CN103077206A - Image semantic classifying searching method based on event-related potential - Google Patents
Image semantic classifying searching method based on event-related potential Download PDFInfo
- Publication number
- CN103077206A CN103077206A CN2012105833953A CN201210583395A CN103077206A CN 103077206 A CN103077206 A CN 103077206A CN 2012105833953 A CN2012105833953 A CN 2012105833953A CN 201210583395 A CN201210583395 A CN 201210583395A CN 103077206 A CN103077206 A CN 103077206A
- Authority
- CN
- China
- Prior art keywords
- picture
- user
- eeg signals
- erp
- semantic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention discloses an image semantic classifying searching method based on event-related potential, which comprises the following steps: (1) showing various semantic related images to a user in turn and causing the user to generate corresponding original electroencephalograms for various semantic related images; (2) while showing the images, collecting the original electroencephalograms generated by the user for the shown images, and amplifying and digitally treating the original electroencephalograms; and (3) performing ERP (Event-Related Potential) detecting analysis on the electroencephalograms treated in the step (2) and confirming the images searched by the user in the semantic related images shown in the step (1) according to an ERP detecting analysis result. According to the image semantic classifying searching method based on the event-related potential provided by the invention, the electroencephalograms for non-semantic images from human are directly collected, the images are quickly classified and searched according to the difference in the electroencephalograms and the efficiency for classifying and searching for the images is increased.
Description
Technical field
The present invention relates to picture classification method, be specifically related to a kind of picture semantic classified search method based on event related potential.
Background technology
The picture of identical semanteme is namely found out in the picture semantic classification from pictures, traditional method has two kinds: the first be by picture is added metadata (such as: the explanation of captions, keyword or image etc.), utilize the note word in the metadata to finish picture classification; The second is to utilize the computer vision correlation technique to finish the classification of picture as substrate take image content (semanteme that picture comprises).
All there is a problem in these two kinds of methods, namely all need to explain or label for picture adds.Second method need to be utilized the relevant learning algorithm of label training of picture, manually to picture explain time-consuming, require great effort and cost dearly, although the method for utilizing computer vision to carry out photo annotation in the prior art can improve note efficient, but the learning process of computerized algorithm needs accurate graphic template, and the greatest differences of fuzzy, the picture of image content and template all can exert an influence to the operation stability of computing machine picture categorizing system.
The human visual system is considered to have natural robustness and good extensive processing power aspect picture processing, even the time that picture is presented is very short, such as a few tens of milliseconds or hundreds of millisecond, human vision system also can " catch " a large amount of Useful Informations, thereby (the list of references: ProceedingsoftheIEEE of the certain content in the identification picture, Vol.98, No.3, March2010,462-463), drawing immediately such as people can sweep a width of cloth picture wherein has nobody, what is more important, and the people can accomplish fine identification under image content presents the situation of various variations.
In order to create the computer vision system that has the identical information processing power with the human visual system, dropped into a large amount of effort, but rarely had successfully, do not possess general applicability.
Event related potential (Event Related Potential, ERP) (list of references: StevenJ.Luck, the event related potential basis, publishing house of East China Normal University, 2009) be a kind of brain electric potential relevant with memory with psychological cognition, reacted the state of the brain in the human cognitive process.The generation of event related potential need to possess two conditions: the first, and the appearance of object event must be a small probability event; The second, for the appearance of all events, the appearance of object event should be at random.
The especially vision induced P300 of event related potential P300 has obtained extensive and deep research and application, when one group of event (comprising object event and non-object event) when occurring continuously, the appearance of object event can the evoked brain potential signal produces P300, utilizes the composition of P300 can infer the appearance of object event.
In the prior art, brain machine interface system based on event related potential generally includes: the EEG signals that applies stimulation to the user is brought out system, gathers the eeg signal acquisition system of user's EEG signals and the electroencephalogramsignal signal analyzing system that the EEG signals that collects is analyzed.
Can realize the mutual of human brain and computing machine based on brain machine interface system, therefore, can consider directly to utilize the human visual system to carry out classification and the search of picture.
Summary of the invention
The invention provides a kind of picture semantic classified search method based on event related potential, directly gather the people for EEG signals that need not semantic picture, according to the difference of EEG signals picture is searched classifiably rapidly, improved the efficient of picture classification search.
A kind of picture semantic classified search method based on event related potential may further comprise the steps:
(1) presents successively the relevant picture of all kinds of semantemes to the user, make the user produce corresponding original EEG signals to the relevant picture of all kinds of semantemes;
(2) when presenting picture, gather the original EEG signals that the user produces the picture that presents, and original EEG signals is amplified and digitized processing;
(3) EEG signals that processing in the step (2) is obtained is carried out ERP and is detected analysis, detects the result who analyzes according to ERP, determines the picture that the user need to search in the picture that the semanteme that presents in step (1) is correlated with.
The picture that the user will search among the present invention is not to refer to a certain pictures, but contains a class picture of identical content, is the picture of identical semanteme yet, and for example, the user need to search for the picture of tiger, and the picture that then comprises tiger all is the search target.
The needs of User and the difference of applied environment can be to the orders of picture appearance, and the parameters such as picture that the number of times that picture occurs and picture occur are made amendment, to arrive best picture search result.
As preferably, present successively the relevant picture of all kinds of semantemes to the user in the described step (1) before, be provided for the countdown time that allows the user focus one's attention on.The user focuses one's attention on, and helps to obtain best picture semantic classifying search results.
If the picture of identical semanteme occurs continuously, then the ERP information strength in the EEG signals can decrease, therefore, in order to guarantee that the ERP information strength in the EEG signals meets the demands, preferably, when presenting successively the relevant picture of all kinds of semantemes to the user in the described step (1), the picture that all kinds of semantemes are relevant occurs and the discontinuous appearance of picture of identical semanteme at random.
As preferably, the time interval of the picture that all kinds of semantemes are relevant in the described step (1) is 500ms, and the picture of semanteme of the same race occurs 10 times.The picture of semanteme of the same race occurs 10 times, can reach best picture semantic classifying search results.
As preferably, in step (3), EEG signals is carried out before ERP detect to analyze, EEG signals is carried out pre-service, described pre-service comprises: remove the baseline wander in the EEG signals, use the relevant electric artefact of eye of subduing in the algorithm removal EEG signals of electro-ocular signal, use band-pass filter.
Through pre-service, increase the signal to noise ratio (S/N ratio) of EEG signals, strengthen the ERP composition that exists in the EEG signals, mainly refer to the P300 composition among the present invention.
As preferably, in the described step (3) EEG signals being carried out the detailed process that ERP detect to analyze is: the time point that occurs according to different semantic pictures, EEG signals is divided into some data segments, the data segment that the picture of identical semanteme is corresponding carries out superposed average, obtain the proper vector of the average ERP that the picture of every kind of semanteme brings out, proper vector according to average ERP obtains corresponding picture, and this picture is the result of picture semantic classified search.
When the picture of target semanteme occurs, can bring out user's generation with the EEG signals of P300 composition, in the proper vector of the average ERP that obtains, detect the P300 composition, and determine the picture that occurs in the corresponding time period according to the time of occurrence of P300, use Successive Regression discriminant analysis method (stepwisediscriminant analysis) in this process, proper vector for given average ERP, find out wherein to the larger characteristic component of picture classification contribution by the Successive Regression discriminant analysis method, but increase the calibration of EEG signals, take into account the P300 signal characteristic of different user, obtain preferably picture semantic classifying search results.
Although the present invention is the picture semantic classified search method under the simple normal form, it itself is a direction highly significant that but the EEG signals of utilizing the people is carried out the picture semantic classified search, especially after following realization single detects ERP (such as the P300 composition among the present invention), can realize the picture semantic classified search of larger pictures, for example can realize the function of similar google picture searching, utilize EEG signals from the picture set of a large amount of the unknowns, to find out the picture of wanting.Picture semantic classified search method of the present invention has potential using value in military affairs.
Description of drawings
Fig. 1 the present invention is based on the system schematic of the picture semantic classified search method of event related potential for realization;
Fig. 2 the present invention is based on the time distribution schematic diagram that picture stimulates in the picture semantic classified search method of event related potential;
Fig. 3 the present invention is based on the process flow diagram of processing original EEG signals in the picture semantic classified search method of event related potential;
Fig. 4 the present invention is based on that ERP detects the process flow diagram of analyzing in the picture semantic classified search method of event related potential.
Embodiment
Below in conjunction with accompanying drawing, a kind of picture semantic classified search method based on event related potential of the present invention is described in detail.
As shown in Figure 1, a kind of picture semantic classified search method based on event related potential may further comprise the steps:
(1) presents successively the relevant picture of all kinds of semantemes to the user, make the user produce corresponding original EEG signals to the relevant picture of all kinds of semantemes;
(2) when presenting picture, gather the original EEG signals that the user produces the picture that presents, and original EEG signals is amplified and digitized processing;
(3) EEG signals that processing in the step (2) is obtained is carried out ERP and is detected analysis, detects the result who analyzes according to ERP, determines the picture that the user need to search in the picture that the semanteme that presents in step (1) is correlated with.
Having selected the picture of 8 kinds of animals (frog, fish, dog, tiger, horse, cat, bird and sheep) in the present embodiment, is the picture of 8 kinds of semantemes also, and every kind of animal has 15 different pictures, all comes from the google picture.
In order to improve the accuracy of picture identification, original image is processed, reject animal other guide in addition, unified picture size.
Before each picture semantic classified search beginning, countdown in 5 seconds is set, in the numeral of the center Screen demonstration countdown that Shows Picture, give the user time to prepare, the user is focused one's attention on, adjust sitting posture.
The present invention presents picture by display device to the user, and final picture semantic classifying search results also feeds back to the user by display device.
Focus one's attention in search procedure for the ease of the user, it is light grey that the screen background that Shows Picture is, and with the central authorities of picture presents at screen, guarantee in the process of picture semantic classified search, user's eyes and head are basic to keep motionless, reduces the myoelectricity that causes because of user's head rotation to the interference of EEG signals.
The time of picture presents and the order of picture presents can be set according to needs, and the time of every width of cloth Image Display is 500ms in the present embodiment, and the picture of same semanteme can not occur continuously.
In each search, the number of times of every kind of animal picture appearance can be set, and the number of times that the picture of every kind of animal occurs equates, generally, the number of times that every kind of animal appearance is set is 10 times, namely picks out 10 from 15 pictures of every kind of animal, amount to 80 pictures, 80 pictures are divided into 10 takes turns successively and presents, and every kind of animal picture occurred once during each was taken turns, and the discontinuous appearance of the picture of every kind of animal in the adjacent wheels.
When the number of times of every kind of animal picture appearance also is the eeg signal classification processing, the stacking fold of EEG signals, in general, for same user, the number of times of stack is fewer, the accuracy that classification is processed is lower, and therefore the setting of the number of times of every kind of animal picture appearance should decide in the feedback of training stage according to the user.When the feedback of training stage did not reach certain accuracy, number of times should tune up; On the contrary, if the feedback of training stage reaches certain accuracy, inferior numerical value can be turned down, under the prerequisite that guarantees search effect, reduce the required time of training.According to experience, the number of times of every kind of animal picture appearance is set at 10 o'clock, for most user preferably training and search effect is arranged.
The original EEG signals of utilizing electrode cap collection user that different pictures are produced, electrode cap is embedded in a plurality of electrodes on the resilient cap with certain space distribution, and scalp brain electricity collects by metal or the metallic compound electrode that contacts with scalp.Have contact resistance between electrode and the user's scalp, contact resistance impedance general control below 20k Ω to obtain noise and the less original EEG signals of baseline wander.
The electrode channel that uses is optional, and reference electrode places the nose place, in addition, be affixed on outward directly over the eyeball and under bipolar eye electrode eliminate nictation to the impact of original EEG signals.
The analog to digital conversion sampling rate of original EEG signals generally is made as 1000Hz, and SC service ceiling is the analog filter low pass filter filters out high frequency interference of 200Hz.
Information (time that picture occurs and kind etc.) with the original EEG signals of the user who collects and every pictures deposits in the PC in real time by USB interface.
As shown in Figure 3, after the amplification of original EEG signals process and the digitized processing, further carry out pre-service, pre-service comprises: utilize once or the Quadratic Spline Interpolation method reduces sampling spot to eliminate baseline wander, use the relevant algorithm of subduing of electro-ocular signal to remove the eye electrical interference, adopt the bandpass filter of 0.3-30Hz to carry out filtering.
The time point that occurs according to every pictures will be divided into some data segments through pretreated EEG signals, choose picture appearance 200ms before during division to the afterwards data segment of 800ms total 1000ms duration occurring, and the 200ms signal before occurring take pictures carries out baseline correction as baseline, the data segment that the picture of identical semanteme is corresponding carries out superposed average, obtains the mean value of the ERP that the picture of every kind of semanteme brings out.
As shown in Figure 2, for example the picture between time point 3 and the time point 4 is the picture of tiger, picture between time point 78 and the time point 79 also is the picture of tiger, ERP in the corresponding time period that then will obtain carries out superposed average, obtain the mean value of ERP of the picture Induced by Stimulation of tiger, the picture of every kind of animal occurs 10 times in the present embodiment, and the ERP of the picture Induced by Stimulation in corresponding 10 time periods is carried out superposed average, obtains the mean value of the ERP that semantic picture brings out.
The proper vector of average ERP can adopt the whole bag of tricks to extract, for example, extract picture in the mean value of the ERP that the picture of every kind of semanteme brings out stimulate occur after the part of 0～800ms, when sample frequency is 1000Hz, have 800 sampled points, adopting the down-sampled factor is that 20 the down-sampled method of running mean is reduced to 40 to sampled point, then 40 sampled points in the mean value of the ERP of a plurality of passages is connected the proper vector that obtains average ERP from beginning to end.
As shown in Figure 4, method training classifier by stepwise regression analysis of the prior art in the present embodiment, calculate the weight vector of SWDA (Stepwise Linear Discriminant Analysis), in the process of stepwise regression analysis, characteristic component to the proper vector of the average ERP that obtains screens, thereby reduce the number of the characteristic component that needs contrast, improve counting yield, but make simultaneously the calibration between the data segment with P300 composition and the data segment that does not have the P300 composition reach local optimum, be conducive to improve the accuracy of picture classification search.
The proper vector of the average ERP that the picture of every kind of semanteme is corresponding is input in the sorter that has trained, obtain output valve corresponding to proper vector of each average ERP, then find out the picture of the appearance in the maximum corresponding time period of output valve, the corresponding animal of this picture is the animal that will search for.
For example, what the output valve of the maximum of the proper vector of average ERP was corresponding is the time period that the tiger picture occurs, and then the picture in requisition for search is the picture of tiger.
During training classifier, before each picture semantic classified search, the user is apprised of a target semanteme (for example picture of tiger), in the browsing pictures process, all pictures (all comprise the picture of tiger) that meet the target semanteme all are object events, owing to the semanteme of corresponding picture is limited in advance, therefore after collecting original EEG signals, can EEG signals be carried out segmentation and stack according to semanteme, and calculate the weight vector (being used for the signal classification of operational phase) of stepwise regression analysis sorter; In operational phase, choosing of picture is the same with the training stage, different is, every target semanteme of taking turns is that the user is from determined, after having browsed picture, the weight vector that sorter obtained according to the training stage compares different semantic signals, obtains the user and wishes the picture semantic of looking for, and demonstrate the picture that all should semanteme.
When using the present invention that picture is searched classifiably, it is very important that the user keeps notice to concentrate in the process of browsing pictures, the quality that is related to EEG signals, the user can count to keep notice by the number of times that the semantic picture of target is occurred, in addition, also can focus one's attention on by clicking the mouse when Target Photo occurs, the record that the user clicks the mouse can be used for assessing the accuracy that degree that user's notice concentrates and picture classification search are carried out.
The record that the user clicks the mouse also can apply in the analytic process of ERP signal, for example, Target Photo is the picture of tiger, when the picture of tiger occurring, the user does not click the mouse, then the EEG signals that the tiger picture occurs is carried out the ERP mean time, remove the corresponding data segment of picture that the user does not click the mouse.
When using the present invention to carry out the picture semantic classified search, the user will often keep the eyeball of not batting an eyelid, be failure to actuate, therefore fatigue conditions can appear after searching for certain hour, affect the accuracy of Search Results, especially user's eyes are easier to fatigue, therefore in use, answer the state of User self to carry out suitable rest.
Claims (6)
1. the picture semantic classified search method based on event related potential is characterized in that, may further comprise the steps:
(1) presents successively the relevant picture of all kinds of semantemes to the user, make the user produce corresponding original EEG signals to the relevant picture of all kinds of semantemes;
(2) when presenting picture, gather the original EEG signals that the user produces the picture that presents, and original EEG signals is amplified and digitized processing;
(3) EEG signals that processing in the step (2) is obtained is carried out ERP and is detected analysis, detects the result who analyzes according to ERP, determines the picture that the user need to search in the picture that the semanteme that presents in step (1) is correlated with.
2. the picture semantic classified search method based on event related potential as claimed in claim 1, it is characterized in that, before presenting successively the relevant picture of all kinds of semantemes to the user in the described step (1), be provided for the countdown time that allows the user focus one's attention on.
3. the picture semantic classified search method based on event related potential as claimed in claim 2, it is characterized in that, when presenting successively the relevant picture of all kinds of semantemes to the user in the described step (1), the picture that all kinds of semantemes are relevant occurs and the discontinuous appearance of picture of identical semanteme at random.
4. the picture semantic classified search method based on event related potential as claimed in claim 3 is characterized in that, the time interval of the picture that all kinds of semantemes are relevant in the described step (1) is 500ms, and the picture of semanteme of the same race occurs 10 times.
5. the picture semantic classified search method based on event related potential as claimed in claim 4, it is characterized in that, in step (3), EEG signals is carried out before ERP detect to analyze, EEG signals is carried out pre-service, described pre-service comprises: remove the baseline wander in the EEG signals, use the relevant electric artefact of eye of subduing in the algorithm removal EEG signals of electro-ocular signal, use band-pass filter.
6. the picture semantic classified search method based on event related potential as claimed in claim 5, it is characterized in that, in the described step (3) EEG signals being carried out the detailed process that ERP detect to analyze is: the time point that occurs according to different semantic pictures, EEG signals is divided into some data segments, the data segment that the picture of identical semanteme is corresponding carries out superposed average, obtain the proper vector of the average ERP that the picture of every kind of semanteme brings out, proper vector according to average ERP obtains corresponding picture, and this picture is the result of picture semantic classified search.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012105833953A CN103077206A (en) | 2012-12-27 | 2012-12-27 | Image semantic classifying searching method based on event-related potential |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012105833953A CN103077206A (en) | 2012-12-27 | 2012-12-27 | Image semantic classifying searching method based on event-related potential |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103077206A true CN103077206A (en) | 2013-05-01 |
Family
ID=48153736
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012105833953A Pending CN103077206A (en) | 2012-12-27 | 2012-12-27 | Image semantic classifying searching method based on event-related potential |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103077206A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103886280A (en) * | 2013-12-30 | 2014-06-25 | 浙江大学 | Rapid face identification method based on one-time ERP detection |
| CN104503580A (en) * | 2014-12-25 | 2015-04-08 | 天津大学 | Identification method of steady-state visual evoked potential brain-computer interface target |
| CN104731341A (en) * | 2015-04-01 | 2015-06-24 | 浙江大学 | Face image retrieval method based on EEG and computer vision |
| CN105072457A (en) * | 2015-08-17 | 2015-11-18 | 无锡天脉聚源传媒科技有限公司 | Method and device for processing television program images |
| CN106420130A (en) * | 2015-08-13 | 2017-02-22 | 深圳市飘爱健康科技有限公司 | Memory assisting device and system |
| CN108345675A (en) * | 2018-02-11 | 2018-07-31 | 广东欧珀移动通信有限公司 | Photograph album display methods and relevant device |
| CN109074487A (en) * | 2016-05-11 | 2018-12-21 | 微软技术许可有限责任公司 | It is read scene cut using neurology into semantic component |
| CN113974658A (en) * | 2021-10-28 | 2022-01-28 | 天津大学 | Semantic visual image classification method and device based on EEG time-sharing spectrum Riemann |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7043474B2 (en) * | 2002-04-15 | 2006-05-09 | International Business Machines Corporation | System and method for measuring image similarity based on semantic meaning |
| CN101339455A (en) * | 2008-08-07 | 2009-01-07 | 北京师范大学 | Brain machine interface system based on human face recognition specific wave N170 component |
| CN101690659A (en) * | 2009-09-29 | 2010-04-07 | 华东理工大学 | Brain wave analysis method |
| CN102722244A (en) * | 2012-05-25 | 2012-10-10 | 西安交通大学 | Steady-state evoked potential brain-computer interface method based on motion turning vision sensing |
-
2012
- 2012-12-27 CN CN2012105833953A patent/CN103077206A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7043474B2 (en) * | 2002-04-15 | 2006-05-09 | International Business Machines Corporation | System and method for measuring image similarity based on semantic meaning |
| CN101339455A (en) * | 2008-08-07 | 2009-01-07 | 北京师范大学 | Brain machine interface system based on human face recognition specific wave N170 component |
| CN101690659A (en) * | 2009-09-29 | 2010-04-07 | 华东理工大学 | Brain wave analysis method |
| CN102722244A (en) * | 2012-05-25 | 2012-10-10 | 西安交通大学 | Steady-state evoked potential brain-computer interface method based on motion turning vision sensing |
Non-Patent Citations (2)
| Title |
|---|
| 吴边等: "基于P300电位的新型BCI中文输入虚拟键盘系统", 《电子学报》 * |
| 陆卫红等: "面孔认知的事件相关电位研究进展", 《人类工效学》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103886280A (en) * | 2013-12-30 | 2014-06-25 | 浙江大学 | Rapid face identification method based on one-time ERP detection |
| CN104503580A (en) * | 2014-12-25 | 2015-04-08 | 天津大学 | Identification method of steady-state visual evoked potential brain-computer interface target |
| CN104503580B (en) * | 2014-12-25 | 2018-04-13 | 天津大学 | A kind of recognition methods to Steady State Visual Evoked Potential brain-computer interface target |
| CN104731341A (en) * | 2015-04-01 | 2015-06-24 | 浙江大学 | Face image retrieval method based on EEG and computer vision |
| CN106420130A (en) * | 2015-08-13 | 2017-02-22 | 深圳市飘爱健康科技有限公司 | Memory assisting device and system |
| CN106420130B (en) * | 2015-08-13 | 2019-03-08 | 深圳市飘爱健康科技有限公司 | Memory aid and system |
| CN105072457A (en) * | 2015-08-17 | 2015-11-18 | 无锡天脉聚源传媒科技有限公司 | Method and device for processing television program images |
| CN109074487A (en) * | 2016-05-11 | 2018-12-21 | 微软技术许可有限责任公司 | It is read scene cut using neurology into semantic component |
| CN108345675A (en) * | 2018-02-11 | 2018-07-31 | 广东欧珀移动通信有限公司 | Photograph album display methods and relevant device |
| CN113974658A (en) * | 2021-10-28 | 2022-01-28 | 天津大学 | Semantic visual image classification method and device based on EEG time-sharing spectrum Riemann |
| CN113974658B (en) * | 2021-10-28 | 2024-01-26 | 天津大学 | Semantic visual image classification method and device based on EEG time-sharing frequency spectrum Riemann |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103077206A (en) | Image semantic classifying searching method based on event-related potential | |
| Bostanov | BCI competition 2003-data sets Ib and IIb: feature extraction from event-related brain potentials with the continuous wavelet transform and the t-value scalogram | |
| CN110811609B (en) | Epileptic spike intelligent detection device based on self-adaptive template matching and machine learning algorithm fusion | |
| Cecotti et al. | Convolutional neural network with embedded Fourier transform for EEG classification | |
| CN111329474A (en) | Electroencephalogram identity recognition method and system based on deep learning and information updating method | |
| Carlson et al. | Multichannel electrophysiological spike sorting via joint dictionary learning and mixture modeling | |
| CN111783942B (en) | Brain cognitive process simulation method based on convolutional recurrent neural network | |
| Liu et al. | Automatic extracellular spike detection with piecewise optimal morphological filter | |
| CN103077205A (en) | Method for carrying out semantic voice search by sound stimulation induced ERP (event related potential) | |
| CN111091074A (en) | Motor imagery electroencephalogram signal classification method based on optimal region common space mode | |
| CN104635934A (en) | Brain-machine interface method based on logic thinking and imaginal thinking | |
| CN109255309A (en) | Brain electricity and eye movement fusion method and device towards Remote Sensing Target detection | |
| CN109770900A (en) | Brain-computer interface based on convolutional neural networks instructs delivery method, system, device | |
| CN104127179A (en) | Electroencephalogram (EEG) feature extraction method based on dominant electrode combination and empirical mode decomposition (EMD) | |
| CN107704881A (en) | A kind of data visualization processing method and processing device based on animal electroencephalogramrecognition recognition | |
| Ghaheri et al. | Extracting common spatial patterns from EEG time segments for classifying motor imagery classes in a Brain Computer Interface (BCI) | |
| Cecotti | Convolutional neural networks for event-related potential detection: impact of the architecture | |
| CN109662710A (en) | A kind of EMG Feature Extraction based on convolutional neural networks | |
| Xiao et al. | Classification of motor imagery EEG signals based on energy entropy | |
| CN104731341A (en) | Face image retrieval method based on EEG and computer vision | |
| CN114601474A (en) | Source domain sample screening method for motor imagery transfer learning | |
| CN114521903A (en) | Electroencephalogram attention recognition system and method based on feature selection | |
| Joadder et al. | A new way of channel selection in the motor imagery classification for BCI applications | |
| Wu et al. | Intelligent artefact identification in electroencephalography signal processing | |
| Goel et al. | On the use of spiking neural network for EEG classification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20130501 |