CN110432899B - Electroencephalogram signal identification method based on depth stacking support matrix machine - Google Patents
Electroencephalogram signal identification method based on depth stacking support matrix machine Download PDFInfo
- Publication number
- CN110432899B CN110432899B CN201910664430.6A CN201910664430A CN110432899B CN 110432899 B CN110432899 B CN 110432899B CN 201910664430 A CN201910664430 A CN 201910664430A CN 110432899 B CN110432899 B CN 110432899B
- Authority
- CN
- China
- Prior art keywords
- layer
- eeg signal
- eeg
- support matrix
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2576/00—Medical imaging apparatus involving image processing or analysis
- A61B2576/02—Medical imaging apparatus involving image processing or analysis specially adapted for a particular organ or body part
- A61B2576/026—Medical imaging apparatus involving image processing or analysis specially adapted for a particular organ or body part for the brain
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Pathology (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Veterinary Medicine (AREA)
- Psychiatry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Psychology (AREA)
- Power Engineering (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 invention provides an electroencephalogram signal identification method based on a depth stack support matrix machine, which comprises the following steps: firstly, preprocessing an Electroencephalogram (EEG) signal and extracting characteristics; training a first layer Support Matrix Machine (SMM) by using the extracted original EEG signal features as input to obtain predicted output of the first layer; projecting the first layer of prediction output to an original EEG feature space by using matrix random projection, superposing the first layer of prediction output with the original EEG signal feature to obtain a second layer of EEG signal feature, and training a second layer of SMM by using the second layer of EEG signal feature as input to obtain the prediction output of the second layer; deeper EEG signal features are obtained in this manner and SMM is trained until accuracy converges to obtain the final classification model. The invention can accurately judge different types of EEG signals and ensure the safe and reliable operation of the BCI system based on the EEG.
Description
Technical Field
The invention relates to an electroencephalogram signal identification method based on a depth stack support matrix machine, and belongs to the field of electroencephalogram signal identification.
Background
An Electroencephalogram (EEG) imaging method has the characteristics of high time resolution, simplicity in acquisition, noninvasive recording equipment and the like, and is widely used for recording the dynamic activity process of the brain. A reasonable mapping relation is established between EEG data and limb movement intentions through a machine learning method, so that the movement intentions of patients in rehabilitation are accurately detected, and the method is one of important problems to be solved urgently in current research work. Most conventional machine learning algorithms today require that the input features must be in vector form, while the EEG signal features are in matrix form. The traditional machine learning algorithm directly converts the matrix into the vector to be input into the classifier, destroys the structural information in EEG signal characteristics and has certain influence on classification results. In order to solve the problem, researchers have proposed a Support Matrix Machine (SMM) for directly processing data in a Matrix form, and the identification accuracy of EEG signals is improved by introducing a kernel normal form to utilize structural information in Matrix form features. But the method still belongs to a shallow learning method, and has limited representation capability due to the structure of a single hidden layer, and has poor generalization capability when processing a complex classification problem. Therefore, the deep stacking support matrix machine is provided, the support matrix machine and the stacking generalization theory are combined, the structural information in EEG signal features is reserved, meanwhile, the model representation capability is enhanced, and the EEG signals are accurately identified.
Disclosure of Invention
The invention provides an EEG signal identification method based on a depth stack support matrix machine, which adopts an SMM (Single-mode modulation) as a basic module, reserves structural information in EEG signal characteristics, combines a stack generalization theory, introduces matrix random projection as a core stack element, and utilizes the prediction output of a front layer module to help open an original EEG data manifold, thereby having stronger representation capability and realizing the accurate identification of EEG signals.
The invention adopts the following technical scheme for solving the technical problems:
an electroencephalogram signal identification method based on a depth stack support matrix machine comprises the following steps:
step 1: acquiring an EEG signal;
step 2: the EEG signal is pre-processed to remove noise and artifacts, and then feature extracted using CSP algorithm, thereby deriving raw EEG signal featuresWhereinRepresenting the ith EEG signal feature, d1,d2Respectively representing the number of channels of the recording electrode and the number of time sampling points, yi∈ {1, -1} are the corresponding class labels;
and step 3: training a first-layer SMM classification model using the original EEG signal features as input to obtain a first-layer prediction output.
And 4, step 4: and projecting the first layer of prediction output to the original EEG feature space by utilizing matrix random projection, superposing the first layer of prediction output with the original EEG signal features to obtain second layer of EEG signal features, and training the second layer of SMM by taking the second layer of EEG signal features as input to obtain the prediction output of the second layer.
And 5: in this manner, per-layer EEG signal features are generated, and per-layer SMM is trained and a predicted output for each layer is derived.
Step 6: and (5) repeating the step 5 until the precision is converged to obtain the final classification model.
Preferably, the depth-stacking support matrix machine adopts SMM as a basic module, and simultaneously combines the stacking generalization theory, introduces matrix random projection as a core stacking element, and utilizes the prediction output of a front-layer module to help open the original EEG data manifold, thereby having stronger representation capability.
Preferably, the deep stack support matrix machine uses SMM as a basic stack unit, and its target function is:
wherein W and b are classification hyperplane parameters, tr (-) represents the trace of the matrix, | | ·| survival*And C and tau are constraint loss terms and penalty coefficients of the kernel norm.
Preferably, the depth stacking support matrix machine introduces a matrix random projection as a core stacking element to construct each layer of EEG signal features, and the construction formula is:wherein pl,mAnd ql,mFor projection of the m-th preceding layer prediction output o for the l-th layermThe elements in the matrix are sampled from a standard normal distribution N (0, 1).
Has the advantages that:
1. by using the SMM as a basic stacking unit, the EEG characteristics in a matrix form can be directly input for training and prediction, the structural information in the EEG signal is reserved, and the identification accuracy of the EEG signal is improved. The traditional machine learning algorithm based on the vector input form is easy to lose the structural information in the EEG signal at present;
2. the method introduces matrix random projection as a core stacking element to construct each layer of EEG signal characteristics, utilizes the prediction output of a front layer module to help open an original EEG data manifold, and has stronger representation capability, while the traditional shallow learning method has limited representation capability due to the structure of a single hidden layer and poor generalization capability when processing complex classification problems.
Drawings
FIG. 1 is a flow chart of the electroencephalogram signal identification method based on a depth stack support matrix machine in the present invention;
FIG. 2 is a schematic diagram of a matrix stochastic projection technique according to the present invention.
Detailed Description
The present invention will be further explained with reference to examples.
The main implementation process of the invention is as follows, and the related flow chart is shown in figure 1.
Step 1: acquiring an EEG signal;
step 2: the EEG signal is pre-processed, including bandpass filtering and artifact removal, to remove noise and artifacts. Feature extraction is then performed using the CSP algorithm, thereby deriving raw EEG signal featuresWhereinRepresenting the ith EEG signal feature, d1,d2Respectively representing the number of channels of the recording electrode and the number of time sampling points, yi∈ {1, -1} are the corresponding class labels;
and step 3: training the first layer SMM using raw EEG signal features as input to obtain a first layer classification model f1=sgn(tr(WTX) + b), where W and b are classification hyperplane parameters. The solution for W and b is as follows:
1. the following objective function was constructed:
wherein tr (W)TW) represents the trace of the matrix, Z is the equivalent value of W, | Z | | Y*And C and tau are constraint loss terms and penalty coefficients of the kernel norm.
The corresponding augmented Lagrange expression of the above formula is:
whereinQ(Z)=τ||Z||*And β is a penalty factor,is a Lagrange multiplier, | ·| non-conducting phosphorFIs Frobenius norm.
2. And (3) alternately iterating and updating by using an alternate direction multiplier method to obtain hyperplane parameters W and b, wherein the specific method comprises the following steps:
1) initialization Z0=0,M00, β > 0, iterative eigenvalue c1=0,m1Control factor μ ∈ (0,1) at 1.
2) Updating parameters by
Mt+1=Mt-β(Wt+1-Zt+1) (5c)
Wherein the solution of formula (5a) isThe solution of formula (5b) isAndwhereinFor singular value threshold operators, i.e.Wherein{u}+=max(0,u)。
3) Computing characteristic value of t-th iterationIf c ist<μct-1Then, thenOtherwise mt+1=1,Zt+1=Zt-1,Mt+1=Mt-1,ct=μ-1ct-1。
4) t +1, repeating 2) -3) until convergence yields hyperplane parameters W and b.
And 4, step 4: the corresponding prediction output is obtained from the first layer classification model, and then the first layer prediction output is projected into the original EEG feature space using matrix stochastic projection and superimposed with the original EEG signal features to obtain a second layer EEG signal features, i.e.,training the second layer SMM with the SMM as input to obtain a predicted output of the second layer.
And 5: each layer of EEG signal features is generated in this manner. Specifically, the formula for constructing the characteristics of the ith layer EEG signal is as follows:wherein p isl,mAnd q isl,mFor projection of the m-th preceding layer prediction output o for the l-th layermThe elements in the matrix are from N (0),1) And (6) sampling to obtain. Training SMM on newly generated EEG signal features yields the l-th layer classification hyperplane model fl-sgn (tr (W)l TX)+bl)。
Step 6: and (5) repeating the step 5 until the precision is converged to obtain the final classification model.
The above are technical embodiments and technical features of the present invention, which are merely used to illustrate the technical solutions of the present invention and are not limited thereto. Modifications and equivalents of the disclosed embodiments may occur to persons skilled in the art based on the teachings and teachings of the present disclosure. Accordingly, the scope of the present invention should not be limited to the embodiments disclosed, but should include various alternatives and modifications without departing from the invention and encompassed by the appended claims.
Claims (4)
1. An electroencephalogram signal identification method based on a depth stack support matrix machine is characterized by comprising the following steps:
step 1: acquiring an EEG signal;
step 2: preprocessing the EEG signal to remove noise and artifacts, and then feature extracting the EEG signal to obtain n original EEG signal featuresWhereinRepresenting the ith EEG signal feature, d1、d2Respectively representing the number of channels of the recording electrode and the number of time sampling points, yi∈ {1, -1} are the corresponding class labels;
and step 3: training a first layer of SMM model by taking the original EEG signal characteristics as input, obtaining the prediction output of the first layer, calculating the classification precision, obtaining a final classification model if the precision is converged, and otherwise continuing the step 4;
and 4, step 4: projecting the first layer of prediction output to an original EEG feature space by using matrix random projection, superposing the first layer of prediction output with the original EEG signal feature to obtain a second layer of EEG signal feature, and training a second layer of SMM by using the second layer of EEG signal feature as input to obtain the prediction output of the second layer;
and 5: repeating the step 4 to generate EEG signal characteristics of each subsequent layer, training each layer of SMM and obtaining prediction output of each layer until the precision is converged to obtain a final classification model;
step 6: and inputting the EEG signal to be classified into the final classification model to obtain the class label of the EEG signal.
2. The method for recognizing EEG signal based on depth stacking support matrix machine as claimed in claim 1, wherein in step 2, the EEG signal is feature extracted by using the co-space mode.
3. The depth-stack support matrix machine-based electroencephalogram signal identification method according to claim 1, wherein the depth-stack support matrix machine utilizes SMM as a basic stack unit, and the objective function is as follows:
wherein W and b are classification hyperplane parameters, tr (-) represents the trace of the matrix, | | ·| survival*Is a nuclear norm, C and tau are respectively a constraint loss term and a penalty coefficient of the nuclear norm, ξiIs the relaxation variable.
4. The EEG signal identification method based on the depth stacking support matrix machine as claimed in claim 1, wherein the depth stacking support matrix machine introduces matrix random projection as core stacking element to construct EEG signal features of each layer, and the construction formula is:wherein η is weight, l is network layer number, pl,mAnd q isl,mFor projection of the m-th preceding layer prediction output o for the l-th layermThe elements in the matrix are sampled from a standard normal distribution N (0, 1).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910664430.6A CN110432899B (en) | 2019-07-23 | 2019-07-23 | Electroencephalogram signal identification method based on depth stacking support matrix machine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910664430.6A CN110432899B (en) | 2019-07-23 | 2019-07-23 | Electroencephalogram signal identification method based on depth stacking support matrix machine |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110432899A CN110432899A (en) | 2019-11-12 |
CN110432899B true CN110432899B (en) | 2020-07-03 |
Family
ID=68431108
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910664430.6A Active CN110432899B (en) | 2019-07-23 | 2019-07-23 | Electroencephalogram signal identification method based on depth stacking support matrix machine |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110432899B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117338313B (en) * | 2023-09-15 | 2024-05-07 | 武汉纺织大学 | Multi-dimensional characteristic electroencephalogram signal identification method based on stacking integration technology |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101219048B (en) * | 2008-01-25 | 2010-06-23 | 北京工业大学 | Method for extracting brain electrical character of imagine movement of single side podosoma |
CN102542283B (en) * | 2010-12-31 | 2013-11-20 | 北京工业大学 | Optimal electrode assembly automatic selecting method of brain-machine interface |
CN102789316B (en) * | 2012-07-12 | 2015-08-12 | 上海海事大学 | A kind of control method based on the motion of Mental imagery brain-computer interface two dimensional cursor |
-
2019
- 2019-07-23 CN CN201910664430.6A patent/CN110432899B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN110432899A (en) | 2019-11-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107122788B (en) | Identity recognition method and device based on electrocardiosignals | |
Wang et al. | Arrhythmia classification algorithm based on multi-head self-attention mechanism | |
CN109165556B (en) | Identity recognition method based on GRNN | |
CN105069434B (en) | A kind of human action Activity recognition method in video | |
CN107133612A (en) | Based on image procossing and the intelligent ward of speech recognition technology and its operation method | |
CN113065526B (en) | Electroencephalogram signal classification method based on improved depth residual error grouping convolution network | |
CN112488081A (en) | Electroencephalogram mental state detection method based on DDADSM (distributed denial of service) cross-test transfer learning | |
CN110477907B (en) | Modeling method for intelligently assisting in recognizing epileptic seizures | |
CN113128552A (en) | Electroencephalogram emotion recognition method based on depth separable causal graph convolution network | |
CN115294658B (en) | Personalized gesture recognition system and gesture recognition method for multiple application scenes | |
CN110558971A (en) | Method for generating countermeasure network electrocardiogram abnormity detection based on single target and multiple targets | |
CN113010013A (en) | Wasserstein distance-based motor imagery electroencephalogram migration learning method | |
CN110210399A (en) | Face recognition method based on uncertainty quantization probability convolution neural network | |
Yao et al. | Interpretation of electrocardiogram heartbeat by CNN and GRU | |
CN115238796A (en) | Motor imagery electroencephalogram signal classification method based on parallel DAMSCN-LSTM | |
CN115601833A (en) | Myoelectric gesture recognition memory network method and system integrating double-layer attention and multi-stream convolution | |
CN110432899B (en) | Electroencephalogram signal identification method based on depth stacking support matrix machine | |
CN112084935B (en) | Emotion recognition method based on expansion of high-quality electroencephalogram sample | |
CN114428555B (en) | Electroencephalogram movement intention recognition method and system based on cortex source signals | |
CN116821764A (en) | Knowledge distillation-based multi-source domain adaptive EEG emotion state classification method | |
CN109685031B (en) | Electroencephalogram signal feature classification method and system in brain-computer interface | |
CN114983447B (en) | Human action recognition, analysis and storage wearable device based on AI technology | |
CN112932504B (en) | Dipole imaging and identifying method | |
CN114818822A (en) | Electroencephalogram migration emotion recognition method combining semi-supervised regression and icon label propagation | |
CN114209342A (en) | Electroencephalogram signal motor imagery classification method based on space-time characteristics |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |