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 PDFInfo
- 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
- modeling
- detection method
- 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
- 239000011159 matrix material Substances 0.000 title claims abstract description 29
- 238000001514 detection method Methods 0.000 title claims abstract description 15
- 238000000034 method Methods 0.000 claims abstract description 28
- 239000013598 vector Substances 0.000 claims abstract description 18
- 238000012706 support-vector machine Methods 0.000 claims abstract description 13
- 239000000284 extract Substances 0.000 claims abstract description 4
- 230000011218 segmentation Effects 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims abstract description 4
- 230000009471 action Effects 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 2
- 230000001131 transforming effect Effects 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims 1
- 230000000638 stimulation Effects 0.000 abstract description 12
- 238000000605 extraction Methods 0.000 abstract description 11
- 230000001149 cognitive effect Effects 0.000 abstract description 3
- 238000012360 testing method Methods 0.000 abstract description 2
- 244000287680 Garcinia dulcis Species 0.000 abstract 1
- 238000010586 diagram Methods 0.000 abstract 1
- 238000012545 processing Methods 0.000 description 5
- 230000008030 elimination Effects 0.000 description 3
- 238000003379 elimination reaction Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- CPLXHLVBOLITMK-UHFFFAOYSA-N Magnesium oxide Chemical compound [Mg]=O CPLXHLVBOLITMK-UHFFFAOYSA-N 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000763 evoking effect Effects 0.000 description 1
- 238000012880 independent component analysis Methods 0.000 description 1
- 230000001936 parietal effect Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 210000004761 scalp Anatomy 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 230000003238 somatosensory effect Effects 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
Technical Field
The invention belongs to the field of cognitive neuroscience, relates to a method for extracting, identifying and detecting characteristics of event-related potential P300, and particularly relates to a single P300 detection method based on matrix gray modeling in a gray theory.
Background
P300 (also known as P3b) is an endogenous component that reflects the process of advanced cognitive processing, and is the largest late positive wave recorded on the scalp with a latency of about 300ms when subject to "target stimuli" and evoked by visual, auditory and somatosensory stimuli.
The P300 potential is widely used in Brain-Computer Interface (BCI) and psychological lie detection studies. The traditional P300 identification method adopts a time domain coherent averaging method, and background random noise can be effectively reduced through superposition averaging. This method can identify P300 well, but has significant drawbacks: the tested object is easy to feel tired after being averaged for many times, the difference between ERP of different test times is ignored, most importantly, the requirement of real-time performance cannot be met, and the method is not beneficial to the practical application of P300.
Researchers have proposed various methods to implement the feature extraction of the minority/single P300, such as a recognition algorithm based on wavelet transform, a recognition algorithm based on independent component analysis, a recognition algorithm based on principal component analysis, and the like. However, these methods have certain limitations, and the detection accuracy of a single extraction of P300 is not very high.
Grey System Theory was first introduced internationally in 1982 by the university of Dunconlon, scholarly in China. The method is particularly suitable for processing atypical regular signals (such as non-stationary, non-Gaussian distribution and non-white noise), and compared with other methods for processing data according to statistical rules and prior rules, the gray method has the obvious advantages of small required data volume, no need of prior knowledge and the like. The gray modeling facing the matrix is a further extension of the traditional gray modeling, and extends the point sequence modeling to the surface, namely the matrix sequence, which not only contains the time information, but also models the space information. Matrix-oriented gray modeling is a spatio-temporal analysis method.
The invention provides a single P300 feature extraction and identification method based on matrix gray modeling.
Disclosure of Invention
(I) technical problem to be solved
In order to overcome the defects of low accuracy and low practicability of a single P300 extraction method in the prior art, the invention provides a P300 feature extraction method based on matrix gray modeling, which models a P300 signal through matrix gray modeling and takes model parameters as features, so that the P300 classification and identification effects are better improved, and the detection and identification rate of the P300 signal is improved under the condition of reducing stimulation repetition times.
(II) technical scheme
The invention is realized by adopting the following technical scheme. The overall flow is shown in figure 1:
1. preprocessing an EEG signal which is originally collected, firstly carrying out 1-15Hz band-pass filtering, and then segmenting according to a data label; then, artifact elimination is carried out on the segmented Epoch; finally, taking the average value of the data 100ms before stimulation as a baseline, and performing baseline correction on the data after stimulation;
2. lead selection, which is to draw a oscillogram of target stimulation and non-target stimulation by observing the collected training data and select 4 leads with the most obvious waveform difference as the optimal electrodes;
3. segmenting the selected 4 lead preprocessed epochs, establishing a matrix gray model of each small segment of data, extracting the gray action quantity of each segmented model as a characteristic parameter, and connecting all segmented model parameters of each Epoch as the characteristic vector of the Epoch, as shown in fig. 2;
4. and (3) feature selection, namely connecting all feature parameters of the Epoch obtained by the method into a feature vector, then calculating the Fisher ratio, and sequencing according to the ratio value to obtain the sequence number corresponding to each sequenced feature. Finally, selecting an optimal feature space according to the feature sorting;
5. and (3) SVM classification and identification, namely selecting a support vector machine as a classifier, selecting a radial basis kernel function by the kernel function, and inputting the feature vector obtained by the steps into the SVM to realize the identification and classification of the P300.
(III) advantageous effects
The invention provides a single P300 feature extraction method based on matrix gray modeling, which can well extract the characteristic feature quantity of P300. Feature vectors are classified by combining feature selection based on a Fisher ratio and a Support Vector Machine (SVM) classifier, and the accuracy of single P300 extraction and identification is effectively improved. The actual measurement of experimental data shows that the method obtains higher identification accuracy on the actual measurement data of most objects, has the advantages of high online operation speed and high precision, and can be suitable for practical application.
Drawings
FIG. 1 is a general flow chart of the process of the present invention
FIG. 2 is a flowchart of DTMGM modeling with Epoch data
FIG. 3 is a characteristic parameter distribution of data segments under target stimulation and non-target stimulation after matrix gray modeling
FIG. 4P 300 recognition accuracy at single extraction and low trial overlap
Detailed description of the preferred embodiments
The invention is further described with reference to the following figures and detailed description.
The invention provides a P300 feature extraction method based on matrix gray modeling, which is used for feature extraction and identification of P300 signals by combining a Fisher ratio feature selection method and a Support Vector Machine (SVM), and extracting model parameters from an EEG data segment corresponding to each stimulus to be used as feature vectors, so that the identification accuracy is improved when P300 is extracted once, and further, the character transmission rate of a BCI system based on P300 and the identification efficiency of P300 in lie detection application can be improved.
The main steps of the invention are shown in figure 1. The method comprises the following steps:
step S1: the EEG data is pre-processed, including filtering, artifact removal and baseline calibration. Firstly, performing 1-15Hz band-pass filtering on acquired original data, and selecting a Butterworth filter as a filter; then, segmenting according to the data labels, selecting data from 0.1s before stimulation to 0.7s after stimulation, and segmenting into epochs of 0.8 s; and then, carrying out artifact elimination on the segmented Epoch, marking the data segment with the difference between the maximum amplitude and the minimum amplitude larger than 140 μ v as a noise artifact, and carrying out elimination without participating in subsequent data processing. Finally, taking the average value of the data 100ms before stimulation as a baseline, and performing baseline correction on the data after stimulation;
step S2: lead selection, the P300 component, is more pronounced in the parietal occipital region, so that a portion of the leads in this region are selected for analysis to select the best lead combination for subsequent processing classification. By observing the acquired training data, target and non-target oscillograms are plotted, and the 4 leads with the most distinct waveform differences are selected as the optimal electrodes.
Step S3: and performing matrix gray modeling on the preprocessed Epoch by k segmentation, and extracting the gray action quantity of each segmented model as a characteristic parameter. All the segment model parameters of each Epoch are concatenated as the feature vector of this one Epoch, as shown in FIG. 2;
step S4: and selecting the features, namely splicing the 2 k-dimensional feature vectors obtained by the method into 2 k-dimensional feature vectors, then calculating the Fisher ratio, and sequencing according to the ratio value to obtain the sequence number corresponding to each sequenced feature. Finally, selecting an optimal feature space according to the feature sorting;
step S5: and (4) classifying and identifying, namely selecting a support vector machine as a classifier, selecting a radial basis kernel function by the kernel function, and inputting the feature vector obtained by the steps into the SVM to realize the identification and classification of the P300.
Further, in an embodiment of the present invention (with a sampling rate of Fs), step S3 includes the following specific steps:
step S31: each sample comprises 4 leads of 0.8s data, which can be represented in a 4 x (Fs x 0.8) matrix form, spatially transforming the 4 leads by a 2 x 2 matrix, so that they are finally written in a three-dimensional form of 2 x 2 (Fs x 0.8);
step S32: 0.8 sample points Fs in time are segmented, and each n points are divided into small segments which can be divided intoSegments, each segment in a data matrix of 2 x n;
step S33: DTMGM (1, 1+2) modeling was performed on the 2 x n data. The specific process is as follows:
(1) x (1), x (2),., x (n)), where x (i) represents a 2 x 2 matrix, i ∈ {1, 2., n }, x £ and x ·, where x · represents a number of x (1)(0)A diagonal matrix transformation sequence being a matrix sequence x, x(1)=MAGOx(0)And then:
x(1)=(x(1)(1),x(1)(2),...,x(1)(n)) (1)
wherein
(2) Establishing a Diagonal Transformation Matrix sequence gray Model (DTMMG):
x(0)(i)+z(1)(i)=diag(a1,a2) (2)
wherein z is(1)(i)=0.5x(1)(i)+0.5x(1)(i-1),z(1)(i)∈z(1)=(z(1)(2),...,z(1)(n))。
(3) Calculating a parameter vector p according todm:
pdm=[a,a1,a2]T=(BTB)-1yn (3)
Wherein,
(4) the acting amount of ash is equal to [ a ]1,a2]As a characteristic parameter of this piece of data.
Step S34: and modeling and extracting parameter characteristics for the k segments according to the method given by S33, wherein k x 2 characteristic parameters can be extracted finally for each Epoch.
In one embodiment of the present invention, the sampling rate Fs is 200Hz, and when n is 8, k is 20. The distribution of the characteristic parameters extracted by the method is shown in figure 3.
The method fully applies the advantages of few data modeling of gray modeling, can effectively enhance the characteristics of P300 after segmented modeling, extracts model parameters as characteristic vectors, performs characteristic space dimensionality reduction by adopting a Fisher ratio, inputs the characteristic vectors into a support vector machine for classification and identification, and effectively improves the identification rate.
The feature vectors are put into an SVM classifier, the radial basis function is taken as a kernel function, cross validation is carried out by adopting a leave-one-out method, the average accuracy of all tested results can reach 91.43%, and the superiority of the method provided by the invention is shown. When the overlapping is performed with few trials, the classification recognition rate can be further improved, as shown in fig. 4.
Claims (2)
1. A single P300 detection method based on matrix gray modeling is characterized by comprising the following steps:
step S1: preprocessing EEG data including filtering, segmentation, artifact removal and baseline calibration;
step S2: drawing a target and non-target oscillogram by observing the collected training data, and selecting 4 leads with the most obvious waveform difference as optimal electrodes;
step S3: performing k segmentation on each preprocessed sample, performing matrix gray modeling on each segment, extracting a gray action quantity with the dimension of 2 of each segmented model as a characteristic parameter, and obtaining 2 x k characteristic parameters from each sample;
step S4: performing Fisher ratio calculation on the feature vectors of 2 x k dimensions obtained by the method, sorting according to the size of ratio values to obtain a serial number corresponding to each sorted feature, and finally selecting an optimal feature space according to feature sorting;
step S5: and (4) performing classification and identification, namely selecting a support vector machine as a classifier, selecting a radial basis kernel function by the kernel function, inputting the feature vector of the optimal feature space obtained through the steps into the SVM, and realizing the identification and classification of the single P300.
2. The single P300 detection method based on matrix gray modeling according to claim 1, wherein the step S3 comprises the following steps:
step S31: each sample comprises 4 leads of 0.8s data, which can be represented in the form of a 4 x (Fs x 0.8) matrix, Fs being the sampling frequency, spatially transforming the 4 leads by a 2 x 2 matrix, so finally written in the form of 2 x (Fs x 0.8) in three dimensions;
step S32: 0.8 sample points Fs in time are segmented, and each n points are divided into small segments which can be divided intoSegments, each segment in a data matrix of 2 x n;
step S33: carrying out DTMGM (1, 1+2) modeling on the 2 x n data, and extracting a model ash action quantity parameter;
step S34: and modeling and extracting parameter characteristics for the k segments according to the method given by S33, wherein each sample can finally extract 2 x k characteristic parameters.
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)
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)
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)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7574254B2 (en) * | 2007-11-13 | 2009-08-11 | Wavesynch Technologies, Inc. | Method for monitoring attentiveness and productivity in a subject |
-
2016
- 2016-01-25 CN CN201610045616.XA patent/CN105718953B/en active Active
Patent Citations (5)
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)
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 |
---|---|---|
CN107157477B (en) | Electroencephalogram signal feature recognition system and method | |
CN103714281B (en) | A kind of personal identification method based on electrocardiosignal | |
CN105718953B (en) | A kind of single P300 detection method based on matrix Motar transport | |
CN110811609B (en) | Epileptic spike intelligent detection device based on self-adaptive template matching and machine learning algorithm fusion | |
CN110991406B (en) | RSVP electroencephalogram characteristic-based small target detection method and system | |
CN108960182B (en) | P300 event related potential classification identification method based on deep learning | |
CN101794372B (en) | Method for representing and recognizing gait characteristics based on frequency domain analysis | |
CN110367980A (en) | EEG signals Emotion identification method based on polynary empirical mode decomposition | |
CN111967363B (en) | Emotion prediction method based on micro-expression recognition and eye movement tracking | |
CN104794440B (en) | A kind of false fingerprint detection method based on the multiple dimensioned LBP of more piecemeals | |
EP2304627A1 (en) | A method and system for classifying brain signals in a bci | |
Naik et al. | Multi run ICA and surface EMG based signal processing system for recognising hand gestures | |
CN105678341B (en) | A kind of woollen and cashmere recognizer based on Gabor wavelet analysis | |
CN108256579A (en) | A kind of multi-modal sense of national identity quantization measuring method based on priori | |
CN113486752B (en) | Emotion recognition method and system based on electrocardiosignal | |
Wen et al. | Virus image classification using multi-scale completed local binary pattern features extracted from filtered images by multi-scale principal component analysis | |
CN112674782B (en) | Device and method for detecting epileptic-like electrical activity of epileptic during inter-seizure period | |
CN107480716A (en) | Method and system for identifying saccade signal by combining EOG and video | |
CN111460953A (en) | Electrocardiosignal classification method based on self-adaptive learning of countermeasure domain | |
CN109009098B (en) | Electroencephalogram signal feature identification method under motor imagery state | |
CN107480635B (en) | Glance signal identification method and system based on bimodal classification model fusion | |
CN107506718A (en) | Facial expression recognizing method based on MB 2DPCA features | |
CN108920699B (en) | Target identification feedback system and method based on N2pc | |
CN107122710B (en) | Finger vein feature extraction method based on scattering convolution network | |
Eltayef et al. | Detection of pigment networks in dermoscopy images |
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 |