CN105718953B - A kind of single P300 detection method based on matrix Motar transport - Google Patents

A kind of single P300 detection method based on matrix Motar transport Download PDF

Info

Publication number
CN105718953B
CN105718953B CN201610045616.XA CN201610045616A CN105718953B CN 105718953 B CN105718953 B CN 105718953B CN 201610045616 A CN201610045616 A CN 201610045616A CN 105718953 B CN105718953 B CN 105718953B
Authority
CN
China
Prior art keywords
matrix
feature
data
carried out
sample
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
Application number
CN201610045616.XA
Other languages
Chinese (zh)
Other versions
CN105718953A (en
Inventor
张娟丽
谢松云
刘畅
段绪
谢辛舟
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.)
Northwestern Polytechnical University
Original Assignee
Northwestern Polytechnical 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 Northwestern Polytechnical University filed Critical Northwestern Polytechnical University
Priority to CN201610045616.XA priority Critical patent/CN105718953B/en
Publication of CN105718953A publication Critical patent/CN105718953A/en
Application granted granted Critical
Publication of CN105718953B publication Critical patent/CN105718953B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Investigating Or Analysing Biological Materials (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

The single P300 detection method based on matrix Motar transport that the invention proposes a kind of, belong to Cognitive Neuroscience field, it is related to the feature extraction and recognition detection method of a kind of event-related potential N400, is specifically related to a kind of single P300 detection method based on the matrix Motar transport in gray theory.The described method includes: 1, acquired original EEG signals are pre-processed；2, lead combination selection selects top most apparent 4 leads of occipital region different wave shape as optimal electrode according to the waveform diagram of training set data goal stimulus and non-targeted stimulation；3, segmentation matrix Motar transport is carried out to the data of 4 leads, and extracts model parameter as feature vector；4, optimal feature selection is carried out using the method for Fisher rate value, while achievees the purpose that reduce feature vector dimension；5, classified using support vector machine classifier to feature vector, realize the detection identification of single P300.Show that the algorithm can be improved the detection discrimination of single detection P300 through experimental data test, and in secondary superposition less, correct recognition rata can be further enhanced.

Description

A kind of single P300 detection method based on matrix Motar transport
Technical field
The invention belongs to Cognitive Neuroscience fields, are related to the feature extraction and identification inspection of a kind of event-related potential N400 Survey method is specifically related to a kind of single P300 detection method based on the matrix Motar transport in gray theory.
Background technique
P300 (also referred to as P3b) is the endogenous component that can reflect higher cognitive treatment process, it is distinguished in subject It, can be by vision, the sense of hearing and body-sensing in the maximum advanced stage positive wave that the incubation period that scalp is recorded is about 300ms when recognizing " target stimulation " Induced by Stimulation.
P300 current potential is widely used in brain-computer interface system (Brain Computer Interface, BCI) and psychology Detect a lie in research.Traditional P300 recognition methods is the method average using time-domain coherence, can be effective by superposed average Reduce background random noise.The method can identify P300 well, but there are apparent defects: multiple averaging be easy to make by Fatigue is felt in examination, and has ignored the different differences tried between time ERP, it is most important that is not able to satisfy the requirement of real-time, no Conducive to the practical application of P300.
Researchers propose the feature extraction of few time/single P300 of accomplished in many ways, such as based on the knowledge of wavelet transformation Other algorithm, the recognizer based on independent component analysis, recognizer based on principal component analysis etc..But these methods are all It has some limitations, the detection accuracy of single-trial extraction P300 is not very high.
Gray system theory (Grey System Theory) is to be taught in nineteen eighty-two by Chinese scholar Deng Julong in the world On first propose.Place particularly suitable for the signal (such as non-stationary, non-gaussian distribution, nonwhite noise) to atypia rule Reason, compared with some other methods by statistical law and priori rule to handle data, data volume needed for gray method is small, It is not required to the apparent advantage such as priori knowledge.And the Motar transport towards matrix is the further expansion to traditional grey modeling, it will Point sequence modeling extends to face, i.e. matrix sequence, not only contains temporal information, but also is simultaneously also to build to spatial information Mould.Motar transport towards matrix is a kind of space-time analysis method.
The invention proposes be related to a kind of single P300 feature extraction and recognition methods based on matrix Motar transport.
Summary of the invention
(1) technical problem to be solved
Low to solve single P300 extracting method accuracy rate in the prior art, the not high defect of practicability, the present invention provides A kind of P300 feature extracting method based on matrix Motar transport, this method model P300 signal by matrix Motar transport, It using model parameter as feature, promotes the effect of P300 Classification and Identification preferably, stimulates number of repetition situation reducing Under, improve the detection discrimination of P300 signal.
(2) technical solution
The present invention is implemented with the following technical solutions.Overall procedure is as shown in Figure 1:
1, the EEG signal of acquired original is pre-processed, the bandpass filtering of 1-15Hz is carried out first, then according to data Label is segmented；Then, artefact rejecting is carried out to the Epoch after segmentation；Finally, being averaged for the data of preceding 100ms will be stimulated Value is used as baseline, carries out Baseline wander to post-stimulatory data；
2, Conduction choice draws the waveform diagram of goal stimulus Yu non-targeted stimulation by the training data of observation acquisition, selects Most apparent 4 leads of different wave shape are selected as optimal electrode；
3,4 pretreated Epoch of lead of selection are segmented, establish the matrix ash model per small segment data, And the grey actuating quantity of each segmented model is extracted as characteristic parameter, all segmented model parameters connection conduct of each Epoch The feature vector of this Epoch, as shown in Figure 2；
4, all characteristic parameters connection by Epoch obtained by the above method is a feature vector by feature selecting, Then the calculating for carrying out Fisher ratio, is ranked up according to the size of rate value, the corresponding sequence of each feature after being sorted Number.Optimal characteristics space is finally selected according to feature ordering；
5, svm classifier identifies, selects support vector machines as classifier, Selection of kernel function Radial basis kernel function will pass through The feature vector that above-mentioned steps obtain is input to SVM, realizes that the identification to P300 is classified.
(3) beneficial effect
The invention proposes a kind of single P300 feature extracting method based on matrix Motar transport, this method can mention well The characteristic feature amount of P300 is taken out.By combining feature selecting and support vector machines (Support based on Fisher ratio Vector Machine, SVM) classifier to feature vector realize classify, effectively increasing single P300 extract and identification just True rate.It is correct to show that this method all achieves higher identification in the measured data of most of objects through experimental data actual measurement Rate has the advantages that on-line operation speed is with high accuracy fastly, enables the invention to be suitable for practical application.
Detailed description of the invention
Fig. 1 is context of methods general flow chart
Fig. 2 is that Epoch data carry out DTMGM modeling procedure figure
Fig. 3 is the characteristic parameter distribution that data segment carries out after matrix Motar transport under goal stimulus and non-targeted stimulation
P300 recognition correct rate of the Fig. 4 in single-trial extraction and few examination time superposition
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and detailed description.
The present invention provides a kind of P300 feature extracting method based on matrix Motar transport, selects in conjunction with Fisher Ratio Features Method and support vector machines (Support Vector Machine, SVM) are used for the feature extraction and identification of P300 signal, to every It is a that corresponding EEG data section is stimulated to extract model parameter as feature vector, it is accurate that identification is improved in single-trial extraction P300 Rate, and then the recognition efficiency of the character transmission rate and P300 of the BCI system based on P300 in application of detecting a lie can be improved.
Key step of the invention is referring to Fig. 1.Method includes the following steps:
Step S1: to EEG data by pretreatment, including filtering, artefact is rejected and baseline calibration.First to the original of acquisition Beginning data carry out the bandpass filtering of 1-15Hz, and filter selects Butterworth filter；Then it is segmented according to data label, The data for choosing 0.1s 0.7s to after stimulating before stimulating, are segmented into the Epoch of 0.8s；Then, the Epoch after segmentation is carried out pseudo- Mark is rejected, and the data segment by maximum amplitude and minimum amplitude difference greater than 140 μ v is labeled as noise artefact, is rejected, is not involved in Subsequent data processing.Finally, the average value of the data of 100ms carries out base to post-stimulatory data as baseline before stimulating Line correction；
Step S2: Conduction choice, P300 ingredient is obvious in top occipital region, therefore the part lead for choosing this region carries out Analysis, is classified with selecting best lead set to share with subsequent processing.By the training data of observation acquisition, target and non-mesh are drawn Target waveform diagram, and select most apparent 4 leads of different wave shape as optimal electrode.
Step S3: k segmentation is carried out to pretreated Epoch and carries out matrix Motar transport, and extracts each segmented model Grey actuating quantity is as characteristic parameter.Each Epoch all segmented model parameters connection as this Epoch feature to Amount, as shown in Figure 2；
Step S4: feature selecting, will by 2 k obtained by the above method dimension feature vector be spliced into 2k dimensional feature to Then amount carries out the calculating of Fisher ratio, is ranked up according to the size of rate value, each feature is corresponding after being sorted Serial number.Optimal characteristics space is finally selected according to feature ordering；
Step S5: Classification and Identification selects support vector machines as classifier, Selection of kernel function Radial basis kernel function, will be through It crosses the feature vector that above-mentioned steps obtain and is input to SVM, realize that the identification to P300 is classified.
Further, in an embodiment of the present invention (sample rate Fs) step S3 comprising the following specific steps
Step S31: each sample includes the data of 4 lead 0.8s, can be expressed as the matrix form of 4* (Fs*0.8), empty Between on 4 leads are carried out to the matrixing of 2*2, so finally being write as the three dimensional form of 2*2* (Fs*0.8)；
Step S32: time upper Fs*0.8 sample point is segmented, every n point is divided into a segment, can be divided intoSection, every one piece of data matrix form are 2*2*n；
Step S33: DTMGM (1,1+2) modeling is carried out to this 2*2*n data.Detailed process are as follows:
(1) x=(x (1), x (2) ..., x (n)), wherein x (i) indicates the matrix of a 2*2, i ∈ { 1,2 ..., n }, x(0)For the diagonal matrix transform sequence of matrix sequence x, x(1)=MAGOx(0), then:
x(1)=(x(1)(1),x(1)(2),...,x(1)(n)) (1)
Wherein
(2) diagonal transformation matrix sequence ash model (Diagonal Transformation Matrix Sequence is established Grey Model, DTMGM):
x(0)(i)+z(1)(i)=diag (a1,a2) (2)
Wherein, z(1)(i)=0.5x(1)(i)+0.5x(1)(i-1), z(1)(i)∈z(1)=(z(1)(2),...,z(1)(n))。
(3) calculating parameter vector p according to the following formuladm:
pdm=[a, a1,a2]T=(BTB)-1yn (3)
Wherein,
(4) by grey actuating quantity M=[a1,a2] characteristic parameter as this segment data.
Step S34: to k segmentation respectively according to the given Method Modeling of S33 and extracting parameter feature, each Epoch is final K*2 characteristic parameter can be extracted.
In an embodiment of the present invention, sample rate Fs is 200Hz, when n takes 8, k=20.It is extracted after context of methods Characteristic parameter distribution it is as shown in Figure 3.
This method sufficiently applies Motar transport to lack the advantage of data modeling, can will effectively enhance its spy after P300 segmentation modeling Sign extracts model parameter as feature vector, is input to support vector machines after carrying out feature space dimensionality reduction using Fisher ratio Classification and Identification is carried out, discrimination effectively improves.
Feature vector is put into SVM classifier, using radial basis function as kernel function, intersect using leaving-one method and test Card, the average accuracy of all subjects can achieve 91.43%, show the superiority of method given by the present invention.When few After examination time superposition, Classification and Identification rate be can further improve, as shown in Figure 4.

Claims (2)

1. a kind of single P300 detection method based on matrix Motar transport, which is characterized in that the described method comprises the following steps:
Step S1: to EEG data by pretreatment, including filtering, segmentation, artefact is rejected and baseline calibration；
Step S2: by the training data of observation acquisition, target and non-targeted waveform diagram are drawn, and select different wave shape most Apparent 4 leads are as optimal electrode；
Step S3: k segmentation is carried out to pretreated each sample, matrix Motar transport then is carried out to each segmentation, and extract The grey actuating quantity that the dimension of each segmented model is 2 is as characteristic parameter, each available 2*k characteristic parameter of sample；
Step S4: the calculating of Fisher ratio will be carried out by the feature vector of 2*k obtained by the above method dimension, according to rate value Size be ranked up, the corresponding serial number of each feature after being sorted finally selects optimal characteristics space according to feature ordering；
Step S5: Classification and Identification selects support vector machines as classifier, and Selection of kernel function Radial basis kernel function will pass through upper The feature vector for stating the optimal characteristics space that step obtains is input to SVM, realizes that the identification to single P300 is classified.
2. based on the single P300 detection method described in claim 1 based on matrix Motar transport, which is characterized in that the step S3 the following steps are included:
Step S31: each sample includes the data of 4 lead 0.8s, can be expressed as the matrix form of 4* (Fs*0.8), and Fs is to adopt 4 leads are spatially carried out the matrixing of 2*2, so finally being write as the three dimensional form of 2*2* (Fs*0.8) by sample frequency；
Step S32: time upper Fs*0.8 sample point is segmented, every n point is divided into a segment, can be divided intoSection, every one piece of data matrix form are 2*2*n；
Step S33: DTMGM (1,1+2) modeling is carried out to this 2*2*n data, and extracts model ash actuating quantity parameter；
Step S34: to k segmentation respectively according to the given Method Modeling of S33 and extracting parameter feature, each sample may finally Extract 2*k characteristic parameter.
CN201610045616.XA 2016-01-25 2016-01-25 A kind of single P300 detection method based on matrix Motar transport Active CN105718953B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610045616.XA CN105718953B (en) 2016-01-25 2016-01-25 A kind of single P300 detection method based on matrix Motar transport

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610045616.XA CN105718953B (en) 2016-01-25 2016-01-25 A kind of single P300 detection method based on matrix Motar transport

Publications (2)

Publication Number Publication Date
CN105718953A CN105718953A (en) 2016-06-29
CN105718953B true CN105718953B (en) 2019-07-05

Family

ID=56154004

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610045616.XA Active CN105718953B (en) 2016-01-25 2016-01-25 A kind of single P300 detection method based on matrix Motar transport

Country Status (1)

Country Link
CN (1) CN105718953B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106713787A (en) * 2016-11-02 2017-05-24 天津大学 Evaluation method for watching comfort level caused by rolling subtitles of different speed based on EEG
CN107180227B (en) * 2017-05-02 2021-02-19 西北工业大学 Cloud prediction and forecast method based on DTMMG + gray model
CN110192875B (en) * 2019-05-24 2020-10-16 中南民族大学 Lie detection method based on multi-lead electroencephalogram signal form factor
TWI734222B (en) * 2019-10-17 2021-07-21 孫光天 A system that combines brain waves and artificial intelligence to diagnose dyslexia

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102521206A (en) * 2011-12-16 2012-06-27 天津大学 Lead optimization method for SVM-RFE (support vector machine-recursive feature elimination) based on ensemble learning thought
CN103150023A (en) * 2013-04-01 2013-06-12 北京理工大学 System and method for cursor control based on brain-computer interface
US8699767B1 (en) * 2006-10-06 2014-04-15 Hrl Laboratories, Llc System for optimal rapid serial visual presentation (RSVP) from user-specific neural brain signals
CN104899573A (en) * 2015-06-18 2015-09-09 福州大学 P300 feature extraction method based on wavelet transformation and Fisher criterion
CN105068644A (en) * 2015-07-24 2015-11-18 山东大学 Method for detecting P300 electroencephalogram based on convolutional neural network

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2009064474A1 (en) * 2007-11-13 2009-05-22 Wavesynch Technologies, Inc. A method of determining whether a test subject is a specific individual

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8699767B1 (en) * 2006-10-06 2014-04-15 Hrl Laboratories, Llc System for optimal rapid serial visual presentation (RSVP) from user-specific neural brain signals
CN102521206A (en) * 2011-12-16 2012-06-27 天津大学 Lead optimization method for SVM-RFE (support vector machine-recursive feature elimination) based on ensemble learning thought
CN103150023A (en) * 2013-04-01 2013-06-12 北京理工大学 System and method for cursor control based on brain-computer interface
CN104899573A (en) * 2015-06-18 2015-09-09 福州大学 P300 feature extraction method based on wavelet transformation and Fisher criterion
CN105068644A (en) * 2015-07-24 2015-11-18 山东大学 Method for detecting P300 electroencephalogram based on convolutional neural network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Single Channel Single Trial P300 Detection Using Extreme Learning Machine Compared with BPNN and SVM;Songyun Xie等;《2014 International Joint Conference on Neural Network》;20140711;第544-549页
单导联脑电信号P300分类研究;黄璐等;《仪器仪表学报》;20140430;第35卷(第4期);第814-819页

Also Published As

Publication number Publication date
CN105718953A (en) 2016-06-29

Similar Documents

Publication Publication Date Title
CN110443798B (en) Autism detection method, device and system based on magnetic resonance image
CN105718953B (en) A kind of single P300 detection method based on matrix Motar transport
CN111329474B (en) Electroencephalogram identity recognition method and system based on deep learning and information updating method
Bhattacharjee et al. Robust technique for the detection of acute lymphoblastic leukemia
CN108959895B (en) Electroencephalogram EEG (electroencephalogram) identity recognition method based on convolutional neural network
CN103714281A (en) Identity recognition method based on electrocardiosignals
CN108256579A (en) A kind of multi-modal sense of national identity quantization measuring method based on priori
CN109344816A (en) A method of based on brain electricity real-time detection face action
EP3449409B1 (en) Biometric method and device for identifying a person through an electrocardiogram (ecg) waveform
CN106874722A (en) A kind of personal identification method and its device based on electrocardiosignal
CN110037693A (en) A kind of mood classification method based on facial expression and EEG
CN107480716A (en) A kind of combination EOG and video pan signal recognition method and system
CN107480635B (en) Glance signal identification method and system based on bimodal classification model fusion
CN111616702A (en) Lie detection analysis system based on cognitive load enhancement
CN104793743B (en) A kind of virtual social system and its control method
Sun et al. A practical cross-domain ECG biometric identification method
CN108338777A (en) A kind of pulse signal determination method and device
Wei et al. Multi-feature fusion for ultrasound breast image classification of benign and malignant
CN116098637A (en) Brain function evaluation device based on ICA (independent component analysis) optimization correction brain electric micro-state
Huang et al. Automatic artifact removal in EEG using independent component analysis and one-class classification strategy
CN105787459A (en) ERP signal classification method based on optimal score sparse determination
Heriawati et al. Blood cells classification for identification of acute lymphoblastic leukemia on microscopic images using image processing
Wang et al. Study on correlation between subjective and objective metrics for multimodal retinal image registration
CN113057654B (en) Memory load detection and extraction system and method based on frequency coupling neural network model
CN113907770A (en) Method and system for detecting and identifying spike-slow complex wave based on feature fusion

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant