CN103345640B - A kind of towards the sorting technique persistently imagining EEG signals - Google Patents
A kind of towards the sorting technique persistently imagining EEG signals Download PDFInfo
- Publication number
- CN103345640B CN103345640B CN201310273395.8A CN201310273395A CN103345640B CN 103345640 B CN103345640 B CN 103345640B CN 201310273395 A CN201310273395 A CN 201310273395A CN 103345640 B CN103345640 B CN 103345640B
- Authority
- CN
- China
- Prior art keywords
- sample
- data set
- transfer point
- distance
- vector
- 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
- 238000000034 method Methods 0.000 title claims description 37
- 238000012546 transfer Methods 0.000 claims abstract description 42
- 238000012360 testing method Methods 0.000 claims abstract description 27
- 238000000746 purification Methods 0.000 claims abstract description 12
- 239000013598 vector Substances 0.000 claims description 26
- 238000012549 training Methods 0.000 claims description 21
- 238000001514 detection method Methods 0.000 claims description 10
- 238000012706 support-vector machine Methods 0.000 claims description 9
- 238000005315 distribution function Methods 0.000 claims description 4
- 238000007619 statistical method Methods 0.000 claims description 3
- 238000000926 separation method Methods 0.000 abstract description 5
- 230000002045 lasting effect Effects 0.000 abstract description 2
- 238000003909 pattern recognition Methods 0.000 abstract description 2
- 238000011160 research Methods 0.000 description 8
- 210000004556 brain Anatomy 0.000 description 7
- 238000002474 experimental method Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 4
- 101150100075 PSD1 gene Proteins 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000003340 mental effect Effects 0.000 description 2
- 101150035846 psd2 gene Proteins 0.000 description 2
- 208000014644 Brain disease Diseases 0.000 description 1
- 241001269238 Data Species 0.000 description 1
- 101000755816 Homo sapiens Inactive rhomboid protein 1 Proteins 0.000 description 1
- 102100022420 Inactive rhomboid protein 1 Human genes 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000013398 bayesian method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 210000003141 lower extremity Anatomy 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000002560 therapeutic procedure Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to area of pattern recognition, disclose a kind of towards the classification policy persistently imagining EEG signals.First, it is assumed that there is transfer point between different imagination tasks, the sample separation that at transfer point, Euclidean distance between sample is greater than between transfer point from, go out transfer point by setting sample separation from threshold test；Secondly, think when persistently imagining same task, can lax because of attention, the factor such as tired causes signal by sound pollution, this strategy adds Sample purification thought, by setting sample separation from scope, from all samples corresponding with task, filter out part sample, and return the classification as all samples of this task of the most of classifications in this part sample.The present invention, by taking into full account the contact between adjacent sample, improves the discrimination of overall sample, is especially suitable for lasting EEG signals is carried out off-line analysis.
Description
Technical field
The invention belongs to area of pattern recognition, relate to a kind of towards the sorting technique persistently imagining EEG signals.
Background technology
The research of brain-computer interface (Brain Computer Interface, BCI) start from brain electricity development, in recent decades, along with
The development of the technical research such as signal processing and machine learning, BCI research is increasingly becoming focus.Mental imagery EEG signals is BCI
A kind of research that field is common, by collection analysis people when imagining to move or carry out certain thinking activities in certain position of health
EEG signals, thus identify the state of people's brain, and then control external device.BCI technology is not only disease of brain patient and provides
A kind of new diagnostic mode, it is often more important that realize the new way of a kind of people and extraneous communication.
In BCI research, it is identified being an important ring in research process to the EEG signals of reaction brain difference thinking mistake area
Joint.The most many scholars conduct in-depth research in terms of Classification and Identification, and achieve certain achievement.Conventional classification
Method has support vector machine, Bayesian Method, nearest neighbor classification and artificial neural network method etc..Li Lijun et al. uses linear point
Class device, support vector machine realize the classification of test set Mental imagery eeg data, and achieve preferable classifying quality.LiuHui
Et al. utilize support vector machine classifier that event-related potential N400 signal is classified, also achieve preferable classifying quality.
The sorter model output that application training data set trains is predicting the outcome of single test sample, for adjacent sample
Between there is the data set of certain contact for, this conventional classification method have ignored the contact between sample, causes classification results
Only rely upon the prediction of grader, and higher discrimination cannot be obtained.
Summary of the invention
It is an object of the invention to overcome the deficiency of above-mentioned conventional classification method, propose a kind of based on transfer point detection and Sample purification
The sorting technique of thought.The method has taken into full account persistently imagines the contact between EEG signals sample, thus achieves higher
Classification accuracy rate.
The described feature persistently imagining EEG signals is that experimenter persistently imagines multiple imagination task, is continuous print, then between task
Can collect continuous print EEG signals in a period of time sequence, collecting flowchart schematic diagram is as shown in Figure 1.Present invention assumes that
Brain electricity sample in certain imagination task meeting certain time, i.e. certain time sequence belongs to same category, different imagination tasks
Between there is transfer point, the sample separation that at transfer point, Euclidean distance between sample is greater than between transfer point from, by setting sample
This spacing threshold value can detect transfer point；Simultaneously, it is believed that when persistently imagining same task, can because of attention lax, tired
Cause signal by sound pollution etc. factor.The present invention adds Sample purification thought, by setting sample separation from scope, from appoint
All samples that business is corresponding filter out part sample, and returns the most of classifications in this part sample as all samples of this task
This classification.The present invention, by taking into full account the contact between adjacent sample, improves the discrimination of overall sample, is especially suitable for
Lasting EEG signals is carried out off-line analysis.
The method of the invention comprises the following steps:
Step one, carries out transfer point detection.
Transfer point detection is mainly used in determining the distance threshold U at transfer point between sample.Assume to there is sample class label known
Training dataset X and unknown test data set Y of sample class label, in data set, sample is belonging respectively to multiple different class
Not, the method for computed range threshold value U is as follows:
(1) calculate in training dataset X the Euclidean distance between adjacent sample two-by-two, and save as vector Dist.
(2) according to the sample class label of training dataset, calculated for step (1) vector Dist is divided into two parts: if
Two adjacent samples belong to same category, and its distance is included into vector XnotransIn, it is designated as sample distance vector in class；If two is adjacent
Sample belongs to different classifications, and its distance is included into vector XtransIn, it is designated as sample distance vector between class.
(3) X is calculatednotransIn the maximum of all distances, be designated as U1。
(4) X is calculatedtransIn the minima of all distances, be designated as U2。
(5) according to step (3) and the result of step (4), by U1And U2In less that as threshold value U, i.e. U=min
(U1,U2).
Step 2, carries out Sample purification.
Sample purification is mainly used in the distance range [a, b] determining between similar sample, and a, b are respectively minima and the maximum of distance
Value.Concrete grammar is as follows:
(1) training dataset X splitting into two parts, a part of X_train is used as grader and learns, and a part of X_test uses
Predict.
(2) use data set Training Support Vector Machines grader, obtain sorter model, then X_test data set is carried out pre-
Survey, obtain the prediction label of X_test data set.
(3) true tag of contrast X_test data set and prediction label, calculate all devices that are classified in data set and just predicting continuously
The true Euclidean distance between same category of adjacent sample, and it is designated as vector D.
(4) the distance vector D obtaining step (3) carries out statistical analysis, determines its distribution function, generally, is somebody's turn to do
Vector Normal Distribution or approximate normal distribution.
(5) distribution function of the vectorial D determined according to step (4), estimate its 95% confidence level hourly value estimate put
Letter is interval as the distance range [a, b] between similar sample.
Step 3, carries out Classification and Identification.
Classification and Identification is on the basis of trying to achieve threshold value U and distance range [a, b], enters test data set Y of unknown sample classification
Row prediction, concrete grammar is as follows:
(1) use training dataset X Training Support Vector Machines grader, test data set Y is predicted, is tested
The sample predictions label vector predict of data set.
(2) the Euclidean distance d between adjacent sample in test data set is calculatedi, i=1,2 ..., N-1, N are in test data set
Number of samples.
(3) if diMore than or equal to U, illustrate to exist between i-th sample and i+1 sample classification transfer point, forward step to
(5).
(4) if diLess than U, illustrate that i-th sample and i+1 sample belong to same category, then judge diWhether in scope
In [a, b], if diIn the range of [a, b], illustrate that the two sample is likely to be classified the sample that device is correctly predicted, then by this two
The prediction label of individual sample adds in the judgement set of an entitled judgeset；If diNot in the range of [a, b], return step
(2).
(5) when a transfer point being detected, calculate judgeset and judge the quantity of each class label in set, return quantity
Most labels, as the class label of all samples between this transfer point and a upper transfer point, then empties judgeset and sentences
Disconnected set, forwards the step (3) detection to next transfer point to.
The invention has the beneficial effects as follows: traditional sorting technique is to be identified single sample, improve the approach of classification accuracy rate
Can only be by promoting the recognition result of single sample, being then lifted out space can be limited by noise.The present invention propose method with
Local replaces entirety, mends inferior position with advantage, effectively overcomes contaminated EEG signals sample and can be substantially reduced overall signal
The shortcoming of discrimination.
Accompanying drawing explanation
Fig. 1 is persistently to imagine eeg signal acquisition schematic flow sheet；
Fig. 2 is the main flow chart of sorting technique involved in the present invention；
Fig. 3 is transfer point detection method FB(flow block) involved in the present invention；
Fig. 4 is Sample purification method flow block diagram involved in the present invention；
Fig. 5 is specimen discerning method flow block diagram involved in the present invention.
Detailed description of the invention
With embodiment, the present invention is described in further detail below in conjunction with the accompanying drawings.
The method of the invention being applied on BCI2005 contest standard data set Data Set V, data pick up from 3 Healthy subjects
(Essential Terms in psychology, the most tested people).During experiment, tested being sitting on a chair, arm is placed on lower limb with loosening.
Experiment has 3 tasks: imagination left hand motion, the imagination right hand moves, the word that the imagination starts with identical random letters.Whole experiment
Process includes the experiment of 4 groups of duplicons, each tested all collects 4 groups of eeg datas.Each tested 4 son experiment data be all
Gather on the same day, every individual sub-Therapy lasted 4 minutes, be spaced 5-10 minute between sub-experiment.First, tested execution one is given appoints
Business, the mission duration is 15 seconds, the most tested perform next main examination (director of experiment) incessantly and is randomly assigned
Task.In experimentation, tested permission is had a rest, to ensure tested carrying out wherein any one task incessantly.Right
The process that initial data carries out feature extraction is as follows: first by surface laplacian filter operator, initial data is carried out space filter
Ripple；The most every 62.5 milliseconds (16 times the most per second) calculate the power spectral density in a 8-30Hz, and frequency resolution is 2Hz, brain
The electrode in 8 central temporo districts on electrical signal collection device is C3, CZ, C4, CP1, CP2, P3, PZ and P4.Finally
To EEG signals, every section of EEG signals sample is 96 dimensional vectors (8 electrode × 12 frequency contents).
Data set comprises three tested data, and each tested all collection has 4 part data sets, be psd1 respectively, psd2, psd3
And psd4, wherein first three data integrates as training dataset, and last data set is test data set.Experimental duties comprise three kinds
Phenomenon task, so data set is three class classification problems, classification is respectively 2,3 and 7.
Method flow diagram of the present invention is as in figure 2 it is shown, specifically include following steps:
Step one, carries out transfer point detection, and its flow chart is as it is shown on figure 3, detailed process is as follows:
(1) calculate training data respectively and concentrate the Euclidean distance between three data set adjacent samples, obtain three distance vectors
Dist1,Dist2,Dist3；
(2) according to the class label that data set is corresponding, respectively three distance vectors being split as two parts, Part I is similar
The very Euclidean distance between this, Part II is the Euclidean distance at transfer point between sample, merges the first of three data sets
Part obtains Xnotrans, the Part II merging three data sets obtains Xtran；
(3) X is calculatednotransIn the maximum of all distances, be designated as U1；
(4) X is calculatedtranIn the minima of all distances, be designated as U2；
(5) according to step (3) and the result of step (4), by U1And U2In less that as threshold value U, i.e. U=min
(U1,U2)；
The transfer point threshold estimation result of three tested data sets is as shown in table 1.
1 three, table tested test data set transfer point testing result
| Tested 1 | Tested 2 | Tested 3 | |
| U1 | 0.1630 | 0.1599 | 0.1082 |
| U2 | 0.0720 | 0.0876 | 0.0858 |
| U | 0.0720 | 0.0876 | 0.0858 |
As shown in Table 1, sample distance set X at the final threshold value of three tested data sets and transfer pointtransMinima phase
Deng.This illustrates that the Euclidean distance of same category of two adjacent samples is likely to more than the distance of sample at transfer point, then, root
The threshold value selected according to this mode may detect the transfer point of vacation, causes same category data to be separated by false transfer point.
Sample purification thought is that the sample between the transfer point detected is purified extraction, same category data is separated and simply increases
Add the process of a Sample purification, so the Research on threshold selection simple possible proposed, it is possible to effectively detect between sample
The transfer point existed.
Step 2, carries out Sample purification, and its flow chart as shown in Figure 4, specifically comprises the following steps that
(1) two data set psd that training data is concentrated are usedi,psdj, i, j=1,2,3, i ≠ j, Training Support Vector Machines grader,
Obtain sorter model, then to another data set psdk, k=1,2,3, k ≠ i, j, it is predicted, obtains the pre-of this data set
Survey tag set pk。
(2) by the prediction tag set p of this data setkWith true class label set TkContrast, if prediction tag set
pkIn i-th (i=1,2 ..., N-1) classification of individual sample is identical with the classification of i+1 sample adjacent thereto, wherein N is data set psdk
The number of sample, and i-th sample and i+1 sample standard deviation are predicted correctly, then calculate the Euclid between the two sample
Distance dist (i).
(3) compare the size of dist (i) and threshold value U, if dist (i) is less than U, then dist (i) added in distance set DistSet,
Set DistSet is i.e. the Euclidean distance collection being classified between the adjacent sample that device prediction is correct in the sample not comprising transfer point
Close.
(4) the distance vector DistSet obtaining step (3) carries out statistical analysis, and this vector Normal Distribution or approximation are just
State is distributed.
(5) according to estimate three tested just dividing distance set confidence level be 95% under normal distribution time distance Estimation of Mean confidence
Interval, and as sample distance condition scope [a, b].
Result is as shown in table 2.
2 three, table tested sample distance estimations scope
| Tested | Confidence interval |
| Tested 1 | [0.0389,0.0394] |
| Tested 2 | [0.0340,0.0344] |
| Tested 3 | [0.0334,0.0337] |
Step 3, carries out Classification and Identification, and its flow chart is as it is shown in figure 5, specifically comprise the following steps that
(1) merging three data set psd1, psd2, psd3 are as training dataset, and Training Support Vector Machines grader, to survey
Examination data set psd4 is predicted, and obtains the sample predictions label vector predict of test data set.
(2) the Euclidean distance d between adjacent sample in test data set is calculatedi, i=1,2 ..., N-1, N are in test data set
Number of samples；
(3) d is judgediWhether more than or equal to U, if being more than or equal to, illustrate to there is class between i-th sample and i+1 sample
Other transfer point, forwards step (5) to.
(4) if diLess than U, illustrate that i-th sample and i+1 sample belong to same category, then judge diWhether in scope
In [a, b].If diIn the scope [a, b], illustrate that the two sample is likely to be classified the sample that device is correctly predicted, then by this two
The prediction label of individual sample adds in an entitled judgeset set；If diNot in scope [a, b], return step (2).
(5) when a transfer point being detected, calculate the quantity of each class label in judgeset set, return quantity most
Label as the class label of all samples between this transfer point and a upper transfer point, then empty judgeset set, turn
To the step 3 detection to next transfer point.
Threshold value obtained above and distance condition scope being all applied in the identification of final sample data set to be predicted, statistics is final
The accuracy of classification.Table 3 gives employing the method for the invention and the accuracy of control methods.Control methods is BCI contest
One one proposed sentences method for distinguishing based on linear discriminant analysis and characteristic statistics.
The accuracy that table 3 is finally classified
| Tested 1 | Tested 2 | Tested 3 | Averagely | |
| Control methods | 79.60% | 70.31% | 56.02% | 68.65% |
| The inventive method | 93.21% | 83.41% | 82.25% | 86.29% |
As shown in Table 3, control methods is to tested 1, and tested 2 is very big with the classification results difference of tested 3, i.e. the method is inapplicable
In tested 3.The inventive method to tested 3 result more than 80%, result substantially gets a promotion, and averagely classification is correct
Rate is the most convincing.Result shows, the inventive method suitability is relatively strong, even if also can obtain the data that signal to noise ratio is higher
Preferably classifying quality.
The Integral Thought of this method is from result, and backstepping obtains Sample purification condition and range, can analyze off-line number well
According to, simultaneously for also having certain exploitativeness for the EEG's Recognition of real-time online, and the method is for solving
It is likely to occur in real time tested attention during brain wave acquisition to lax or the problem such as tired provides a kind of new thinking, has certain
Using value.
Claims (1)
1. one kind towards the sorting technique persistently imagining EEG signals, it is characterised in that comprise the following steps:
Step one, carries out transfer point detection；
Transfer point detection is mainly used in determining the distance threshold U at transfer point between sample；Assume to there is sample class label known
Training dataset X and unknown test data set Y of sample class label, in data set, sample is belonging respectively to multiple different class
Not, the method for computed range threshold value U is as follows:
(1) calculate in training dataset X the Euclidean distance between adjacent sample two-by-two, and save as vector Dist；
(2) according to the sample class label of training dataset, calculated for step (1) vector Dist is divided into two parts: if
Two adjacent samples belong to same category, and its distance is included into vector XnotransIn, it is designated as sample distance vector in class；If two is adjacent
Sample belongs to different classifications, and its distance is included into vector XtransIn, it is designated as sample distance vector between class；
(3) X is calculatednotransIn the maximum of all distances, be designated as U1；
(4) X is calculatedtransIn the minima of all distances, be designated as U2；
(5) according to step (3) and the result of step (4), by U1And U2In less that as threshold value U, i.e. U=min
(U1,U2)；
Step 2, carries out Sample purification；
Sample purification is mainly used in the distance range [a, b] determining between similar sample, and a, b are respectively minima and the maximum of distance
Value；Concrete grammar is as follows:
(1) training dataset X splitting into two parts, a part of X_train is used as grader and learns, and a part of X_test uses
Predict；
(2) use data set Training Support Vector Machines grader, obtain sorter model, then X_test data set is carried out pre-
Survey, obtain the prediction label of X_test data set；
(3) true tag of contrast X_test data set and prediction label, calculate all devices that are classified in data set and just predicting continuously
The true Euclidean distance between same category of adjacent sample, and it is designated as vector D；
(4) the distance vector D obtaining step (3) carries out statistical analysis, determines its distribution function, generally, is somebody's turn to do
Vector Normal Distribution or approximate normal distribution；
(5) distribution function of the vectorial D determined according to step (4), estimate its 95% confidence level hourly value estimate put
Letter is interval as the distance range [a, b] between similar sample；
Step 3, carries out Classification and Identification；
Classification and Identification is on the basis of trying to achieve threshold value U and distance range [a, b], the test data set unknown to sample class label
Being predicted, concrete grammar is as follows:
(1) use training dataset X Training Support Vector Machines grader, test data set Y is predicted, is tested
The sample predictions label vector predict of data set；
(2) the Euclidean distance d between adjacent sample in test data set is calculatedi, i=1,2 ..., N-1, N are in test data set
Number of samples；
(3) if diMore than or equal to U, illustrate to exist between i-th sample and i+1 sample classification transfer point, forward step to
(5)；
(4) if diLess than U, illustrate that i-th sample and i+1 sample belong to same category, then judge diWhether in scope
In [a, b], if diIn the range of [a, b], illustrate that the two sample is likely to be classified the sample that device is correctly predicted, by the two
The prediction label of sample adds in the judgement set of an entitled judgeset；If diNot in the range of [a, b], return step (2)；
(5) when a transfer point being detected, calculate judgeset and judge the quantity of each class label in set, return quantity
Most labels, as the class label of all samples between this transfer point and a upper transfer point, then empties judgeset and sentences
Disconnected set, forwards the step (3) detection to next transfer point to.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310273395.8A CN103345640B (en) | 2013-07-02 | 2013-07-02 | A kind of towards the sorting technique persistently imagining EEG signals |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310273395.8A CN103345640B (en) | 2013-07-02 | 2013-07-02 | A kind of towards the sorting technique persistently imagining EEG signals |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103345640A CN103345640A (en) | 2013-10-09 |
| CN103345640B true CN103345640B (en) | 2016-08-10 |
Family
ID=49280435
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310273395.8A Active CN103345640B (en) | 2013-07-02 | 2013-07-02 | A kind of towards the sorting technique persistently imagining EEG signals |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103345640B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104361345A (en) * | 2014-10-10 | 2015-02-18 | 北京工业大学 | Electroencephalogram signal classification method based on constrained extreme learning machine |
| CN107024987B (en) * | 2017-03-20 | 2020-04-14 | 南京邮电大学 | Real-time human brain attention testing and training system based on EEG |
| CN110824979B (en) * | 2019-10-15 | 2020-11-17 | 中国航天员科研训练中心 | Unmanned equipment control system and method |
| CN113283062A (en) * | 2021-05-10 | 2021-08-20 | 杭州电子科技大学 | Multivariable modeling method for motor imagery electroencephalogram signals |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101488189A (en) * | 2009-02-04 | 2009-07-22 | 天津大学 | Brain-electrical signal processing method based on isolated component automatic clustering process |
| CN102306303A (en) * | 2011-09-16 | 2012-01-04 | 北京工业大学 | Electroencephalography signal characteristic extraction method based on small training samples |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011123059A1 (en) * | 2010-03-31 | 2011-10-06 | Agency For Science, Technology And Research | Brain- computer interface system and method |
-
2013
- 2013-07-02 CN CN201310273395.8A patent/CN103345640B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101488189A (en) * | 2009-02-04 | 2009-07-22 | 天津大学 | Brain-electrical signal processing method based on isolated component automatic clustering process |
| CN102306303A (en) * | 2011-09-16 | 2012-01-04 | 北京工业大学 | Electroencephalography signal characteristic extraction method based on small training samples |
Non-Patent Citations (2)
| Title |
|---|
| research on heuristic feature extraction and classification of eeg signal based on bci data set;lijuan duan etc.;《Research journal of applied sciences,Engineering and technology》;20130121;第1008-1014页 * |
| 一种快速的脑电信号特征提取与分类方法;吴婷 等;《系统仿真学报》;20070930;第19卷(第18期);第4342-4344页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103345640A (en) | 2013-10-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP1864246B1 (en) | Spatio-temporal self organising map | |
| Lewicki | A review of methods for spike sorting: the detection and classification of neural action potentials | |
| CN110236536A (en) | A kind of brain electricity high-frequency oscillation signal detection system based on convolutional neural networks | |
| Hosseini et al. | The comparison of different feed forward neural network architectures for ECG signal diagnosis | |
| Saif-ur-Rehman et al. | SpikeDeeptector: a deep-learning based method for detection of neural spiking activity | |
| CN109948647A (en) | A kind of electrocardiogram classification method and system based on depth residual error network | |
| CN110139597A (en) | The system and method for being iterated classification using neuro-physiological signals | |
| CN110353702A (en) | A kind of emotion identification method and system based on shallow-layer convolutional neural networks | |
| CN103345640B (en) | A kind of towards the sorting technique persistently imagining EEG signals | |
| CN105590099B (en) | A kind of more people's Activity recognition methods based on improvement convolutional neural networks | |
| CN104166548B (en) | Deep learning method based on Mental imagery eeg data | |
| Jambhekar | Red blood cells classification using image processing | |
| CN105608432A (en) | Instantaneous myoelectricity image based gesture identification method | |
| CN106874929B (en) | Pearl classification method based on deep learning | |
| CN103971106A (en) | Multi-view human facial image gender identification method and device | |
| CN111091074A (en) | Motor imagery electroencephalogram signal classification method based on optimal region common space mode | |
| CN106127191B (en) | Brain electricity classification method based on WAVELET PACKET DECOMPOSITION and logistic regression | |
| CN113010013A (en) | Wasserstein distance-based motor imagery electroencephalogram migration learning method | |
| Parsaei et al. | SVM-based validation of motor unit potential trains extracted by EMG signal decomposition | |
| CN104463916B (en) | Eye movement fixation point measurement method based on random walk | |
| Qian et al. | Decision-level fusion of EEG and pupil features for single-trial visual detection analysis | |
| Alahmari et al. | Food state recognition using deep learning | |
| CN104679967A (en) | Method for judging reliability of psychological test | |
| CN107480635A (en) | Glance signal identification method and system based on bimodal classification model fusion | |
| CN113642525A (en) | Infant neural development assessment method and system based on skeletal points |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20131009 Assignee: Luoyang Miao Copper Trading Co.,Ltd. Assignor: Beijing University of Technology Contract record no.: X2024980000222 Denomination of invention: A Classification Method for Continuous Imagination EEG Signals Granted publication date: 20160810 License type: Common License Record date: 20240105 Application publication date: 20131009 Assignee: Luoyang Lexiang Network Technology Co.,Ltd. Assignor: Beijing University of Technology Contract record no.: X2024980000083 Denomination of invention: A Classification Method for Continuous Imagination EEG Signals Granted publication date: 20160810 License type: Common License Record date: 20240104 |