CN108764008B - Method for detecting movement intention based on combination of dynamic stopping strategy and integrated learning - Google Patents

Method for detecting movement intention based on combination of dynamic stopping strategy and integrated learning Download PDF

Info

Publication number
CN108764008B
CN108764008B CN201810193673.1A CN201810193673A CN108764008B CN 108764008 B CN108764008 B CN 108764008B CN 201810193673 A CN201810193673 A CN 201810193673A CN 108764008 B CN108764008 B CN 108764008B
Authority
CN
China
Prior art keywords
data
state data
time
motion state
key pressing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810193673.1A
Other languages
Chinese (zh)
Other versions
CN108764008A (en
Inventor
明东
王坤
许敏鹏
张珊珊
何川
陈龙
柯余峰
周鹏
何峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN201810193673.1A priority Critical patent/CN108764008B/en
Publication of CN108764008A publication Critical patent/CN108764008A/en
Application granted granted Critical
Publication of CN108764008B publication Critical patent/CN108764008B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • 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)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

The invention discloses a method for detecting a movement intention based on a dynamic stopping strategy combined with integrated learning, which comprises the following steps: an off-line modeling stage: after extracting features of the resting state data and the first pre-motion state data, establishing a linear discriminant analysis model for identifying whether a current user is about to move or not; after extracting the characteristics of the second pre-motion state data, respectively constructing sub-classifiers with the motion state data characteristics for detecting the accurate moment of motion; and (3) an online testing stage: and after extracting the features from the real-time intercepted data, sending the data into an LDA model, outputting a decision value, if the decision value is greater than a set threshold value, sending the data into a constructed sub-classifier, outputting a result if the obtained vote is greater than the set threshold value by using a voting strategy, and otherwise, waiting for the input of the next data feature. The invention realizes the on-line movement intention detection and can obtain a better detection result on the basis of the calculation cost of the control system.

Description

Method for detecting movement intention based on combination of dynamic stopping strategy and integrated learning
Technical Field
The invention relates to the field of movement intention detection, in particular to a method for detecting movement intention based on dynamic stopping strategy combined with integrated learning.
Background
Brain-Computer Interface (BCI) is a communication control system that does not rely on the normal output channels of the peripheral nerves and muscles of the Brain. Motor intent is the instructional decision a person makes to control peripheral nerves and skeletal muscles to perform an intended action in preparation for performing or imagining motor, invoking cognitive resources associated with the brain and motor. Colloquially, motor intent is the mental preparation of a person's brain for movement prior to the onset of movement, or the initial mental state of the central nervous system planning to engage in movement. Research shows that the movement intention can be detected by analyzing relevant characteristics of the electroencephalogram signals. Therefore, the movement intention can be used as a control instruction of the BCI, can be used for controlling the artificial limb to assist the movement of a patient with impaired limb movement function but normal brain function activity, and can be used for controlling the mechanical exoskeleton to enhance the movement capability of the healthy user (such as a soldier) in a special scene.
Motor cortex-related potentials (MRCPs) are low-frequency potentials generated when a human body invokes motor-related cognitive resources in the brain while performing suggestive or voluntary exercise. MRCPs are rich in brain motor intention information and therefore are of great interest to researchers. MRCPs mainly comprise three important components: (1) the motion preparation potential is generated 1-2 s before the motion starting moment and is expressed as a negative potential with a gently reduced amplitude and then a sharply reduced amplitude; (2) the motor potential, which usually occurs 150ms before the start of the motor, is followed by the motor reserve potential, which is usually expressed as the minimum of the negative potential before the motor; (3) motion detection potential: the electric potential following the movement lasts for 1s or longer after the movement, and is a complex cortical electric potential with the positive and negative values changing alternately.
Many researchers have designed experiments to perform motor intent detection by analyzing MRCPs. However, the current related research results are not ideal and still cannot meet the requirement of controlling the use of BCI in real life. With the gradual development of feature extraction and machine learning algorithms, the movement intention detection with high accuracy rate becomes possible.
Ensemble learning by combining multiple learners often results in a significantly superior generalization capability over a single learner. However, the integrated learning algorithm has high computational cost and requires a high-performance hardware platform.
Disclosure of Invention
The invention provides a method for detecting movement intention based on a dynamic stopping strategy and integrated learning, which realizes on-line movement intention detection, can obtain a better detection result on the basis of the calculation cost of a control system, and is described in detail as follows:
a method of detecting motor intent in conjunction with ensemble learning based on a dynamic stopping strategy, the method comprising the steps of:
firstly, an off-line modeling stage:
collecting a sample of 200 times of autonomous key pressing of a subject, and randomly intercepting data with the time interval between a data center before key pressing and the key pressing time not less than 150ms as resting state data and data with the time interval between the data center before key pressing and the key pressing time not more than 150ms as first pre-movement state data by taking the key pressing time as 0 time and 1s as window width;
after extracting features of the resting state data and the first pre-motion state data, establishing a linear discriminant analysis model for identifying whether a current user is about to move or not;
intercepting data of-500 ms-500 s as motion state data, and intercepting data at a certain moment before and after the key as second pre-motion state data;
after extracting the characteristics of the second pre-motion state data, respectively constructing sub-classifiers with the motion state data characteristics for detecting the accurate moment of motion;
II, an online testing stage:
and after extracting the features from the real-time intercepted data, sending the data into an LDA model, outputting a decision value, if the decision value is greater than a set threshold value, sending the data into a constructed sub-classifier, outputting a result if the obtained vote is greater than the set threshold value by using a voting strategy, and otherwise, waiting for the input of the next data feature.
The method further comprises the following steps:
collecting electroencephalograms by using an electrode cap and an electroencephalogram amplifier produced by Neuroscan, and collecting 18 channel electroencephalograms at positions corresponding to the scalp of a functional exercise area by taking the top of the head as a reference and the forehead leaf as the ground;
the method is characterized in that the lead is placed according to the 10-20 international standard lead position, the sampling frequency is 1000Hz, and a 50Hz wave trap is adopted to remove power frequency interference.
The intercepting of the data at a certain moment before and after the key is specifically the second pre-motion state data:
the time difference between the data center before and after the key pressing and the time of the key pressing is-140 ms, -130ms, -120ms, -110ms, -100ms, -90ms, -80ms, -70ms, -60ms, -50ms, -40ms, -30ms, -20ms, -10ms, 10ms and 20ms respectively, and the data are second pre-motion state data.
The method sequentially adopts discrimination space mode filtering, typical correlation analysis and template matching feature identification methods.
Further, the method further comprises: and (3) down-sampling the electroencephalogram signal to 200Hz, and filtering the signal by using a Chebyshev filter at 0.5-45 Hz.
The technical scheme provided by the invention has the beneficial effects that:
1. the invention realizes real-time detection of movement intention, and overcomes the problem of weak generalization ability of the original single learner by combining a plurality of learners through integrated learning;
2. the dynamic stopping strategy based on linear discriminant analysis makes up the defect of high integrated learning calculation cost, and simultaneously ensures the accuracy and the operation time;
3. the invention is expected to provide key technical support for the application of the movement intention detection system in actual life;
4. further research can open up a new development direction for the development of BCI, and is expected to obtain considerable social and economic benefits.
Drawings
FIG. 1 is a flow chart of a method for detecting intent to move based on dynamic stopping strategy in combination with ensemble learning;
FIG. 2 is a schematic diagram of the motion intent detection system design;
FIG. 3 is a schematic diagram of offline data interception and modeling;
fig. 4 is a flow chart of feature extraction.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention are described in further detail below.
Example 1
The embodiment of the invention provides a method for detecting a movement intention based on a dynamic stopping strategy and integrated learning, which is described in detail in the following description with reference to fig. 1 and 2:
when a person has an intention to exercise, a specific signal pattern existing in the brain electricity, i.e., motor-related cortical potentials (MRCPs), is generated before and after the exercise. Because the device is rich in a large amount of motion information and has strict time locking and phase locking characteristics, the device has attracted extensive attention of researchers. The embodiment of the invention designs a dynamic stopping strategy combined integrated learning method based on linear discriminant analysis, which can improve the accuracy of movement intention detection.
101: an online experiment platform is set up, electroencephalogram data of a user are read in real time, collected data are preprocessed, and features are extracted;
102: and selecting whether to perform the next motion intention detection analysis or not by using a dynamic stopping strategy based on linear discriminant analysis, detecting and outputting the motion intention by using an integrated learning algorithm, and finally analyzing the accuracy of the system.
In summary, the embodiment of the present invention designs a method for detecting an exercise intention based on a dynamic stopping strategy combined with ensemble learning, so as to reduce the calculation cost when a user has no exercise intention, and is of great significance for bringing the exercise intention to practical use in BCI.
Example 2
The scheme of example 1 is further described below with reference to fig. 2-4 and specific examples, which are described in detail below:
fig. 2 is a schematic system design diagram according to an embodiment of the present invention. The design mainly comprises: brain electrical signal collection and computer signal processing.
An electrode cap and an electroencephalogram amplifier manufactured by Neuroscan company are used for collecting electroencephalograms, 18 channels of electroencephalogram signals (Fc5, Fc3, Fc1, Fc2, Fc4, Fc6, C5, C3, C1, C2, C4, C6, Cp5, Cp3, Cp1, Cp2, Cp4 and Cp6, which are arranged according to 10-20 international standard lead positions) are collected at positions corresponding to the functional areas of motor work by taking the vertex of the head as a reference and the forehead as the ground, and the sampling frequency is 1000Hz, and a 50Hz trap is adopted to remove power frequency interference. Computer signal processing uses MATLAB software to implement various signal processing algorithms.
In the experimental process, a user sits on a comfortable chair, presses keys randomly according to the independent movement intention, and triggers an event synchronization unit (label) at the same time of the keys, namely, the time for marking the occurrence of the keys is recorded together with an electroencephalogram signal and is used for post-signal processing and intercepting data and system performance analysis. The experiment is divided into two stages: an off-line modeling phase and an on-line testing phase.
Firstly, an off-line modeling stage:
the method comprises the steps of collecting samples of 200 times of self-key pressing of a subject, wherein the key pressing time is 0 time, 1s is a window width (a data interception schematic diagram is shown in fig. 3(a)), data before key pressing (the time interval between the data center and the key pressing time is not less than 150ms) is randomly intercepted to be resting state data (200 samples), data before key pressing (the time interval between the data center and the key pressing time is not more than 150ms) is randomly intercepted to be first pre-movement state data (200 samples), and after the characteristics of the two types of samples (namely the resting state data and the first pre-movement state data) are extracted, a Linear Discriminant Analysis (LDA) model is established for identifying whether a current user is about to move or not (shown in fig. 3 (b)).
Intercepting data of-500 ms to 500s as motion state data (200 samples, the data center is the key press time), intercepting data before and after the key press (the time difference between the data center and the key press time is respectively-140 ms, -130ms, -120ms, -110ms, -100ms, -90ms, -80ms, -70ms, -60ms, -50ms, -40ms, -30ms, -20ms, -10ms, 10ms, 20ms) as second pre-motion state data (16 x 200 samples, wherein 16 is the time difference number between the data center and the key press time, 200 samples are intercepted in each time difference), extracting characteristics from the second pre-motion state data, and respectively comparing the extracted characteristics with motion state data characteristics (the process of extracting characteristics from motion state data is the same as the process of extracting characteristics from the rest state data and the first pre-motion state data), the embodiment of the present invention does not describe this, and a sub-classifier is constructed for detecting the accurate time of the motion (see fig. 3 (c)).
II, an online testing stage:
as shown in fig. 2, after extracting features from the captured data in real time, sending the captured data to the LDA model, outputting a decision value, if the decision value is greater than a set threshold, sending the decision value to a sub-classifier for ensemble learning, and performing further processing, otherwise, waiting for the input of the next data feature; and (4) inputting the features of the sub-classifiers, outputting a result if the obtained vote is greater than a set threshold value by using a voting strategy, and otherwise, waiting for the input of the next data feature.
In summary, the data processing flow for detecting the exercise intention in the present invention is described.
In an embodiment of the present invention, an algorithm for feature extraction includes: three methods, namely, Discrimination Spatial Pattern (DSP) filtering, Canonical Correlation Analysis (CCA), and template matching, and LDA is used for Pattern recognition. The electroencephalogram signal is preprocessed before feature extraction. As the sampling frequency of the electroencephalogram signal is 1000Hz, the electroencephalogram signal is firstly down-sampled to 200 Hz. And filtering the signal by using a Chebyshev filter at 0.5-45 Hz.
Firstly, feature extraction
The DSP algorithm belongs to a spatial filter with the aim of maximizing the feature difference between different modes. The method is an extension of Fisher linear discriminant analysis idea in space analysis, and obtains a space filter (such as formula (1)) under the condition of emphasizing maximum inter-class dispersion projection and minimum intra-class dispersion projection, wherein X in the formula isiAnd YiRepresenting training data for both types of tasks. The method is suitable for detecting the amplitude difference of low-frequency slow potential such as MRCPs.
Figure BDA0001592438270000051
Wherein the content of the first and second substances,
Figure BDA0001592438270000052
is the average value of the first type task samples,
Figure BDA0001592438270000053
The mean value of the second type task sample and Cov are covariance; omega is a spatial filter; n is the number of samples; t is matrix transposition.
The basic principle of the CCA algorithm is: in order to grasp the correlation between the two sets of indexes as a whole, two representative comprehensive variables U and V (each being a linear combination of each variable in the two variable sets) are extracted from the two sets of variables, respectively, and the overall correlation between the two sets of indexes is reflected by the correlation between the two comprehensive variables.
In the embodiment of the present invention, the test data and the two types of training data are filtered by DSP to be TestData, Template1 and Template2 in FIG. 4. The test data and the two types of data are respectively constructed into CCA filters and are respectively filtered (indicated in a dashed box in FIG. 4).
In the template matching process, pearson correlation coefficients (R11, R12, R21, R22) are calculated to represent the similarity between test data and training data, the coefficients being feature vectors.
Second, pattern recognition
LDA is often used for classification recognition of brain electrical signals, and the general form of the model is:
f(x)＝wTx+b (2)
wherein x is [ x ]1,x2,...,xn]TIs n-dimensional feature vector, w is weight vector or parameter variable, f (x) is decision value, and b is constant term. And training the LDA model by using the characteristics extracted from the training samples, and determining the values of w and b. For the second classification, the characteristics of the test sample are input, f (x) is calculated, and the test sample belongs to the class according to the comparison result of the f (x) and the threshold value. Normally, the threshold is set to 0, and the threshold may also be selected according to the probability density distribution curve of the decision value of the training sample to obtain a better classification result.
In summary, the method for detecting the exercise intention based on the dynamic stopping strategy and the integrated learning designed in the embodiment of the invention can be used for assisting the output of the disabled and the special people and performing information interaction with the outside.
In addition, the decoding method provided by the embodiment of the invention is further researched to obtain a perfect brain-computer interface system based on movement intention detection, and is expected to obtain considerable social benefit and economic benefit in the fields of electronic entertainment, industrial control and the like.
In the embodiment of the present invention, except for the specific description of the model of each device, the model of other devices is not limited, as long as the device can perform the above functions.
Those skilled in the art will appreciate that the drawings are only schematic illustrations of preferred embodiments, and the above-described embodiments of the present invention are merely provided for description and do not represent the merits of the embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (5)

1. A method for detecting motor intent based on dynamic stopping strategy combined with ensemble learning, the method comprising the steps of:
firstly, an off-line modeling stage:
collecting a sample of 200 times of autonomous key pressing of a subject, and randomly intercepting data with the time interval between a data center before key pressing and the key pressing time not less than 150ms as resting state data and data with the time interval between the data center before key pressing and the key pressing time not more than 150ms as first pre-movement state data by taking the key pressing time as 0 time and 1s as window width;
after extracting features of the resting state data and the first pre-motion state data, establishing a linear discriminant analysis model for identifying whether a current user is about to move or not;
intercepting data of-500 ms-500 s as motion state data, and intercepting data at a certain moment before and after the key as second pre-motion state data;
after extracting the characteristics of the second pre-motion state data, respectively constructing sub-classifiers with the motion state data characteristics for detecting the accurate moment of motion;
II, an online testing stage:
and after extracting the features from the real-time intercepted data, sending the data into an LDA model, outputting a decision value, if the decision value is greater than a set threshold value, sending the data into a constructed sub-classifier, outputting a result if the obtained vote is greater than the set threshold value by using a voting strategy, and otherwise, waiting for the input of the next data feature.
2. The method of claim 1, further comprising:
collecting electroencephalograms by using an electrode cap and an electroencephalogram amplifier produced by Neuroscan, and collecting 18 channel electroencephalograms at positions corresponding to the scalp of a functional exercise area by taking the top of the head as a reference and the forehead leaf as the ground;
the method is characterized in that the lead is placed according to the 10-20 international standard lead position, the sampling frequency is 1000Hz, and a 50Hz wave trap is adopted to remove power frequency interference.
3. The method for detecting exercise intention based on dynamic stopping strategy combined with ensemble learning according to claim 1, wherein the intercepting data at a certain time before and after the key is second pre-exercise state data specifically comprises:
the time difference between the data center before and after the key pressing and the time of the key pressing is-140 ms, -130ms, -120ms, -110ms, -100ms, -90ms, -80ms, -70ms, -60ms, -50ms, -40ms, -30ms, -20ms, -10ms, 10ms and 20ms respectively, and the data are second pre-motion state data.
4. A method for detecting motor intention based on dynamic stopping strategy combined with ensemble learning according to any of claims 1-3,
the method sequentially adopts discrimination space mode filtering, typical correlation analysis and template matching feature identification methods.
5. The method for detecting motion intention based on dynamic stop strategy combined with ensemble learning according to claim 2,
and (3) down-sampling the electroencephalogram signal to 200Hz, and filtering the signal by using a Chebyshev filter at 0.5-45 Hz.
CN201810193673.1A 2018-03-09 2018-03-09 Method for detecting movement intention based on combination of dynamic stopping strategy and integrated learning Active CN108764008B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810193673.1A CN108764008B (en) 2018-03-09 2018-03-09 Method for detecting movement intention based on combination of dynamic stopping strategy and integrated learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810193673.1A CN108764008B (en) 2018-03-09 2018-03-09 Method for detecting movement intention based on combination of dynamic stopping strategy and integrated learning

Publications (2)

Publication Number Publication Date
CN108764008A CN108764008A (en) 2018-11-06
CN108764008B true CN108764008B (en) 2021-10-26

Family

ID=63980171

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810193673.1A Active CN108764008B (en) 2018-03-09 2018-03-09 Method for detecting movement intention based on combination of dynamic stopping strategy and integrated learning

Country Status (1)

Country Link
CN (1) CN108764008B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112545536B (en) * 2020-12-02 2023-07-04 复旦大学附属华山医院 Action auxiliary device based on brain plasticity and control method and circuit thereof

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103610447A (en) * 2013-12-04 2014-03-05 天津大学 Mental workload online detection method based on forehead electroencephalogram signals
CN104606030A (en) * 2015-01-06 2015-05-13 上海交通大学 Lower limb on-line walking rehabilitation system and method fused with autokinetic movement consciousness

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120328148A1 (en) * 2011-06-27 2012-12-27 Grey William Doherty Method and system for secure image management

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103610447A (en) * 2013-12-04 2014-03-05 天津大学 Mental workload online detection method based on forehead electroencephalogram signals
CN104606030A (en) * 2015-01-06 2015-05-13 上海交通大学 Lower limb on-line walking rehabilitation system and method fused with autokinetic movement consciousness

Also Published As

Publication number Publication date
CN108764008A (en) 2018-11-06

Similar Documents

Publication Publication Date Title
CN103793058B (en) A kind of active brain-computer interactive system Mental imagery classification of task method and device
Kumari et al. Brainwave based user identification system: A pilot study in robotics environment
Fabiani et al. Conversion of EEG activity into cursor movement by a brain-computer interface (BCI)
CN105956624B (en) Mental imagery brain electricity classification method based on empty time-frequency optimization feature rarefaction representation
CN109497990B (en) Electrocardiosignal identity recognition method and system based on canonical correlation analysis
CN103425249A (en) Electroencephalogram signal classifying and recognizing method based on regularized CSP and regularized SRC and electroencephalogram signal remote control system
CN109512442A (en) A kind of EEG fatigue state classification method based on LightGBM
Alyasseri et al. The effects of EEG feature extraction using multi-wavelet decomposition for mental tasks classification
CN107656612B (en) Large instruction set brain-computer interface method based on P300-SSVEP
Gao et al. Multi-ganglion ANN based feature learning with application to P300-BCI signal classification
CN107411738A (en) A kind of mood based on resting electroencephalogramidentification similitude is across individual discrimination method
AU2013100576A4 (en) Human Identification with Electroencephalogram (EEG) for the Future Network Security
CN101703405A (en) Wavelet analysis and Welch algorithm personal identification method
Jiang et al. Analytical comparison of two emotion classification models based on convolutional neural networks
Fang et al. Recent advances of P300 speller paradigms and algorithms
CN108764008B (en) Method for detecting movement intention based on combination of dynamic stopping strategy and integrated learning
CN110472595A (en) Identification model construction method, device and the recognition methods of EEG signals, device
Milan et al. Adaptive brain interfaces for physically-disabled people
CN114557708A (en) Device and method for detecting somatosensory stimulation consciousness based on electroencephalogram dual-feature fusion
CN112869743B (en) Exercise initiation intention neural analysis method considering cognitive distraction
CN108491792B (en) Office scene human-computer interaction behavior recognition method based on electro-oculogram signals
CN109214325B (en) Motion-related potential detection method based on spatial filtering and template matching
CN109814720A (en) A kind of brain control method and system of equipment
CN113017648B (en) Electroencephalogram signal identification method and system
Syal et al. Comparative analysis of KNN, SVM, DT for EOG based human computer interface

Legal Events

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