CN109948640A - Electromyographic signal classification method based on two-parameter core Optimization-type extreme learning machine - Google Patents
Electromyographic signal classification method based on two-parameter core Optimization-type extreme learning machine Download PDFInfo
- Publication number
- CN109948640A CN109948640A CN201811603034.4A CN201811603034A CN109948640A CN 109948640 A CN109948640 A CN 109948640A CN 201811603034 A CN201811603034 A CN 201811603034A CN 109948640 A CN109948640 A CN 109948640A
- Authority
- CN
- China
- Prior art keywords
- learning machine
- extreme learning
- output
- parameter
- hidden layer
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of myoelectricity recognition methods based on two-parameter core Optimization-type extreme learning machine, the present invention is extracted 4 tunnel electromyography signals first and is extracted corresponding average amplitude, variance, Wilson's amplitude, wavelet energy coefficient, then these features are merged, fused feature is finally transported to two-parameter Optimization-type extreme learning machine.Two-parameter Optimization-type extreme learning machine is on the basis of extreme learning machine, introduce gaussian kernel function, by the minimum to output weight matrix, optimization parameters are set, the problem of constructing neural network structure, and extreme learning machine is minimized output error is changed into the problem of minimum output weight.This method has the Function approximation capabilities more more powerful than conventional limit learning machine, while the ability for handling Nonlinear Classification is also stronger, also has higher accuracy rate and less operation time compared to other common classifier algorithms.
Description
Technical field
The invention belongs to area of pattern recognition, are related to a kind of myoelectricity identification based on two-parameter core Optimization-type extreme learning machine
Method.
Background technique
Mode identification method is typically divided into six classes: statistical recognition methods, syntactic structure method of identification, Fuzzy Recognition, artificial
Neural network recognization method, template matching method of identification and support vector machines method of identification [41].Wherein more has statistics, mould
The recognition methods of paste, neural network and support vector machines.
For single linear discriminant analysis (LDA), this method is although accurate and quick, but for multi input and multi output
For system, become complicated using linear discriminant.In order to solve this problem, many scholars introduce " kernel function " skill, will
Kernel function and linear identification are combined research.Most in conjunction with kernel function is support vector machines mode (Support
Vector Machine, SVM).Kakoty et al. grasps class using the linear kernel SVM with wavelet transform to classify six kinds
Type, discrimination are 84 ± 2.4 %.SVM has a high accuracy of identification, excellent mathematics adaptability, intuitive geometric interpretation, with
And the advantages that being avoided that overfitting, it is more advanced classification method.Fuzzy neural network (Fuzzy Neutral Network,
FNN) be fuzzy diagnosis mode and artificial neural network identification method perfect combination.Such as Multistage fuzzy min-max classifier,
Mainly overlapping region problem is handled using multi-level tree structure.In conjunction with other classification methods such as fuzzy C-mean algorithm of fuzzy diagnosis
(Fuzzy C-Means, FCM) etc. can divide the clustering procedure of classifying rules automatically, reach lower limb movement general classification accuracy
94%~99%.
Extreme learning machine (Extreme Learning Machine, ELM) is a kind of improved list that Huang et al. is proposed
Hidden layer feedforward neural network (Single-hidden Layer Feedforward Neural, SLFN) learning algorithm.ELM mould
Type is a kind of emerging learning art, and the modeling provides the solution sides, global optimum of a Fast Learning and good Generalization Capability
Case.The model is easy to use, it is only necessary to the quantity of hidden layer neuron be arranged.
Summary of the invention
In view of the deficiencies of the prior art, the present invention proposes a kind of myoelectricities based on two-parameter core Optimization-type extreme learning machine
Recognition methods.
The present invention is extracted 4 tunnel electromyography signals first and is extracted corresponding average amplitude, variance, and Wilson's amplitude is small
Then wave energy coefficient of discharge merges these features, fused feature is finally transported to the two-parameter Optimization-type limit
Habit machine (Double Parameter Kernel Optimizing Method based on Extreme Learning
Machine, DPK-OMELM).Two-parameter Optimization-type extreme learning machine introduces Gaussian kernel letter on the basis of extreme learning machine
Number is arranged optimization parameters by the minimum to output weight matrix, constructs neural network structure, and the limit is learnt
The problem of machine minimum output error, is changed into the problem of minimum output weight.This method has more than conventional limit learning machine
For powerful Function approximation capabilities, while the ability for handling Nonlinear Classification is also stronger, compared to other common classifier algorithms
Also there are higher accuracy rate and less operation time.
In order to achieve the goal above, the method for the present invention mainly comprises the steps that
Step 1 acquires human body gastrocnemius, tibialis anterior, vastus medials, musculus vastus lateralis totally 4 when human body does daily behavior movement
Road electromyography signal and plantar pressure；Daily behavior movement includes that static conversion acts, gait movement, tumble movement, gait movement
It walks, go upstairs, go downstairs, run including level land；Static conversion movement include station-seat, seat-stand, stand-crouching, crouching-are stood, are sat-lie,
It lies-sits；Tumble movement include level land walk, upstairs, downstairs, race.
Step 2 extracts the average amplitude of 4 tunnel electromyography signals, variance, Wilson's amplitude, wavelet energy coefficient respectively；
Step 3 constructs two-parameter core optimization extreme learning machine classifier, and the feature of extraction is input to the classifier
In classify；
The two-parameter core optimization extreme learning machine classifier building is as follows:
(1) Architecture of Feed-forward Neural Network for determining extreme learning machine is made of input layer, hidden layer and output layer.Each
Neuron is linked using the weighing vector of w.Other parameters include deviation b, provide additional adjustable parameter and the input of model
Parameter x.Therefore, network can be described with triple f=(w, b, x).
The input/output relation of extreme learning machine can be expressed from the next:
Matrix form:
Wherein xi=[xi1,xi2,...,xin]T∈RnIt is n dimension input, yi=[yi1,yi2,...,yim]T∈RmIt is that m dimension is defeated
Out, bi=[bi1,bi2,...,bik]T∈RkIt is hidden layer vector, wi=[ωi1,ωi2,...,ωin] it is j-th of hidden layer mind
The weight vector being connect through member with input neuron；βi=[βi1,βi2,...,βin]TIt is j-th of hidden layer neuron and output nerve
The weight vector of member connection；K is the number of hidden layer neuron, bjIt is the deviation of j-th of hidden layer neuron, ωj·xiIt is ωj
And xiInner product, f is a function that infinitely can be micro-, can be using it as the activation primitive of hidden layer neuron.H is hidden layer
Output matrix, Y is output matrix, and β is output weight coefficient matrix.
(2) it is randomly provided the inclined of the input layer of feedforward neural network and the connection weight ω of hidden layer and hidden layer neuron
Poor b, and calculate the output matrix H of hidden layer；
(3) keep output error minimum by following formula:
Wherein, foIt (x) is target output function.Guarantee the performance of itself by the minimum to output weighting function β.
(4) the problem of extreme learning machine being minimized output error is changed into the problem of minimum output weight:
(5) solve output weight β least squares norm solution, output weight matrix β can by by input weight matrix with
The least square method of deviation calculates gained:
Wherein H+For the generalized inverse matrix of state matrix, canonical constant 1/C can improve the stability and generalization ability of result.
(6) gaussian kernel function K (x, x are introducedi), since Gaussian kernel extreme learning machine is to the regular parameter C in weight matrix
There is selection to require with kernel function center width parameter σ, then
By σ and C to weight minimum value must be exported to optimize ELM, referred to as two-parameter core Optimization-type extreme learning machine:
The fall detection method based on electromyography signal that the present invention designs, has the advantages that
ELM substantially increases the performance of classifier, relatively as a kind of learning method that is quick, possessing stronger generalization
In traditional classification method, ELM greatly reduces training and spends the time on the basis of guaranteeing certain recognition accuracy.ELM phase
Faster and parameter is insensitive to SVM speed, it is easier to dispose.Kernel function is a kind of mapping function, and can make can not linear separation
Eigenvector projection to more implicit and higher-dimension a space, and the coordinate without calculating data in new space will be away from
Simple kernel inner product is substituted for from calculating.It is much simpler that this single stepping usually carries out distance calculating to coordinate than directly.It is double
Parameter core, which optimizes extreme learning machine, has the Function approximation capabilities more more powerful than conventional limit learning machine, while handling non-linear
The ability of classification is also stronger.
Detailed description of the invention
Fig. 1 is implementation flow chart of the invention；
Fig. 2 is ELM complete structure figure；
Fig. 3 is that static conversion respectively acts the sensitivity under three classifiers, specificity and accuracy rate；
Fig. 4 is each tumble type sensitivity under three classifiers, specificity and accuracy rate.
Specific embodiment
As shown in Figure 1, the present embodiment includes the following steps:
Step 1 acquires human body gastrocnemius, tibialis anterior, vastus medials, musculus vastus lateralis totally 4 when human body does daily behavior movement
Road electromyography signal and plantar pressure；
Step 2 extracts the VAR of 4 tunnel electromyography signals, WAMP, EWT, MA and plantar pressure AR coefficient respectively；
Step 3, by combining the broad sense canonical correlation method (WGA-GCCA) of genetic algorithm to 5 features of step 2
Carry out Fusion Features；
Step 4 determines the Architecture of Feed-forward Neural Network of extreme learning machine, and determines the number of hidden layer neuron；
Step 5 is randomly provided the input layer of feedforward neural network and the connection weight ω and hidden layer neuron of hidden layer
Deviation b, and calculate the output matrix H of hidden layer；
Step 6 solves the least squares norm solution of output weight β, and output weight matrix β can be by by input weight square
The least square method of battle array and deviation calculates gained:
Step 7, in conjunction with gaussian kernel function, by σ and C to weight minimum value must be exported to optimize ELM:
As shown in Figure 2；
The fused feature vector of step 3 is input in two-parameter core Optimization-type extreme learning machine and is divided by step 8
Class identification.
Further to verify the reliable classification results of institute's elevator electricity classifier of the present invention, the support of Gaussian kernel is introduced herein
Vector machine (Gaussian Kernel Support Vector Machine, GK-SVM) classifier, and be all neural network
Minimax fuzzy neural network (Fuzzy Min-Max Neural Network, FMMNN) classifier.It successively will be each quiet
State switching motion is used as negative sample, obtains respective sensitive under three classifiers as positive sample, the movement of remaining static conversion
Degree, specificity and accuracy rate.
1 static conversion of table respectively acts the sensitivity at DPK-OMELM, GK-SVM, tri- classifiers of FMMNN, specificity and
Accuracy rate (%)
Statistics indicate that DPK-OMELM classifier algorithm proposed in this paper is better than SVM and FMMNN in table 1.It can from Fig. 3
Out, the phase reciprocal process sat and lain being acted, recognition effect is best, this is because the vola overall pressure ratio after lying is 0,
If Fusion Features weighting is correct, discrimination all can be higher.Remaining switching motion has certain similitude, especially for
Stand-crouching, crouching-stand this group of reversible process, the recognition accuracy of GK-SVM is small less than the recognition accuracy of 95%, FMMNN classifier
In 90%, but this paper DPK-OMELM classifier can also reach 97% or so.
Table 2 is successively using every kind of tumble type as positive sample, remaining tumble Type division is a kind of as negative sample, is obtained
Respective sensitivity, specificity and accuracy rate under DPK-OMELM and SVM and FMMNN classifier out.
The sensitivity at DPK-OMELM, tri- classifiers of GK-SVM, FMMNN of each tumble type of table 2, it is specific and accurate
Rate (%)
Fig. 4 shows that discrimination of the identification classification of tumble classification generally than normal ADLs is low.Relatively high tumble class
Be not downstairs-fall, in part because of downstairs fall tumble mode it is different from other several modes fallen forward.
FMMNN is poor to the mode recognition effect of several tumbles, and context of methods is still the optimal sorting algorithm of effect.
The testing result of the trifle is larger to the practical significance of Prevention of fall and safeguard, thus the time detected is direct
Affect the implementation of safeguard procedures.Tumble process be generally divided into collision before, collision and collision after.From fall to
That time before collision on the ground becomes collision early period, to safeguard procedures of tumble, such as the inflation measure of air bag etc.
It all needs during this period of time to complete, thus, detection time is shorter, and the time for implementing safeguard procedures is well-to-do.Table 3 has chosen ten times
Identification process simultaneously has recorded each DPK-OMELM, the operation time of GK-SVM, FMMNN, it can be found that the fortune of these three classifiers
Scanning frequency degree is not on one level.Fast 1 order of magnitude of operation ratio GK-SVM of DPK-OMELM, faster than FMMNN 2 quantity
Grade.
Runing time (× 10 of the 3 10 groups of tests of table at DPK-OMELM, tri- classifiers of GK-SVM, FMMNN2s)
Claims (2)
1. the electromyographic signal classification method based on two-parameter core Optimization-type extreme learning machine, which is characterized in that this method includes such as
Lower step:
Step 1 acquires human body gastrocnemius, tibialis anterior, vastus medials, musculus vastus lateralis totally 4 road flesh when human body does daily behavior movement
Electric signal and plantar pressure；
Step 2 extracts the average amplitude of 4 tunnel electromyography signals, variance, Wilson's amplitude and wavelet energy coefficient respectively；
Step 3, constructs two-parameter core optimization extreme learning machine classifier, and the feature of extraction be input in the classifier into
Row classification；
The two-parameter core optimization extreme learning machine classifier building is as follows:
(1) Architecture of Feed-forward Neural Network for determining extreme learning machine is made of input layer, hidden layer and output layer；Each nerve
Member is linked using the weighing vector of w；Other parameters include deviation b, provide the additional adjustable parameter and input parameter of model
x；Therefore, network is described with triple f=(w, b, x)；
The input/output relation of extreme learning machine is expressed from the next:
Matrix form: Y=H β
Wherein xi=[xi1,xi2,...,xin]T∈RnIt is n dimension input, yi=[yi1,yi2,...,yim]T∈RmIt is m dimension output, bi=
[bi1,bi2,...,bik]T∈RkIt is hidden layer vector, wi=[ωi1,ωi2,...,ωin] be j-th of hidden layer neuron with it is defeated
Enter the weight vector of neuron connection；βi=[βi1,βi2,...,βin]TIt is that j-th of hidden layer neuron is connect with output neuron
Weight vector；K is the number of hidden layer neuron, bjIt is the deviation of j-th of hidden layer neuron, ωj·xiIt is ωjAnd xiIt is interior
Product, f is a function that infinitely can be micro-, using it as the activation primitive of hidden layer neuron；H is the output matrix of hidden layer, Y
For output matrix, β is output weight coefficient matrix；
(2) it is randomly provided the deviation b of the input layer of feedforward neural network and the connection weight ω of hidden layer and hidden layer neuron,
And calculate the output matrix H of hidden layer；
(3) keep output error minimum by following formula:
Wherein, foIt (x) is target output function；Guarantee the performance of itself by the minimum to output weighting function β；
(4) the problem of extreme learning machine being minimized output error is changed into the problem of minimum output weight:
(5) the least squares norm solution of output weight β is solved, output weight matrix β passes through by input weight matrix and deviation
Least square method calculates gained:
Wherein H+For the generalized inverse matrix of state matrix, canonical constant 1/C can improve the stability and generalization ability of result；
(6) gaussian kernel function K (x, x are introducedi), since Gaussian kernel extreme learning machine is to the regular parameter C and core letter in weight matrix
Number center width parameter σ has selection to require, then
By σ and C to weight minimum value must be exported to optimize ELM, referred to as two-parameter core Optimization-type extreme learning machine:
2. the electromyographic signal classification method according to claim 1 based on two-parameter core Optimization-type extreme learning machine, special
Sign is: daily behavior movement includes that static conversion acts, gait movement, tumble movement, and gait movement includes that level land is walked, upstairs
Ladder goes downstairs, runs；Static conversion movement include station-seat, seat-stand, stand-crouching, crouching-are stood, are sat-lie, lie-sit；Tumble movement packet
Include level land walk, upstairs, downstairs, run.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811603034.4A CN109948640A (en) | 2018-12-26 | 2018-12-26 | Electromyographic signal classification method based on two-parameter core Optimization-type extreme learning machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811603034.4A CN109948640A (en) | 2018-12-26 | 2018-12-26 | Electromyographic signal classification method based on two-parameter core Optimization-type extreme learning machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109948640A true CN109948640A (en) | 2019-06-28 |
Family
ID=67007270
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811603034.4A Pending CN109948640A (en) | 2018-12-26 | 2018-12-26 | Electromyographic signal classification method based on two-parameter core Optimization-type extreme learning machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109948640A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110738093A (en) * | 2019-08-16 | 2020-01-31 | 杭州电子科技大学 | Classification method based on improved small world echo state network electromyography |
| CN111789592A (en) * | 2020-06-30 | 2020-10-20 | 杭州电子科技大学 | Electroencephalogram recognition method based on topological feature fusion |
| CN111796519A (en) * | 2020-06-14 | 2020-10-20 | 武汉理工大学 | Automatic control method of multi-input multi-output system based on extreme learning machine |
| CN112766140A (en) * | 2021-01-15 | 2021-05-07 | 云南电网有限责任公司电力科学研究院 | Transformer fault identification method based on kernel function extreme learning machine |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120184838A1 (en) * | 2009-07-30 | 2012-07-19 | South African Medical Research Council | Non-invasive deep muscle electromyography |
| CN102622605A (en) * | 2012-02-17 | 2012-08-01 | 国电科学技术研究院 | Surface electromyogram signal feature extraction and action pattern recognition method |
| CN103984962A (en) * | 2014-05-30 | 2014-08-13 | 河北工业大学 | Exoskeleton walking mode identification method based on electromyographic signals |
| CN104127181A (en) * | 2014-07-22 | 2014-11-05 | 杭州电子科技大学 | Electromyographic signal tumble detection method based on WKFDA |
| CN104382595A (en) * | 2014-10-27 | 2015-03-04 | 燕山大学 | Upper limb rehabilitation system and method based on myoelectric signal and virtual reality interaction technology |
| CN104680236A (en) * | 2015-02-13 | 2015-06-03 | 西安交通大学 | FPGA implementation method of kernel function extreme learning machine classifier |
| CN105740619A (en) * | 2016-01-28 | 2016-07-06 | 华南理工大学 | On-line fault diagnosis method of weighted extreme learning machine sewage treatment on the basis of kernel function |
| CN108267429A (en) * | 2017-11-30 | 2018-07-10 | 南京信息工程大学 | The quantitative analysis method of full iron and basicity in a kind of sinter |
| CN108776783A (en) * | 2018-05-31 | 2018-11-09 | 陕西师范大学 | The gesture electromyography signal recognition methods of extreme learning machine-Hidden Markov Model |
-
2018
- 2018-12-26 CN CN201811603034.4A patent/CN109948640A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120184838A1 (en) * | 2009-07-30 | 2012-07-19 | South African Medical Research Council | Non-invasive deep muscle electromyography |
| CN102622605A (en) * | 2012-02-17 | 2012-08-01 | 国电科学技术研究院 | Surface electromyogram signal feature extraction and action pattern recognition method |
| CN103984962A (en) * | 2014-05-30 | 2014-08-13 | 河北工业大学 | Exoskeleton walking mode identification method based on electromyographic signals |
| CN104127181A (en) * | 2014-07-22 | 2014-11-05 | 杭州电子科技大学 | Electromyographic signal tumble detection method based on WKFDA |
| CN104382595A (en) * | 2014-10-27 | 2015-03-04 | 燕山大学 | Upper limb rehabilitation system and method based on myoelectric signal and virtual reality interaction technology |
| CN104680236A (en) * | 2015-02-13 | 2015-06-03 | 西安交通大学 | FPGA implementation method of kernel function extreme learning machine classifier |
| CN105740619A (en) * | 2016-01-28 | 2016-07-06 | 华南理工大学 | On-line fault diagnosis method of weighted extreme learning machine sewage treatment on the basis of kernel function |
| CN108267429A (en) * | 2017-11-30 | 2018-07-10 | 南京信息工程大学 | The quantitative analysis method of full iron and basicity in a kind of sinter |
| CN108776783A (en) * | 2018-05-31 | 2018-11-09 | 陕西师范大学 | The gesture electromyography signal recognition methods of extreme learning machine-Hidden Markov Model |
Non-Patent Citations (1)
| Title |
|---|
| XUGANG XI ET AL: "Evaluation of Feature Extraction and Recognition for Activity Monitoring and Fall Detection Based on Wearable sEMG Sensors", 《SENSORS》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110738093A (en) * | 2019-08-16 | 2020-01-31 | 杭州电子科技大学 | Classification method based on improved small world echo state network electromyography |
| CN110738093B (en) * | 2019-08-16 | 2022-07-01 | 杭州电子科技大学 | Classification method based on improved small world echo state network electromyography |
| CN111796519A (en) * | 2020-06-14 | 2020-10-20 | 武汉理工大学 | Automatic control method of multi-input multi-output system based on extreme learning machine |
| CN111796519B (en) * | 2020-06-14 | 2022-05-06 | 武汉理工大学 | Automatic control method of multi-input multi-output system based on extreme learning machine |
| CN111789592A (en) * | 2020-06-30 | 2020-10-20 | 杭州电子科技大学 | Electroencephalogram recognition method based on topological feature fusion |
| CN111789592B (en) * | 2020-06-30 | 2022-04-05 | 杭州电子科技大学 | Electroencephalogram recognition method based on topological feature fusion |
| CN112766140A (en) * | 2021-01-15 | 2021-05-07 | 云南电网有限责任公司电力科学研究院 | Transformer fault identification method based on kernel function extreme learning machine |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109948640A (en) | Electromyographic signal classification method based on two-parameter core Optimization-type extreme learning machine | |
| CN106599797B (en) | A kind of infrared face recognition method based on local parallel neural network | |
| Cristin et al. | Deep neural network based Rider-Cuckoo Search Algorithm for plant disease detection | |
| Gutta et al. | Gender and ethnic classification of face images | |
| Ghyar et al. | Computer vision based approach to detect rice leaf diseases using texture and color descriptors | |
| CN104598890B (en) | A kind of Human bodys' response method based on RGB D videos | |
| CN108256488A (en) | A kind of radar target identification method based on micro-Doppler feature extraction and deep learning | |
| CN104523264B (en) | Electrocardiosignal processing method | |
| CN107657279A (en) | A kind of remote sensing target detection method based on a small amount of sample | |
| CN107316294A (en) | One kind is based on improved depth Boltzmann machine Lung neoplasm feature extraction and good pernicious sorting technique | |
| CN106529504B (en) | A kind of bimodal video feeling recognition methods of compound space-time characteristic | |
| CN103473539A (en) | Gait recognition method and device | |
| Vilasini et al. | CNN Approaches for Classification of Indian Leaf Species Using Smartphones. | |
| CN107301409B (en) | System and method for selecting Bagging learning to process electrocardiogram based on Wrapper characteristics | |
| CN106778501A (en) | Video human face ONLINE RECOGNITION method based on compression tracking with IHDR incremental learnings | |
| CN110363152B (en) | Method for identifying road condition of lower limb prosthesis based on surface electromyographic signals | |
| CN108267724A (en) | A kind of unknown object recognition methods of radar target recognition | |
| CN104268507A (en) | Manual alphabet identification method based on RGB-D image | |
| Gutta et al. | Gender and ethnic classification of human faces using hybrid classifiers | |
| Mao et al. | Discriminant analysis neural networks | |
| CN109886113A (en) | A kind of spacious view pedestrian recognition methods again based on region candidate network | |
| CN106529544A (en) | Fabric flatness objective evaluation method and fabric flatness objective evaluation device based on unsupervised machine learning | |
| Sabzi et al. | The use of soft computing to classification of some weeds based on video processing | |
| CN108509861A (en) | A kind of method for tracking target and device combined based on sample learning and target detection | |
| Wallace et al. | Intelligent initialization of resource allocating RBF networks |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190628 |
|
| RJ01 | Rejection of invention patent application after publication |