CN110222634A - A kind of human posture recognition method based on convolutional neural networks - Google Patents
A kind of human posture recognition method based on convolutional neural networks Download PDFInfo
- Publication number
- CN110222634A CN110222634A CN201910481323.XA CN201910481323A CN110222634A CN 110222634 A CN110222634 A CN 110222634A CN 201910481323 A CN201910481323 A CN 201910481323A CN 110222634 A CN110222634 A CN 110222634A
- Authority
- CN
- China
- Prior art keywords
- neural networks
- convolutional neural
- human body
- layer
- convolutional
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/23—Recognition of whole body movements, e.g. for sport training
Abstract
The invention discloses a kind of human posture recognition methods based on convolutional neural networks, first acquisition human body attitude data set, and carry out the pretreatment that video is cut into picture frame to it；Then convolutional neural networks model is built, by introducing sparsity in RELU excitation function input, reduces the unnecessary input of excitation function；Then network is trained by iteration undated parameter in conjunction with sparse item optimization conventional target loss function, to obtain optimal solution；Finally, identifying according to the resulting network model of training to human body attitude, human body attitude classification is exported.The beneficial effects of the present invention are: the method that the present invention uses can accelerate convergence rate, improve the generalization ability of network while keeping compared with lofty stance discrimination.
Description
Technical field
The invention belongs to technical field of computer vision, and in particular to a kind of human body attitude knowledge based on convolutional neural networks
Other method.
Background technique
In recent years, universal with the development of information technology and artificial intelligence technology, human body attitude identification technology starts extensively
General application.Related researcher attempts to explore validity feature with the human posture's data set being collected into and classify.Tradition
Gesture recognition method mainly there are two step: (1) complicated manual features are extracted from original input picture；(2) from acquisition
Feature in training classifier.
During traditional gesture recognition, need to extract complicated manual features from original input picture.Although
It is effective on accuracy of identification, but due to the high complexity of human body, the spy extracted from bone key point and deep image information
It is often relatively high to levy size.Most of depth image is pre-processed, and causes feature extraction difficult, and recognition efficiency is low,
Convergence time is long.Conventional method is not best practice.
However, the more traditional part joint modeling method of artificial neural network is compared, there is Nonlinear Modeling and adaptive
Ability can excavate image profound level information, Lai Jinhang gesture recognition by a large amount of still images of training；Neural network method pair
The local feature of objects in images is characterized, and classification has more robustness.In addition, reducing and swashing by introducing sparse regularization
It encourages function and inputs unnecessary increase, reduce the complexity of model, improve the generalization ability of convolutional neural networks, thus guaranteeing
The convergence rate of model is improved while higher discrimination.
Summary of the invention
In view of the shortcomings of the prior art, the purpose of the present invention is to provide a kind of people based on convolutional neural networks
Body gesture recognition method is guaranteeing higher discrimination by designing a kind of convolutional neural networks model for introducing sparse regularization
While, convergence rate is improved, the generalization ability of model is enhanced.
To achieve the goals above, the present invention is to realize by the following technical solutions:
In the training stage, construct the self study complex network structures model of data-driven, by original human body pose presentation and
Corresponding posture tag block outputting and inputting as network carries out the study for having supervision tutor's mode to network.In verifying rank
Section gives unknown input original image, carries out gesture recognition.
A kind of human posture recognition method based on convolutional neural networks, comprising the following steps: S01 obtains human body attitude
Sets of video data carries out the pretreatment that video is cut into picture frame to it, and the image data set for being cut into picture frame is divided into
Training set and verifying collect；
S02, construct neural network model, RELU excitation function input introduce sparsity, convolutional neural networks it is defeated
Enter for pretreated image in the step S01, exports as human body attitude classification；The convolutional neural networks are instructed
Practice；
S03 identifies human body attitude using the neural network model in the S02, in disclosed human body attitude number
According to the test for carrying out model training and performance on collection KTH；When there is unknown video input, the step S01 is called to carry out first
Then pretreatment carries out gesture recognition using the neural network model in the step S02, obtain human body attitude classification.
A kind of above-mentioned human posture recognition method based on convolutional neural networks obtains human body in the step S01
Posture sets of video data specifically includes the following steps:
S11: disclosed KTH human body attitude sets of video data is obtained；
S12: video is cut into framing, and saves the image of every frame；
S13: it is filtered out from image and complete human body occurs and the image of the human body behavior of making outgoing label respective action, deleted
Blank does not occur complete human body or does not make the image of the human body behavior of outgoing label respective action, and to the image after screening into
Row classification marker；It includes boxing, waving, applauding, jogging, running, being careful that the label, which corresponds to posture,；The classification marker is specially
According to box, wave, applaud, jog, run, this 6 class posture of being careful is marked；
S14: extracting prospect using gauss hybrid models, that is, the human body moved；
S15: the image after screening in the step S13 is normalized；
S16: the image set after normalized in the step S15 is randomly divided into training set in the ratio of 8:2 and is tested
Card collection.
A kind of above-mentioned human posture recognition method based on convolutional neural networks, in the step S02, the convolution
Neural network includes 4 convolutional layers, 4 pond layers, 2 full articulamentums and 1 classification layer；Convolution kernel in the first two convolutional layer
5 × 5 are dimensioned to, latter two is set as 3 × 3；All pond layers are dimensioned to 2 × 2, and pond layer is using maximum pond
Change；Convolution kernel number in first convolutional layer 32, convolution kernel number 64 in the second convolutional layer, third and fourth volume
Convolution kernel number in lamination is 128；Neuron number in first full articulamentum is set as 1024, and second connects entirely
It connects the neuronal quantity in layer and is set as 512；In convolutional layer and full articulamentum, use RELU function as activation primitive.
The characteristic information of each figure layer of pond layer main decomposition, and divide feature set.Sample size is big when initialization, calculates multiple
It is miscellaneous, but as continuous renewal, calculation amount can be reduced, local feature rises to global characteristics.For each characteristic pattern, what it was extracted
It is maximum value, relative to average pond, maximum pondization can extract more effective feature.
Full articulamentum collects every layer of characteristic information, the global characteristics of entire neural network is finally obtained, to subsequent sample
This image classification and identification play key effect.
Layer of classifying is used for final decision, is the last layer in network model.It calculates network training punishment prediction and reality
Deviation between the label of border.Soft-max function is used in classification layer, this function is used for more assorting processes.
A kind of above-mentioned human posture recognition method based on convolutional neural networks, convolutional neural networks are a kind of special
Deep neural network includes convolutional layer and pond layer.Convolutional layer and the neuron of preceding layer by part connection and value share into
Row connection, reduces the quantity of training parameter.
Convolutional layer is the core of convolutional neural networks.Low-level image feature progressive alternate is updated to high vision feature.Often
A convolutional layer is made of neuron；Assuming that n-th of node O of m layers of convolutional layerm,nInput value be { xn-1,1；
xn-1,2；…；xn-1,k, then the value of output unit is as follows:
Wherein, hw,bIt (x) is excitation value, f (x) is excitation function, xiIt is i-th of input value of node, wiIt is defeated i-th of node
Enter the weight of value, k is the number of present node input value, and b is bias term；
Linear unit R ELU is corrected to be defined as follows:
F (x)=max (0, x) (2)
The pattern field function that linear unit R ELU is convolutional layer is corrected, provides nonlinear activation energy for convolutional neural networks
Power, and other convolutional layers will not be interfered.
A kind of above-mentioned human posture recognition method based on convolutional neural networks, passes through the instruction to convolutional neural networks
Practice, obtain the parameter of convolutional neural networks, specifically:
If there is p sample set, they are expressed as { (x1,y1),(x2,y2),…,xp,yp), for each sample, each
Sample set loss function is defined as follows:
Wherein, hω,b(x) it is predicted value after network training, that is, excitation value, y are real output values, w is input value
Weight；
For whole sample set, whole sample set loss function is defined as follows:
Wherein, N is total number of plies of convolutional neural networks, skIt is -1 layer of kth of number of nodes, a is regularization coefficient, i i-th
A node, j are j-th of feature convolution kernel, and k is convolutional neural networks kth layer；For convolutional neural networks kth i-th of section of layer
Connection weight between point and j-th of feature convolution kernel；
The target of convolutional neural networks model training is to minimize the value of whole sample set loss function；
Parameter more new formula is as follows:
Wherein, L is target loss function, and μ is learning rate,X is indicated to y derivation, parameter uses small lot method more
Newly.
A kind of above-mentioned human posture recognition method based on convolutional neural networks, learning rate are set as 0.001, use
Adamoptimer algorithm optimizes.
A kind of above-mentioned human posture recognition method based on convolutional neural networks swashs in the step S02 in RELU
Function input is encouraged, i.e., at the output of linear filter, introduces sparsity, specifically:
The input h of kth layer RELU excitation function in the convolutional neural networkskDegree of rarefication indicate are as follows:
Wherein, hkIt is the input of kth layer RELU excitation function, S (hk) it is that kth layer RELU motivates letter in convolutional neural networks
Several input hkDegree of rarefication；
The objective function of optimization is defined, to determine convolutional neural networks parameter, objective function is defined as follows:
E=L+ λ ∑kS(hk) (8)
Wherein, E is the objective function of optimization, i.e., final result of the invention；L is the objective function for being not introduced into sparsity,
Namely target loss function；λ is the tuner parameters for controlling degree of rarefication.
By introducing sparsity in the input RELU, the unnecessary increase that can have not only prevented RELU from exporting, but also can reduce
The unnecessary negative input of RELU, can be improved model generalization ability.
The invention has the benefit that
(1) present invention is by designing a kind of convolutional neural networks for introducing sparsity, to identify human body attitude, to be applied to
Human-computer interaction, Activity recognition, the classification of motion, unusual checking and automatic Pilot etc. are guaranteeing the same of higher discrimination
When, convergence rate is improved, the generalization ability of model is enhanced；
(2) present invention utilizes the Artificial Neural Network mentioned in background technique and sparse regularization method is introduced,
And the advantages of combining the two, completely new convolutional neural networks model is designed, and divide human body attitude using this kind of model
Class identification.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is the flow chart of the method for the present invention；
Fig. 2 is convolutional neural networks model structure block diagram proposed by the present invention.
Specific embodiment
To be easy to understand the technical means, the creative features, the aims and the efficiencies achieved by the present invention, below with reference to
Specific embodiment, the present invention is further explained.
As described in Figure 1, human posture recognition method proposed by the present invention, comprising the following steps:
S01, obtains human body attitude video KTH data set, the data set totally 6 kinds of postures, box, wave, applauding, jogging,
It runs, be careful, the pretreatment that video is cut into picture frame is carried out to it, and the image data set is divided into training set and verifying collection；
S11: KTH human body attitude sets of video data is obtained；
S12: video is cut into framing, and saves the image of every frame；
S13: filtering out from image and complete human body occur and make the image that outgoing label corresponds to the human body behavior of posture, deletes
Blank does not occur complete human body or does not make the image that outgoing label corresponds to the human body behavior of posture, and to the image after screening into
Row classification marker；It includes box, waving, applauding, jogging, running, being careful that label, which corresponds to posture, this in 6 posture be from KTH human body appearance
State video data concentrates the posture obtained；Classification marker is specially according to boxing, wave, applaud, jog, run, this 6 class appearance of being careful
State is marked；
S14: extracting prospect using gauss hybrid models, that is, the human body moved；
S15: image is normalized；
S16: image set is randomly divided into training set in the ratio of 8:2 and verifying collects.Equal 120 of every class posture training set has
Imitate picture, verifying 30 effective pictures of collection.
S02 constructs neural network model, introduces sparsity in RELU excitation function input；Convolutional neural networks it is defeated
Enter for pretreated image, exports as human body attitude classification；Convolutional neural networks are trained.
S21: building neural network, the convolutional neural networks framework that the present invention uses is as shown in Fig. 2, the network architecture shares 7
Layer, 4 convolutional layers (including pond layer), 2 full articulamentums and 1 classification layer.The first two convolution kernel is dimensioned to 5 × 5,
Latter two is set as 3 × 3.All pond layers are dimensioned to 2 × 2, and pond layer is using maximum pond.In first convolutional layer
Convolution kernel number 32, convolution kernel number 64 in the second convolutional layer, the convolution nucleus number in third and fourth convolutional layer
Mesh is 128.Neuron number in first full articulamentum is set as 1024, the neuron number in second full articulamentum
Amount is set as 512.In convolutional layer and full articulamentum, use RELU function as activation primitive.Learning rate is set as
0.001, it is optimized using adamoptimer algorithm.
Assuming that n-th of node O of m layers of convolutional layerm,nInput value be { xn-1,1；xn-1,2；…；xn-1,k, then export list
The value of member is as follows:
Wherein, hw,bIt (x) is excitation value, f (x) is excitation function, xiIt is i-th of input value of node, wiIt is defeated i-th of node
Enter the weight of value, k is the number of present node input value, and b is bias term；
Linear unit R ELU is corrected to be defined as follows:
F (x)=max (0, x) (2)
The pattern field function that linear unit R ELU is convolutional layer is corrected, provides nonlinear activation energy for convolutional neural networks
Power, and other convolutional layers will not be interfered.
By the training to convolutional neural networks, the parameter of convolutional neural networks is obtained, specifically:
If there is p sample set, they are expressed as { (x1,y1),(x2,y2),…,(xp,yp), for each sample, often
A sample set loss function is defined as follows:
Wherein, hω,b(x) it is predicted value after network training, that is, excitation value, y are real output values, w is input value
Weight；
For whole sample set, whole sample set loss function is defined as follows:
Wherein, N is total number of plies of convolutional neural networks, skIt is -1 layer of kth of number of nodes, a is regularization coefficient, i i-th
A node, j are j-th of feature convolution kernel, and k is convolutional neural networks kth layer；For convolutional neural networks kth i-th of section of layer
Connection weight between point and j-th of feature convolution kernel；
The target of convolutional neural networks model training is to minimize the value of whole sample set loss function；
Parameter more new formula is as follows:
Wherein, L is target loss function, and μ is learning rate,X is indicated to y derivation, parameter uses small lot method more
Newly.
S22: introducing sparsity, in RELU excitation function input, i.e., at the output of linear filter, introduces sparsity,
To degree of rarefication S (hk) handled greater than 0.6 layer.By introducing sparsity in the input RELU, both RELU can be prevented defeated
Unnecessary increase out, and the unnecessary negative input of RELU can be reduced, model generalization ability can be improved.
In RELU excitation function input, i.e., at the output of linear filter, sparsity is introduced, specifically:
The input h of kth layer RELU excitation function in convolutional neural networkskDegree of rarefication indicate are as follows:
Wherein, hkIt is the input of kth layer RELU excitation function, S (hk) it is that kth layer RELU motivates letter in convolutional neural networks
Several input hkDegree of rarefication；
The objective function of optimization is defined, to determine convolutional neural networks parameter, objective function is defined as follows:
E=L+ λ ∑kS(hk) (8)
Wherein, E is the objective function of optimization, and L is the objective function for being not introduced into sparsity, that is, target loss function, λ
It is the tuner parameters for controlling degree of rarefication.
S03 identifies human body attitude using the neural network model in S02, in disclosed human body attitude data set
The test of model training and performance is carried out on KTH.When new unknown video input, S01 pretreatment is first passed around, is then passed through
Neural network forecast judges posture.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, method, article or the device that include a series of elements not only include those elements, and
And further include other elements that are not explicitly listed, or further include for this process, method, article or device institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including being somebody's turn to do
There is also other identical elements in the process, method of element, article or device.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
In embodiment provided herein, it should be understood that disclosed system and method can pass through others
Mode is realized.For example, system embodiment described above is only schematical, for example, the division of the unit, only
A kind of logical function partition, there may be another division manner in actual implementation, for example, multiple units or components can combine or
Person is desirably integrated into another system, or some features can be ignored or not executed.
Professional further appreciates that, module described in conjunction with the examples disclosed in the embodiments of the present disclosure
And algorithm steps, it can be realized with electronic hardware, computer software, or a combination of the two, professional technician can be to every
A specific application uses different methods to achieve the described function, but this realizes it is not considered that beyond of the invention
Range.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention.
Various modifications to these embodiments will be readily apparent to those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the invention
It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one
The widest scope of cause.
The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention.Industry description
Merely illustrate the principles of the invention, without departing from the spirit and scope of the present invention, the present invention also have various change and
It improves, these changes and improvements all fall within the protetion scope of the claimed invention.The claimed scope of the invention is by appended power
Sharp claim and its equivalent thereof.
Claims (7)
1. a kind of human posture recognition method based on convolutional neural networks, which comprises the following steps:
S01 obtains human body attitude sets of video data, the pretreatment that video is cut into picture frame is carried out to it, and will be cut into image
The image data set of frame is divided into training set and verifying collects；
S02 constructs neural network model, introduces sparsity in RELU excitation function input, the input of convolutional neural networks is
Pretreated image in the step S01 exports as human body attitude classification；The convolutional neural networks are trained；
S03 identifies human body attitude using the neural network model in the S02, in disclosed human body attitude data set
The test of model training and performance is carried out on KTH；When there is unknown video input, the step S01 is called to be located in advance first
Then reason carries out gesture recognition using the neural network model in the step S02, obtain human body attitude classification.
2. a kind of human posture recognition method based on convolutional neural networks according to claim 1, it is characterised in that:
In the step S01, obtain human body attitude sets of video data specifically includes the following steps:
S11: disclosed KTH human body attitude sets of video data is obtained；
S12: video is cut into framing, and saves the image of every frame；
S13: it is filtered out from image and complete human body occurs and make the image that outgoing label corresponds to the human body behavior of posture, delete blank
Or does not occur complete human body or do not make the image that outgoing label corresponds to the human body behavior of posture, and the image after screening is divided
Class label；It includes boxing, waving, applauding, jogging, running, being careful that the label, which corresponds to posture,；The classification marker be specially according to
It boxes, wave, applauding, jogging, running, this 6 class posture of being careful is marked；
S14: extracting prospect using gauss hybrid models, that is, the human body moved；
S15: the image after screening in the step S13 is normalized；
S16: the image set after normalized in the step S15 is randomly divided into training set in the ratio of 8:2 and verifying collects.
3. a kind of human posture recognition method based on convolutional neural networks according to claim 1, it is characterised in that:
In the step S02, the convolutional neural networks include 4 convolutional layers, 4 pond layers, 2 full articulamentums and 1 classification layer；
Convolution kernel is dimensioned to 5 × 5 in the first two convolutional layer, latter two is set as 3 × 3；All pond layers are dimensioned to 2
× 2, pond layer is using maximum pond；Convolution kernel number in first convolutional layer 32, the convolution nucleus number in the second convolutional layer
64, mesh, the convolution kernel number in third and fourth convolutional layer is 128；Neuron number in first full articulamentum is set
1024 are set to, the neuronal quantity in second full articulamentum is set as 512；In convolutional layer and full articulamentum, use
RELU function is as activation primitive.
4. a kind of human posture recognition method based on convolutional neural networks according to claim 3, it is characterised in that: false
If n-th of node O of m layers of convolutional layerm,nInput value be { xn-1,1；xn-1,2；…；xn-1,k, then the value of output unit is as follows:
Wherein, hw,bIt (x) is excitation value, f (x) is excitation function, xiIt is i-th of input value of node, wiIt is i-th of input value of node
Weight, k is the number of present node input value, and b is bias term；
Linear unit R ELU is corrected to be defined as follows:
F (x)=max (0, x) (2)
The pattern field function that linear unit R ELU is convolutional layer is corrected, provides nonlinear activation ability for convolutional neural networks,
And other convolutional layers will not be interfered.
5. a kind of human posture recognition method based on convolutional neural networks according to claim 4, it is characterised in that: logical
The training to convolutional neural networks is crossed, the parameter of convolutional neural networks is obtained, specifically:
If there is p sample set, they are expressed as { (x1,y1),(x2,y2),…,xp,yp), for each sample, each sample
Collection loss function is defined as follows:
Wherein, hω,b(x) it is predicted value after network training, that is, the excitation value, y are real output values, w is input value
Weight；
For whole sample set, whole sample set loss function is defined as follows:
Wherein, N is total number of plies of convolutional neural networks, skIt is -1 layer of kth of number of nodes, a is regularization coefficient, and i is i-th of section
Point, j are j-th of feature convolution kernel, and k is convolutional neural networks kth layer；For i-th of node of convolutional neural networks kth layer with
Connection weight between j-th of feature convolution kernel；
The target of convolutional neural networks model training is to minimize the value of whole sample set loss function；
Parameter more new formula is as follows:
Wherein, L is target loss function, and μ is learning rate,X is indicated to y derivation, parameter is updated using small lot method.
6. a kind of human posture recognition method based on convolutional neural networks according to claim 5, it is characterised in that: learn
Habit rate is set as 0.001, is optimized using adamoptimer algorithm.
7. a kind of human posture recognition method based on convolutional neural networks, feature exist according to claim 1 or 5
In: in the step S02, in RELU excitation function input, i.e., at the output of linear filter, sparsity is introduced, specifically
Are as follows:
The input h of kth layer RELU excitation function in the convolutional neural networkskDegree of rarefication indicate are as follows:
Wherein, hkIt is the input of kth layer RELU excitation function, S (hk) it is kth layer RELU excitation function in convolutional neural networks
Input hkDegree of rarefication；
The objective function of optimization is defined, to determine convolutional neural networks parameter, objective function is defined as follows:
E=L+ λ ∑kS(hk) (8)
Wherein, E is the objective function of optimization, and L is the objective function for being not introduced into sparsity, that is, the target loss function, λ
It is the tuner parameters for controlling degree of rarefication.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910481323.XA CN110222634B (en) | 2019-06-04 | 2019-06-04 | Human body posture recognition method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910481323.XA CN110222634B (en) | 2019-06-04 | 2019-06-04 | Human body posture recognition method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110222634A true CN110222634A (en) | 2019-09-10 |
| CN110222634B CN110222634B (en) | 2022-11-01 |
Family
ID=67819290
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910481323.XA Active CN110222634B (en) | 2019-06-04 | 2019-06-04 | Human body posture recognition method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110222634B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110688933A (en) * | 2019-09-23 | 2020-01-14 | 中国计量大学 | Novel convolutional neural network and weighted assignment human body posture estimation algorithm |
| CN110688980A (en) * | 2019-10-12 | 2020-01-14 | 南京工程学院 | Human body posture classification method based on computer vision |
| CN110929242A (en) * | 2019-11-20 | 2020-03-27 | 上海交通大学 | Method and system for carrying out attitude-independent continuous user authentication based on wireless signals |
| CN111178237A (en) * | 2019-12-27 | 2020-05-19 | 上海工程技术大学 | Pavement state recognition method |
| CN111753736A (en) * | 2020-06-24 | 2020-10-09 | 北京软通智慧城市科技有限公司 | Human body posture recognition method, device, equipment and medium based on packet convolution |
| CN111950496A (en) * | 2020-08-20 | 2020-11-17 | 广东工业大学 | Identity recognition method for masked person |
| CN112801176A (en) * | 2021-01-25 | 2021-05-14 | 西安交通大学 | Deep learning identification method for helicopter flight attitude imbalance data |
| CN113269047A (en) * | 2021-04-29 | 2021-08-17 | 江苏大学 | Three-dimensional human body posture estimation method based on convolutional neural network and spark |
| CN113850565A (en) * | 2021-09-24 | 2021-12-28 | 广东诚誉工程咨询监理有限公司 | Maturity model-based overall process consultation project management monitoring system and method |
| CN117649631A (en) * | 2024-01-29 | 2024-03-05 | 广州宇中网络科技有限公司 | Client image processing method and system based on improved convolutional neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106096506A (en) * | 2016-05-28 | 2016-11-09 | 重庆大学 | Based on the SAR target identification method differentiating doubledictionary between subclass class |
| CN106529442A (en) * | 2016-10-26 | 2017-03-22 | 清华大学 | Pedestrian identification method and apparatus |
| CN107169117A (en) * | 2017-05-25 | 2017-09-15 | 西安工业大学 | A kind of manual draw human motion search method based on autocoder and DTW |
| CN109731302A (en) * | 2019-01-22 | 2019-05-10 | 深圳职业技术学院 | Athletic posture recognition methods, device and electronic equipment |
-
2019
- 2019-06-04 CN CN201910481323.XA patent/CN110222634B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106096506A (en) * | 2016-05-28 | 2016-11-09 | 重庆大学 | Based on the SAR target identification method differentiating doubledictionary between subclass class |
| CN106529442A (en) * | 2016-10-26 | 2017-03-22 | 清华大学 | Pedestrian identification method and apparatus |
| CN107169117A (en) * | 2017-05-25 | 2017-09-15 | 西安工业大学 | A kind of manual draw human motion search method based on autocoder and DTW |
| CN109731302A (en) * | 2019-01-22 | 2019-05-10 | 深圳职业技术学院 | Athletic posture recognition methods, device and electronic equipment |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110688933A (en) * | 2019-09-23 | 2020-01-14 | 中国计量大学 | Novel convolutional neural network and weighted assignment human body posture estimation algorithm |
| CN110688980B (en) * | 2019-10-12 | 2023-04-07 | 南京工程学院 | Human body posture classification method based on computer vision |
| CN110688980A (en) * | 2019-10-12 | 2020-01-14 | 南京工程学院 | Human body posture classification method based on computer vision |
| CN110929242A (en) * | 2019-11-20 | 2020-03-27 | 上海交通大学 | Method and system for carrying out attitude-independent continuous user authentication based on wireless signals |
| CN111178237A (en) * | 2019-12-27 | 2020-05-19 | 上海工程技术大学 | Pavement state recognition method |
| CN111753736A (en) * | 2020-06-24 | 2020-10-09 | 北京软通智慧城市科技有限公司 | Human body posture recognition method, device, equipment and medium based on packet convolution |
| CN111950496A (en) * | 2020-08-20 | 2020-11-17 | 广东工业大学 | Identity recognition method for masked person |
| CN111950496B (en) * | 2020-08-20 | 2023-09-15 | 广东工业大学 | Mask person identity recognition method |
| CN112801176A (en) * | 2021-01-25 | 2021-05-14 | 西安交通大学 | Deep learning identification method for helicopter flight attitude imbalance data |
| CN112801176B (en) * | 2021-01-25 | 2022-12-06 | 西安交通大学 | Deep learning identification method for helicopter flight attitude imbalance data |
| CN113269047A (en) * | 2021-04-29 | 2021-08-17 | 江苏大学 | Three-dimensional human body posture estimation method based on convolutional neural network and spark |
| CN113269047B (en) * | 2021-04-29 | 2024-03-22 | 江苏大学 | Three-dimensional human body posture estimation method based on convolutional neural network and spark |
| CN113850565A (en) * | 2021-09-24 | 2021-12-28 | 广东诚誉工程咨询监理有限公司 | Maturity model-based overall process consultation project management monitoring system and method |
| CN117649631A (en) * | 2024-01-29 | 2024-03-05 | 广州宇中网络科技有限公司 | Client image processing method and system based on improved convolutional neural network |
| CN117649631B (en) * | 2024-01-29 | 2024-04-05 | 广州宇中网络科技有限公司 | Client image processing method and system based on improved convolutional neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110222634B (en) | 2022-11-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110222634A (en) | A kind of human posture recognition method based on convolutional neural networks | |
| CN104361363B (en) | Depth deconvolution feature learning network, generation method and image classification method | |
| CN106560848B (en) | Novel neural network model for simulating biological bidirectional cognitive ability and training method | |
| CN109829541A (en) | Deep neural network incremental training method and system based on learning automaton | |
| CN109597891A (en) | Text emotion analysis method based on two-way length Memory Neural Networks in short-term | |
| CN104239897B (en) | Visual feature representing method based on autoencoder word bag | |
| CN107609460A (en) | A kind of Human bodys' response method for merging space-time dual-network stream and attention mechanism | |
| CN106570477A (en) | Vehicle model recognition model construction method based on depth learning and vehicle model recognition method based on depth learning | |
| CN106326899A (en) | Tobacco leaf grading method based on hyperspectral image and deep learning algorithm | |
| CN106919920A (en) | Scene recognition method based on convolution feature and spatial vision bag of words | |
| CN106951858A (en) | A kind of recognition methods of personage's affiliation and device based on depth convolutional network | |
| CN106909938A (en) | Viewing angle independence Activity recognition method based on deep learning network | |
| CN109255340A (en) | It is a kind of to merge a variety of face identification methods for improving VGG network | |
| CN106709482A (en) | Method for identifying genetic relationship of figures based on self-encoder | |
| CN109711426A (en) | A kind of pathological picture sorter and method based on GAN and transfer learning | |
| CN108229503A (en) | A kind of feature extracting method for clothes photo | |
| CN106980830A (en) | One kind is based on depth convolutional network from affiliation recognition methods and device | |
| Feng et al. | Bag of visual words model with deep spatial features for geographical scene classification | |
| CN110490242A (en) | Training method, eye fundus image classification method and the relevant device of image classification network | |
| CN106980831A (en) | Based on self-encoding encoder from affiliation recognition methods | |
| CN108154156A (en) | Image Ensemble classifier method and device based on neural topic model | |
| CN112016601A (en) | Network model construction method based on knowledge graph enhanced small sample visual classification | |
| CN110096976A (en) | Human behavior micro-Doppler classification method based on sparse migration network | |
| CN109165698A (en) | A kind of image classification recognition methods and its storage medium towards wisdom traffic | |
| CN113159067A (en) | Fine-grained image identification method and device based on multi-grained local feature soft association aggregation |
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 | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Li Jian Inventor after: Zhang Yuan Inventor after: Luo Ying Inventor after: Zhang Yixin Inventor before: Zhang Yuan Inventor before: Li Jian Inventor before: Luo Ying Inventor before: Zhang Yixin |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |