabilitation system brain-computer interface key technical method based on generation of an antagonistic network, combines and improves two algorithms of CSP and DCGANs, and performs secondary feature extraction and classification on motor imagery electroencephalogram signals. Compared with the method of directly inputting the original electroencephalogram signals, the method not only increases the discrimination between the signals, but also solves the problem of unbalanced sample resources by using the DCGANs, and finally outputs a plurality of motor imagery classifications. As shown in fig. 1, a stroke rehabilitation system brain-computer interface key technology method based on generation of an antagonistic network includes the following steps:
step 1: preprocessing an electroencephalogram signal, comprising two processes of wavelet packet transformation and quick independent component analysis:
firstly, WPT is carried out on original motor imagery electroencephalogram information, the decomposition layer number of the WPT is determined, when high-frequency noise in the electroencephalogram signals is filtered through wavelet packet transformation, the enhancement or attenuation frequency range of a power spectrum representing an electroencephalogram signal ERD/ERS phenomenon is mainly reflected in 8-30 Hz, therefore, the decomposition layer number of the WPT can be determined to be five layers for a sampling frequency of 128Hz, and the frequency range is determined to be 8-30 Hz through wavelet packet decomposition. Finally, the frequency band where the high-frequency noise is to be filtered is determined, and the corresponding frequency band is set to zero, as shown in fig. 3, which is a schematic diagram of wavelet packet decomposition.
Then, performing FastICA transformation on the signal subjected to high-frequency noise filtering to obtain independent components of the EEG signals of all channels, calculating the correlation coefficients of the independent components and the C3 and C4 channels of the original EEG signals respectively, reserving the independent components with the correlation coefficients higher than a set threshold value, and performing FastICA inverse transformation to obtain the EEG signals subjected to noise filtering. The WPT algorithm and the FastICA algorithm preprocess the electroencephalogram signals, noise and partial interference signals of the electroencephalogram signals are filtered, and subsequent feature extraction and classification are facilitated.
And 2, performing feature extraction on the multi-class motor imagery EEG signals by using an OVR-CSP algorithm.
Assuming that the number of projection matrixes required to be solved by the W-type signals is W, after each sample is projected, W groups of space domain results can be obtained, the results are the characteristics of the extracted motor imagery electroencephalogram signals of each type, and finally, the variance is used as the input of the classifier. The extracted features may form one-dimensional feature data. The convolutional neural network structure is greatly simplified: not only can the discrimination between EEG signals be increased, but also the data size of the convolutional neural network input samples can be reduced. Fig. 4 shows feature extraction of four classes of motor imagery tasks (Data sets IIIa Data sets are left hand, right hand, foot and tongue, respectively) based on OVR-CSP.
The OVR-CSP divides four types of motor imagery tasks into four new two types of classification problems to obtain four projection matrixes, and four groups of corresponding spatial domain characteristics can be obtained after projection. The specific calculation process is as follows, setting XiAnd (i is 1,2,3 and 4) is the N x T electroencephalogram signals of the four tasks, N is the number of channels for collecting signals, T is the number of sampling points of each channel, and T is more than or equal to N. The basic principle of the OVR-CSP algorithm is to calculate the normalized covariance matrix R of four types of data respectivelyiNamely:
the resulting hybrid spatial covariance matrix R is:
whereinThe mean covariance matrix of multiple experiments for four types of tasks. And (3) decomposing the characteristic value of R to obtain:
R＝UVUT
wherein U is a feature vector of R; v is a matrix of eigenvalues of R. And (3) performing descending arrangement on the eigenvalue matrix, and performing the same adjustment on the eigenvectors according to the sorted positions, so that the whitening matrix P is as follows:
P＝V-1/2UT
unlike the classic two classes of CSP algorithms, OVR-CSP classifies one of them into one class and the remaining three classes into another class when computing the projection matrix, namely:
the preprocessed EEG signal X is divided into two new classes, X is1，X1' projecting in the projection direction in the i-th mode to obtain:
Z1＝(U1′)TP1X1,Z1′＝(U1′)TP1X1′(i＝1,2,3,4)
the covariance matrix value of the matrix after projection of the four types of data isThen, normalizing the eigenvector of the covariance matrix to obtain:
(n is the number of columns of the feature vector)
The feature vectors are used for classification learning.
And step 3: and performing secondary feature extraction and classification by using the modified DCGANs network.
The invention designs a DCGANs structure for classifying motor imagery electroencephalogram signals, which is named as an EEGGANs network structure, and the EEGGANs network structure is shown as figures 5 and 6. The network comprises a network generation network G and a discrimination network D, wherein the network generation network G and the discrimination network D are both constructed by adopting 4 layers of CNN models, and the last layer 1 is a Softmax classification layer. When the semi-supervised classification task is realized, the structure of the DCGANs needs to be changed to a certain extent, namely, an output layer of a D network is replaced by a Softmax classifier. Assuming that the training data has K classes, when the semi-supervised learning model is trained, the G network generation samples are classified into K +1 th class, and the Softmax classifier also adds an output neuron for representing the probability that the input of the D network is false data. This model is called "semi-supervised" classification because it can be learned from unlabeled generated data as well as from labeled training samples. In addition, the subject adds condition extension to the condition model, and if both G-network and D-network are suitable for some extra condition c (e.g. some kind of label y), then the data generation process can be guided by adding c to the network for adjustment.
The network structure has a decision network D as a Classifier (Classifier) and a generation network G for generating pseudo samples X from random noisefakeThe training set contains a labeled sample XlabelUnlabeled sample XunlabelAnd a dummy sample Xfake. Wherein XlabelComprises a BCI competition data set and an EEG data set with a label of a real patient; xunlabelA real EEG data set representing a patient under-fit or a physician unlabeled category. Classifier accepts samplesFor the classification problem of K (K ═ 30), outputting a K + 1-dimensional estimation, and obtaining a probability set P through a Softmax function: the first K dimension corresponds to the original motor imagery category and the last one dimensionCorresponding to the "pseudo sample" class, piCorresponds to the class estimation label yi. The optimization function of the system is as follows
For accurately calculating the class output of the motor imagery, three loss functions are defined for the subject, and a labeled sample X in a training set is subjected tolabelCalculating the probability of whether the estimated label is correct, and recording as Llabel。
For unlabeled sample X in training setunlabelWhether the estimated value is 'true' or not is examined, namely, the probability that the estimated value is not K +1 class is calculated and is marked as Lunlabel。
For the pseudo sample X generated by the generatorfakeWhether or not to estimate as "false" is examined. I.e. calculating the probability of estimating as K +1 class, denoted as Lfake。
It is assumed that the input data of the EEGGANs can be represented as a feature map having three dimensions, where each dimension represents the number of samples N, the sample latitude H, and the sample latitude W, respectively. Assume that each element in the feature map is denoted as hnijThen the normalized value of the output is recorded as
Where γ and β represent the scaling factor and the offset factor, respectivelyk、w′kCorresponding to the weighting coefficients when BN, IN and LN are adopted.
In the formula, λkFor controlling the parameter, the initial value is 1, and the control parameter lambda is controlled by using the Softmax functionkIs normalized, i.e. sigmak∈Ωwk1. And a self-adaptive normalization mode is adopted, so that the system can learn a proper normalization mode in a self-adaptive manner according to the number of samples and the batch size, and high-precision mathematical expression is kept.
In addition, in order to accelerate network convergence, a weight epsilon is introduced into the classifier discrimination network D to obtain a classifier optimization target:
wherein f isi，jTraining a sample for the input motor imagery features, namely feature vectors of a covariance matrix after OVR-CSP normalization; y is a class label (determined class) corresponding to the sample; f (f)i，j) Is the network output through the EEGGANs. The output layer then begins backpropagating to adjust the weights of the convolution kernels in the network. Finally, through EEGGANs training, L is outputDValue of (A), LDThe smaller the value of (A), the better the training effect is, and the more accurate the classification result is. f (f)i，j) The output of (1) is the classification of left, right hand, tongue, foot. And the classification result of the motor imagery is used as a BCI control command to be transmitted to the upper and lower mechanical limbs to be used as a quantitative evaluation index. Compared with the evaluation index only based on the experience of the doctor, the method provides objective data support and feeds back the data support to the rehabilitation doctor. Rehabilitation doctorAnd continuously updating the evaluation model to be used as a new method for guiding treatment, and performing effective rehabilitation training on the stroke patient until the patient is helped to recover early.
In conclusion, in the process of extracting the EEG characteristics of the scalp electroencephalogram signals, the concept of patches is introduced to decompose the signals into small data fragments to form a covariance matrix, and then the classification characteristic vectors of the motor imagery electroencephalogram signals based on the OVR-CSP method are extracted through matrix estimation; after the high-dimensional feature vectors are mapped to a low-dimensional space, a high-robustness semi-supervised learning model EEGGANs is designed to carry out motor imagery accurate classification. Therefore, the system is applied to the field of rehabilitation training of patients with motor dysfunction, helps the patients to participate in active rehabilitation training, is beneficial to improving the motor function recovery effect, and provides objective data support for rating the rehabilitation degree of the patients.
Claims (4)
1. A stroke rehabilitation system brain-computer interface key technical method based on a generation countermeasure network is characterized by comprising the following steps:
step 1: preprocessing the electroencephalogram signal to obtain the electroencephalogram signal after noise is filtered;
step 2: the improved OVR-CSP algorithm can perform feature extraction on the multi-class motor imagery EEG signals after noise is filtered out to obtain the features of each class of motor imagery EEG signals to form one-dimensional feature data, and meanwhile, the variance is used as the input of a classifier;
and step 3: and (4) utilizing a DCGANs network model, wherein a convolutional neural network is added into the generative model to perform quadratic feature extraction and classification.
2. The brain-computer interface-based co-spatial mode and deep learning method for assisted rehabilitation therapy according to claim 1, wherein the step 1 specifically comprises: wavelet packet transformation and fast independent component analysis:
firstly, WPT is carried out on an original EEG, the number of decomposition layers of the WPT is determined, a proper wavelet basis function is selected according to the characteristics of the EEG and noise, finally, a frequency band where high-frequency noise is to be filtered is determined, and the corresponding frequency band is set to be zero;
and performing FastICA transformation, FastICA inverse transformation and FastICA inverse transformation on the signal subjected to high-frequency noise filtering to obtain the electroencephalogram signal subjected to noise filtering.
3. The brain-computer interface-based co-spatial mode and deep learning method for assisted rehabilitation therapy according to claim 2, wherein the step 2 is specifically as follows: the OVR-CSP divides four types of motor imagery tasks into four new two types of classification problems to obtain four projection matrixes, and five groups of corresponding airspace characteristics can be obtained after projection;
the specific calculation process is as follows, setting Xi(i is 1,2,3,4) is N x T brain electrical signals of four types of tasks, N is the number of channels for collecting signals, T is the number of sampling points of each channel, T is more than or equal to N, and the basic principle of the OVR-CSP algorithm is to respectively calculate the normalized covariance matrix R of the four types of dataiNamely:
the resulting hybrid spatial covariance matrix R is:
wherein,and (3) carrying out characteristic value decomposition on R for an average covariance matrix of multiple experiments of four types of tasks to obtain:
R＝UVUT
wherein U is an eigenvector of R, and V is an eigenvalue matrix of R; and (3) performing descending arrangement on the eigenvalue matrix, and performing the same adjustment on the eigenvectors according to the sorted positions, so that the whitening matrix P is as follows:
P＝V-1/2UT
the OVR-CSP classifies one of the categories as one category and the remaining four categories as another category when calculating the projection matrix, namely:
the preprocessed EEG signal X is divided into two new classes, X is1，X1' projecting in the projection direction in the i-th mode to obtain:
Z1＝(U1′)TP1X1,Z1′＝(U1′)TP1X1′(i＝1,2,3,4)
the covariance matrix value of the matrix after projection of the four types of data is RZi＝ZiZi T(i＝1,2,3,4),
Then, normalizing the eigenvector of the covariance matrix to obtain:
wherein n is the number of columns of the feature vector,
the feature vectors are used for classification learning.
4. The stroke rehabilitation system brain-computer interface key technology method based on generation of the countermeasure network as claimed in claim 1, wherein step 3 is specifically: designing a DCGANs structure for classifying motor imagery electroencephalogram signals, wherein the DCGANs structure is named as an EEGGANs network structure, the network generation network G and the judgment network D are composed, the network generation network G and the judgment network D are both constructed by adopting 4 layers of CNN models, and the last layer 1 is a Softmax classification layer;
the network structure has a decision network D as a Classifier (Classifier) and a generation network G for generating pseudo samples X from random noisefakeThe training set contains a labeled sample XlabelUnlabeled sample XunlabelAnd a dummy sample XfakeWherein X islabelComprises a BCI competition data set and an EEG data set with a label of a real patient; xunlabelA real EEG data set representing a patient under-fit or a physician unlabeled category. Classifier accepts samplesFor the classification problem of K (K ═ 30), outputting a K + 1-dimensional estimation, and obtaining a probability set P through a Softmax function: the first K dimension corresponds to the original motor imagery class, and the last one dimension corresponds to the 'pseudo sample' class, piCorresponds to the class estimation label yiThe optimization function of the system is as follows:
for accurately calculating the class output of the motor imagery, three loss functions are defined, and for labeled samples X in a training setlabelCalculating the probability of whether the estimated label is correct, and recording as Llabel，
For unlabeled sample X in training setunlabelWhether the estimated value is 'true' or not is examined, namely, the probability that the estimated value is not K +1 class is calculated and is marked as Lunlabel，
For the pseudo sample X generated by the generatorfakeWhether the estimation is false or not is examined, namely the probability of estimating as K +1 class is calculated and is marked as Lfake，
It is assumed that the input data of EEGGANs can be represented as a feature map having three dimensions, where each dimension represents the number of samples N, the sample latitude H, and the sample latitude W, respectively, and each element in the feature map is represented as HnijThen the normalized value of the output is recorded as
Where γ and β represent the scaling factor and the offset factor, respectively, wk、w′kCorresponding to the weighting coefficients when BN, IN and LN are adopted.
In the formula, λkFor controlling the parameter, the initial value is 1, and the control parameter lambda is controlled by using the Softmax functionkIs normalized, i.e. sigmak∈Ωwk1, adopting an adaptive normalization mode to lead the system to adaptively learn a proper normalization mode according to the number of samples and the batch size,
the network convergence is accelerated, and the weight epsilon is introduced into the classifier discrimination network D to obtain the optimization target of the classifier:
wherein f isi，jTraining samples for input motor imagery features, namely covariance matrix after OVR-CSP normalizationThe feature vector of (2); y is a category label corresponding to the sample; f (f)i，j) Is the network output through the EEGGANs. Then the output layer starts to reversely propagate and adjust the weight of the convolution kernel in the network, finally the L is output through the training of EEGGANsDValue of (A), LDThe smaller the value of (f) is, the better the training effect is, the more accurate the classification result is, f (f)i，j) The output of the motor imagery is the classification of the left hand, the right hand, the tongue and the foot, and the classification result of the motor imagery is used as a BCI control command to be transmitted to the upper and lower mechanical limbs to be used as a quantitative evaluation index.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811492294.9A CN109620223A (en) | 2018-12-07 | 2018-12-07 | A kind of rehabilitation of stroke patients system brain-computer interface key technology method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811492294.9A CN109620223A (en) | 2018-12-07 | 2018-12-07 | A kind of rehabilitation of stroke patients system brain-computer interface key technology method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109620223A true CN109620223A (en) | 2019-04-16 |
Family
ID=66071938
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811492294.9A Pending CN109620223A (en) | 2018-12-07 | 2018-12-07 | A kind of rehabilitation of stroke patients system brain-computer interface key technology method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109620223A (en) |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110251137A (en) * | 2019-06-05 | 2019-09-20 | 长沙湖湘医疗器械有限公司 | A kind of sleep detection method for noninvasive ventilator and the ventilator using this method |
| CN110338906A (en) * | 2019-07-10 | 2019-10-18 | 清华大学深圳研究生院 | Smart therapeutics system and method for building up for photo-crosslinking operation |
| CN110379506A (en) * | 2019-06-14 | 2019-10-25 | 杭州电子科技大学 | The cardiac arrhythmia detection method of binaryzation neural network is used for ECG data |
| CN110531861A (en) * | 2019-09-06 | 2019-12-03 | 腾讯科技（深圳）有限公司 | The treating method and apparatus and storage medium of Mental imagery EEG signals |
| CN110569727A (en) * | 2019-08-06 | 2019-12-13 | 华南理工大学 | Transfer learning method combining intra-class distance and inter-class distance based on motor imagery classification |
| CN110888526A (en) * | 2019-10-30 | 2020-03-17 | 西安理工大学 | Brain-computer interface technology based on integrated support vector learning |
| CN110895964A (en) * | 2019-12-12 | 2020-03-20 | 浙江迈联医疗科技有限公司 | Electroencephalogram-based rehabilitation training assessment method and system |
| CN111000555A (en) * | 2019-11-29 | 2020-04-14 | 中山大学 | Training data generation method, automatic recognition model modeling method and automatic recognition method for epilepsia electroencephalogram signals |
| CN111523520A (en) * | 2020-06-11 | 2020-08-11 | 齐鲁工业大学 | Method for analyzing electroencephalogram signals of brain patients with motor imagery stroke by using cycleGAN |
| CN111544856A (en) * | 2020-04-30 | 2020-08-18 | 天津大学 | Brain-myoelectricity intelligent full limb rehabilitation method based on novel transfer learning model |
| CN111990992A (en) * | 2020-09-03 | 2020-11-27 | 山东中科先进技术研究院有限公司 | Electroencephalogram-based autonomous movement intention identification method and system |
| CN112370066A (en) * | 2020-09-30 | 2021-02-19 | 北京工业大学 | Brain-computer interface method of stroke rehabilitation system based on generation of countermeasure network |
| CN112545532A (en) * | 2020-11-26 | 2021-03-26 | 中国人民解放军战略支援部队信息工程大学 | Data enhancement method and system for classification and identification of electroencephalogram signals |
| CN112560964A (en) * | 2020-12-18 | 2021-03-26 | 深圳赛安特技术服务有限公司 | Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning |
| CN112884062A (en) * | 2021-03-11 | 2021-06-01 | 四川省博瑞恩科技有限公司 | Motor imagery classification method and system based on CNN classification model and generation countermeasure network |
| CN112932505A (en) * | 2021-01-16 | 2021-06-11 | 北京工业大学 | Symbol transfer entropy and brain network characteristic calculation method based on time-frequency energy |
| CN112971786A (en) * | 2021-02-05 | 2021-06-18 | 郑州大学 | Apoplexy rehabilitation evaluation method based on brain electromyographic signal wavelet coherence coefficient |
| CN113128384A (en) * | 2021-04-01 | 2021-07-16 | 北京工业大学 | Brain-computer interface software key technical method of stroke rehabilitation system based on deep learning |
| CN113499524A (en) * | 2021-07-23 | 2021-10-15 | 华南理工大学 | Auxiliary rehabilitation training system using motor imagery electroencephalogram detection |
| CN114082169A (en) * | 2021-11-22 | 2022-02-25 | 江苏科技大学 | Disabled hand soft body rehabilitation robot motor imagery identification method based on electroencephalogram signals |
| CN114343673A (en) * | 2021-11-29 | 2022-04-15 | 北京机械设备研究所 | Cross-tested motor imagery electroencephalogram signal processing method, medium and equipment |
| CN115034272A (en) * | 2022-06-29 | 2022-09-09 | 上海术理智能科技有限公司 | Motor imagery electroencephalogram signal classification method based on iterative learning |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160128592A1 (en) * | 2014-11-11 | 2016-05-12 | Hyperfine Research, Inc. | Low field magnetic resonance methods and apparatus |
| CN107392973A (en) * | 2017-06-06 | 2017-11-24 | 中国科学院自动化研究所 | Pixel-level handwritten Chinese character automatic generation method, storage device, processing unit |
| CN107958213A (en) * | 2017-11-20 | 2018-04-24 | 北京工业大学 | A kind of cospace pattern based on the medical treatment of brain-computer interface recovering aid and deep learning method |
| WO2018141061A1 (en) * | 2017-02-01 | 2018-08-09 | Cerebian Inc. | System and method for measuring perceptual experiences |
| CN108564039A (en) * | 2018-04-16 | 2018-09-21 | 北京工业大学 | A kind of epileptic seizure prediction method generating confrontation network based on semi-supervised deep layer |
| CN108665058A (en) * | 2018-04-11 | 2018-10-16 | 徐州工程学院 | A kind of generation confrontation network method based on segmentation loss |
-
2018
- 2018-12-07 CN CN201811492294.9A patent/CN109620223A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160128592A1 (en) * | 2014-11-11 | 2016-05-12 | Hyperfine Research, Inc. | Low field magnetic resonance methods and apparatus |
| WO2018141061A1 (en) * | 2017-02-01 | 2018-08-09 | Cerebian Inc. | System and method for measuring perceptual experiences |
| CN107392973A (en) * | 2017-06-06 | 2017-11-24 | 中国科学院自动化研究所 | Pixel-level handwritten Chinese character automatic generation method, storage device, processing unit |
| CN107958213A (en) * | 2017-11-20 | 2018-04-24 | 北京工业大学 | A kind of cospace pattern based on the medical treatment of brain-computer interface recovering aid and deep learning method |
| CN108665058A (en) * | 2018-04-11 | 2018-10-16 | 徐州工程学院 | A kind of generation confrontation network method based on segmentation loss |
| CN108564039A (en) * | 2018-04-16 | 2018-09-21 | 北京工业大学 | A kind of epileptic seizure prediction method generating confrontation network based on semi-supervised deep layer |
Non-Patent Citations (4)
| Title |
|---|
| KAY GREGOR HARTMANN, ROBIN TIBOR SCHIRRMEISTER, TONIO BALL: "EEG-GAN: Generative adversarial networks for electroencephalograhic (EEG) brain signals", 《ELECTRICAL ENGINEERING AND SYSTEMS SCIENCE》 * |
| S. PALAZZO, C. SPAMPINATO, I. KAVASIDIS, D. GIORDANO AND M. SHAH: "Generative Adversarial Networks Conditioned by Brain Signals", 《2017 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV)》 * |
| 田晓霞: "运动想象EEG的识别方法及在上肢康复中的应用", 《中国优秀硕士学位论文全文数据库》 * |
| 罗平、任家敏、彭章琳: "深度剖析 | 可微分学习的自适配归一化", 《知乎》 * |
Cited By (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110251137A (en) * | 2019-06-05 | 2019-09-20 | 长沙湖湘医疗器械有限公司 | A kind of sleep detection method for noninvasive ventilator and the ventilator using this method |
| CN110379506A (en) * | 2019-06-14 | 2019-10-25 | 杭州电子科技大学 | The cardiac arrhythmia detection method of binaryzation neural network is used for ECG data |
| CN110379506B (en) * | 2019-06-14 | 2022-06-14 | 杭州电子科技大学 | Arrhythmia detection method using binarization neural network for electrocardiogram data |
| CN110338906A (en) * | 2019-07-10 | 2019-10-18 | 清华大学深圳研究生院 | Smart therapeutics system and method for building up for photo-crosslinking operation |
| CN110569727A (en) * | 2019-08-06 | 2019-12-13 | 华南理工大学 | Transfer learning method combining intra-class distance and inter-class distance based on motor imagery classification |
| CN110569727B (en) * | 2019-08-06 | 2022-03-29 | 华南理工大学 | Transfer learning method combining intra-class distance and inter-class distance for motor imagery classification |
| CN110531861A (en) * | 2019-09-06 | 2019-12-03 | 腾讯科技（深圳）有限公司 | The treating method and apparatus and storage medium of Mental imagery EEG signals |
| CN110888526A (en) * | 2019-10-30 | 2020-03-17 | 西安理工大学 | Brain-computer interface technology based on integrated support vector learning |
| CN110888526B (en) * | 2019-10-30 | 2023-04-07 | 西安理工大学 | Brain-computer interface technology based on integrated support vector learning |
| CN111000555B (en) * | 2019-11-29 | 2022-09-30 | 中山大学 | Training data generation method, automatic recognition model modeling method and automatic recognition method for epileptic electroencephalogram signals |
| CN111000555A (en) * | 2019-11-29 | 2020-04-14 | 中山大学 | Training data generation method, automatic recognition model modeling method and automatic recognition method for epilepsia electroencephalogram signals |
| CN110895964A (en) * | 2019-12-12 | 2020-03-20 | 浙江迈联医疗科技有限公司 | Electroencephalogram-based rehabilitation training assessment method and system |
| CN110895964B (en) * | 2019-12-12 | 2022-11-22 | 浙江迈联医疗科技有限公司 | Electroencephalogram-based rehabilitation training assessment method and system |
| CN111544856A (en) * | 2020-04-30 | 2020-08-18 | 天津大学 | Brain-myoelectricity intelligent full limb rehabilitation method based on novel transfer learning model |
| CN111523520A (en) * | 2020-06-11 | 2020-08-11 | 齐鲁工业大学 | Method for analyzing electroencephalogram signals of brain patients with motor imagery stroke by using cycleGAN |
| CN111990992A (en) * | 2020-09-03 | 2020-11-27 | 山东中科先进技术研究院有限公司 | Electroencephalogram-based autonomous movement intention identification method and system |
| CN112370066A (en) * | 2020-09-30 | 2021-02-19 | 北京工业大学 | Brain-computer interface method of stroke rehabilitation system based on generation of countermeasure network |
| CN112545532B (en) * | 2020-11-26 | 2023-05-16 | 中国人民解放军战略支援部队信息工程大学 | Data enhancement method and system for electroencephalogram signal classification and identification |
| CN112545532A (en) * | 2020-11-26 | 2021-03-26 | 中国人民解放军战略支援部队信息工程大学 | Data enhancement method and system for classification and identification of electroencephalogram signals |
| CN112560964A (en) * | 2020-12-18 | 2021-03-26 | 深圳赛安特技术服务有限公司 | Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning |
| CN112932505A (en) * | 2021-01-16 | 2021-06-11 | 北京工业大学 | Symbol transfer entropy and brain network characteristic calculation method based on time-frequency energy |
| CN112971786A (en) * | 2021-02-05 | 2021-06-18 | 郑州大学 | Apoplexy rehabilitation evaluation method based on brain electromyographic signal wavelet coherence coefficient |
| CN112884062A (en) * | 2021-03-11 | 2021-06-01 | 四川省博瑞恩科技有限公司 | Motor imagery classification method and system based on CNN classification model and generation countermeasure network |
| CN112884062B (en) * | 2021-03-11 | 2024-02-13 | 四川省博瑞恩科技有限公司 | Motor imagery classification method and system based on CNN classification model and generated countermeasure network |
| CN113128384A (en) * | 2021-04-01 | 2021-07-16 | 北京工业大学 | Brain-computer interface software key technical method of stroke rehabilitation system based on deep learning |
| CN113128384B (en) * | 2021-04-01 | 2024-04-05 | 北京工业大学 | Brain-computer interface software key technical method of cerebral apoplexy rehabilitation system based on deep learning |
| CN113499524A (en) * | 2021-07-23 | 2021-10-15 | 华南理工大学 | Auxiliary rehabilitation training system using motor imagery electroencephalogram detection |
| CN114082169A (en) * | 2021-11-22 | 2022-02-25 | 江苏科技大学 | Disabled hand soft body rehabilitation robot motor imagery identification method based on electroencephalogram signals |
| CN114343673B (en) * | 2021-11-29 | 2024-03-29 | 北京机械设备研究所 | Method, medium and equipment for processing brain wave signals crossing tested motor imagery |
| CN114343673A (en) * | 2021-11-29 | 2022-04-15 | 北京机械设备研究所 | Cross-tested motor imagery electroencephalogram signal processing method, medium and equipment |
| CN115034272A (en) * | 2022-06-29 | 2022-09-09 | 上海术理智能科技有限公司 | Motor imagery electroencephalogram signal classification method based on iterative learning |
| CN115034272B (en) * | 2022-06-29 | 2024-06-28 | 上海术理智能科技有限公司 | Motor imagery electroencephalogram signal classification method based on iterative learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109620223A (en) | A kind of rehabilitation of stroke patients system brain-computer interface key technology method | |
| CN107958213A (en) | A kind of cospace pattern based on the medical treatment of brain-computer interface recovering aid and deep learning method | |
| Xu et al. | Wavelet transform time-frequency image and convolutional network-based motor imagery EEG classification | |
| Jiao et al. | Sparse group representation model for motor imagery EEG classification | |
| Alkan et al. | Identification of EMG signals using discriminant analysis and SVM classifier | |
| Chatterjee et al. | Ensemble learning approach to motor imagery EEG signal classification | |
| Abbas et al. | DeepMI: Deep learning for multiclass motor imagery classification | |
| CN112370066A (en) | Brain-computer interface method of stroke rehabilitation system based on generation of countermeasure network | |
| Ieracitano et al. | A hybrid-domain deep learning-based BCI for discriminating hand motion planning from EEG sources | |
| Chen et al. | A discriminant bispectrum feature for surface electromyogram signal classification | |
| CN110013248A (en) | Brain electricity tensor mode identification technology and brain-machine interaction rehabilitation system | |
| Sun et al. | Golden subject is everyone: A subject transfer neural network for motor imagery-based brain computer interfaces | |
| Rejer | EEG feature selection for BCI based on motor imaginary task | |
| Yue et al. | Exploring BCI control in smart environments: intention recognition via EEG representation enhancement learning | |
| Xu et al. | EEG decoding method based on multi-feature information fusion for spinal cord injury | |
| Li et al. | Multi-view fusion network-based gesture recognition using sEMG data | |
| CN116522106A (en) | Motor imagery electroencephalogram signal classification method based on transfer learning parallel multi-scale filter bank time domain convolution | |
| Al-dabag et al. | EEG motor movement classification based on cross-correlation with effective channel | |
| Zou et al. | An inter-subject model to reduce the calibration time for motion imagination-based brain-computer interface | |
| CN114305453A (en) | Multi-source manifold electroencephalogram feature transfer learning method | |
| Wang et al. | Deep convolutional neural network for decoding EMG for human computer interaction | |
| CN110321856B (en) | Time-frequency multi-scale divergence CSP brain-computer interface method and device | |
| Yu et al. | The research of sEMG movement pattern classification based on multiple fused wavelet function | |
| Bhalerao et al. | FBSE-based automated classification of motor imagery EEG signals in brain–computer interface | |
| Rammy et al. | Recurrent deep learning for eeg-based motor imagination recognition |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20240115 Address after: 310007 station 30, 2nd floor, east area, building a, Zhejiang University Science Park, No. 525 Xixi Road, Xihu District, Hangzhou, Zhejiang Province Applicant after: Hangzhou Xingyuan intelligent Biotechnology Co.,Ltd. Address before: 100124 No. 100 Chaoyang District Ping Tian Park, Beijing Applicant before: Beijing University of Technology Applicant before: Wang Zhuozheng Effective date of registration: 20240115 Address after: 100124 No. 100 Chaoyang District Ping Tian Park, Beijing Applicant after: Beijing University of Technology Applicant after: Wang Zhuozheng Address before: 100124 No. 100 Chaoyang District Ping Tian Park, Beijing Applicant before: Beijing University of Technology |